Ask Unix Co-Creator Rob Pike

Today we return to our Slashdot interview roots with a "Call for questions" for Rob "Commander" Pike, who has been involved in the development of many modern programming concepts, GUI advances, character sets, and operating systems. We'll email 10 - 12 of the highest-moderated questions to Rob and post his answers as soon as he gets them back to us.

Comments

How do I

[techsoldaten]

How do I get to my C:\ drive on a Unix box?

M

Re:How do I

[noselasd]

ln -s /mnt/whatever /c\:

Re:How do I

[viva_fourier]

And, can you put an AOL icon on Unix somewhere, so that I can access the internet?

Re:How do I

[techsoldaten]

Yeah, and how about putting Bonzai Buddy and my Comet Cursor in Unix too?

M

Re:How do I

[Eric_Cartman_South_P]

How do I get to my C:\ drive on a Unix box?

VMWare!

Re:How do I

[Anonymous Coward]

make mount
make install

Plan9

[mirko]

Plan 9 was supposed to go even further than Unix went, does it really looks like to you that it's been conceived according to a similar approach ???

Re:Plan9

[Spyffe]

Rob,

Right now, there are a large number of research kernels. Plan 9, Inferno, AtheOS, Syllable, K42, Mach, L4, etc. all have their own ideas about the future of the kernel. But they all end up implementing a POSIX interface because the UNIX userland is the default.

The kernel space needs to be invigorated using a new userland that demands new and innovative functionality from the underlying system. Suppose you were to design a user environment for the next 30 years. What would the central abstractions be? What sort of applications would it support?

Re:Plan9

[Anonymous Coward]

It would most probably support the STFU filesystem.

Re:Plan9

[Anonymous Coward]

AtheOS and Syllable aren't research operating systems. Second, Inferno doesn't implement a POSIX interface. Third, the POSIX emulation support in Plan 9 isn't used much (Thankfully). Finally, I suggest you actually check out Plan 9 if you want to see what the environment of the next 10 years (not 30, I don't want to see it around that long) should look like. Don't just flash the Plan 9 to try and swoon Rob.

Re:Plan9

[Anonymous Coward]

I suggest you check out an old Lisp Machine for a better environment.

Re:Plan9

[serialhex]

ooh... yes, this ones real good

Re:Plan9

[Spyffe]

I'm not trying to "swoon" Rob. I was merely illustrating the fact that there are a ton of kernels out there. I'll grant that Plan 9 does implement a new userland, but would also say that it is probably not the one that's going to be driving OS research in the future.

The reason is that Plan 9's strength (from what I've seen) is presentation of data on a finite, fixed-size collection of known computers. The resources of these computers are made virtually indistinguishable in a very elegant manner.

The cluster, however, doesn't seem to be the foundation of the user computing experience nowdays. (Granted, research labs and company server farms use them, but they are not the driving force of user-space innovation, generally preferring highy specialized management tools.) For the average user, the great computing resources are:

1. their personal, usually single-processor, workstation,
2. the Internet, an amorphous mass of processors offering widely varying services and capabilities, and possibly
3. portable devices of various sorts.

What I'm interested in is how Rob thinks the userland will shift to adjust to these realities.

What do you think of the SCO vs IBM?

[johansalk]

Gotta ask the question :-)

Re:What do you think of the SCO vs IBM?

[stecoop]

Add to the question about what he thinks of the government forcing Bell to sell of the Unix OS (because the parent company was considered a monopoly) inlight of today's litigation wrangling.

Programming language

[Rosco P. Coltrane]

Dear Rob,

One question: do you like programming in Pike?

The Microsoft timecapsule

[Anonymous Coward]

Why make unix when you already had Windows?

</joke>

Biggest problem with Unix

[akaina]

Recently on the Google Labs Aptitude Test there was a question: "What's broken with Unix? How would you fix it?"

What would you have put?

Re:Biggest problem with Unix

[Anonymous Coward]

You should check the plan9 faq, there is an answer there.
http://www.ecf.toronto.edu/plan9/plan9faq. html#pla n9design

basically the argument is that what was added later on (network & gui) was not done the way it should. plan 9 (unix os rework from ground 0) is the answer to this.

a.c.

Re:Biggest problem with Unix

[Anonymous Coward]

If you know what's wrong and know how to fix it, why don't you fix it already! The source is out there.

Re:Biggest problem with Unix

[Anonymous Coward]

is the source out there? where is the source for solaris 9? aix? hpux? *nix doesn't mean superior. unless there is source code for aix hpux or solaris 9, then where can i find it.

Re:Biggest problem with Unix

[El]

Well, for one thing, "creat" should have had an "e" on the end ;-)

Re:Biggest problem with Unix

[Just Some Guy]

Recently on the Google Labs Aptitude Test there was a question: "What's broken with Unix? How would you fix it?"

And along those lines, what should I list as my favorite color?

Re:Biggest problem with Unix

[hackstraw]

Recently on the Google Labs Aptitude Test there was a question: "What's broken with Unix? How would you fix it?"

I saw this google labs apt question, and while I've become numb to most of UNIX's issues and cannot think of a generic across the board (ie, cross vendor) "broken" thing except why the hell is UNIX so picky about 1) unmounting filesystems "that are in use" and 2) why the hell there is a 'D' run state that is completely uninterruptable?* The 2nd one really baffles me, and the first is just annoying, and fuser or some vendor specific tool can (sometimes) point you to the offending process that is using the filesystem. I found out today that fuser does not work on linux with the kernel NFS daemon sharing a filesystem and I try to unmount it. Annoying, but not as fundamentally broken as #2 in my opinion.

Another thing that I see as "broken" in UNIX is that there is no normal/standardized/sane way of installing software. Debian gets it the closest, but the LSB picked RPM for some insane reason for package mismanagement on Linux.

* For those that don't know, if there is something wrong with a disk subsystem, and a process tries to access that disk subsystem, the process is in an uninterruptable "disk wait state", that cannot be corrected without rebooting the computer. One can ususally safely ignore the processes stuck in this state, but its kinda irritating because it can often bring the system load up by one for each stuck process, yet it does not appear to hurt performance any.

Re:Biggest problem with Unix

[Anonymous Coward]

"What's broken with eunuchs? How would you fix them?"

Re:Biggest problem with Unix

[Snard]

Blue.... NO! Yellooooooooooooowwwwww...

Re:Biggest problem with Unix

[menkhaura]

There *is* source code for all those unices, just go ask the vendors... I don't believe they will hand it to you, though.

Re:Biggest problem with Unix

[CaptnMArk]

One thing that has been recently broken on UNIX is character encodings.

With the switch to UTF-8, all the classic text file processing tools stop being useful on binary files (like grep, ...). The 'old way' was that there is no difference between binary and text files, with the UTF-8 transition this is no longer true.

This also affects Perl (and other popular 'scripting' languages), because the lack of separation between strings and byte arrays.

On some days this makes me strongly prefer iso-8859-1.

One thing needs to be done is a cleaner separation between 'server' code that needs to know nothing at all about locales and 'client'/ui code that needs to do the UI the way locale is configured.

Re:Biggest problem with Unix

[Anonymous Coward]

Null terminated strings. strlen.

Re:Biggest problem with Unix

[CondeZer0]

I think that he and the Bell Labs folks already answered those questions over 10 years ago:

http://plan9.bell-labs.com/sys/doc/9.html
(See specially the first section: Motivation)

uriel

Re:Biggest problem with Unix

[MesiahTaz]

1) File systems that are "in use" cannot be unmounted because there may be uncommitted writes to them. When an application writes to a filesystem, the data is not immediately written to the disk for performance reasons.

2) The 'D' state does not exist because there is a "problem" with the file system. Processes are but into the D state because two processes are trying to access the same file or resource at the same time.

Re:Biggest problem with Unix

[OrangeTide]

grep works quite nicely with utf8. even an ancient version of grep. as long as it's 8-bit clean and you give it search queries in the same encoding as your file. (like 8859-1 for 8859-1 files, utf8 for utf8 files).

The whole idea of utf8 is to have the least impact on existing software. There are encoding sets that do break text editors and grep. Like some of the asian multibyte encodings.

if you grep for a utf8 sequence in a binary file it will work exactly the same as if you grep for a latin1 sequence in a binary file or a US ASCII.

Some problems with utf8 are that strlen() can no longer be used to determine how many monospaced positions a string will fill. you have to use mblen() which requires you to know the lenght of your multibyte string. (multibyte strings don't have to be null terminated in all encoding systems, utf8 happens to support it, but mblen() isn't interested in that).

binary and text files are still the same on *nix. you can't have a newline as part of a utf8 multibyte sequence for example. fgets() works on utf8 the same as it always has. The following code will work with usascii, utf8 and iso-8859-x. the code obviously is unaware of the encoding scheme, but the important parts are the same in at least thoses encodings for it to work.

#include
int main() {
char buf[1024];
while(fgets(buf,sizeof buf,stdin)) {
fputs(buf, stdout);
}
return 0;
}

ps- if you want to grep binary files I would recommend piping the binary though the "strings" command first. it can yield better results. You can specify -e if you want "strings" to sniff out alternate encoding types. (including utf16)

Re:Biggest problem with Unix

[geeber]

Given that Rob Pike works for Google, I wouldn't be surprised if he wrote that question himself...

Re:Biggest problem with Unix

[Eraser_]

I do seem to recall the classic example of this being Solaris and or Sun tar. I forget who the offending party was, but when a tape drive would error and not report it back (physical failure of some sort of the "file system") tar would never time out, merely wait forever for the write call to be returned.

This effectivly locked the tar process, that memory, and your tape drive until reboot. It does make sense for I/O to have unkillable states, so long as there are proper timeouts, probably in the kernel, for hung I/O.

Re:Biggest problem with Unix

[99BottlesOfBeerInMyF]

nother thing that I see as "broken" in UNIX is that there is no normal/standardized/sane way of installing software. Debian gets it the closest, but the LSB picked RPM for some insane reason for package mismanagement on Linux.

I think the OSX solution is by far the most elegant. Self contained packages with everything needed, drop and go.

Re:Biggest problem with Unix

[hackstraw]

I think the OSX solution is by far the most elegant. Self contained packages with everything needed, drop and go.

Tell me about it. And the beautiful part is in order to do something crazy like uninstall software, you just drag the puppy in the trash.

Re:Biggest problem with Unix

[hey]

But I've seen cases when I can't unmount a CD-ROM.
(No pending writes). Just because somebody had cd-ed to a directory in the CD-ROM. This is Linux.

I agree with the guy who started this thread, when you ask to unmount something it should unmount.

Re:Biggest problem with Unix

[hackstraw]

1) File systems that are "in use" cannot be unmounted because there may be uncommitted writes to them. When an application writes to a filesystem, the data is not immediately written to the disk for performance reasons.

I don't care. I'm unmounting a filesystem for a reason, and I'm going to kill the offending processes anyway. I'm root dammit! Heh.

Also, its just inconsistant that I cannot unmount a filesystem, but I can do rm -rf on the entire filesystem. Another thing that gets me is that a filesystem can be "in use" just because an app is using it as its current working directory. No open files, no uncommited writes, just because some program is sitting there it can't be unmounted, but again I can rm -rf it. (Yes, I do know that rm -rf only removes the linkcount by one, and the files might actually really still be there and not there at the same time.)

2) The 'D' state does not exist because there is a "problem" with the file system. Processes are but into the D state because two processes are trying to access the same file or resource at the same time.

Thats not true. A 'D' state is when an app is waiting for a disk or tape, it has nothing to do with filesystems, its a device issue. Regardless, I should be able to kill the process vs having to reboot to get rid of it. There should be no software reason for rebooting a computer besides a fundamental OS change.

Re:Biggest problem with Unix

[mikefe]

Actually, the refuse to umount if there are users in the mount is part of the POSIX or SUS specifications.

Linux does not have a problem with it. That's why it has the -l option.

-l Lazy unmount. Detach the filesystem from the filesystem hierarchy now, and cleanup all references to the filesystem as soon as it is not busy anymore. (Requires kernel 2.4.11 or later.)

Re:Biggest problem with Unix

[mikefe]

And you end up using more memory because of this simplicity because your programs can now have multiple copies of the same library in memory if they have any libraries in common.

Re:Biggest problem with Unix

[evil_engin33r]

Who cares--memory and hdd space are cheap.

Re:Biggest problem with Unix

[Anonymous Coward]

Except if the software includes a device driver, like the Palm desktop software does (to talk to the Palm, which is a USB device). In that case, it's no longer all in one place, and uninstalling it is just as much of a pain as similar stuff on other systems. :-(

Re:Biggest problem with Unix

[runderwo]

And the beautiful part is in order to do something crazy like uninstall software, you just drag the puppy in the trash.

What's even more beautiful is that a virus can now infect your software because it's all installed in single folders under your user account. Bring on the Windows virus ports!

Re:Biggest problem with Unix

[Anonymous Coward]

Another thing that I see as "broken" in UNIX is that there is no normal/standardized/sane way of installing software. Debian gets it the closest, but the LSB picked RPM for some insane reason for package mismanagement on Linux.

For a start: I'd have kicked your stupid ass right out of the interview for comparing "Debian" (by which you mean both apt and dpkgs) and RPM. RPM is a package format -- you need a management tool to go on top of it, and you can jsut as well use apt (or yum) for that purpose. Fuckwits like you continue to confuse the issue -- when, in fact, RPM as a *packaging* format is considerably superior to dpkgs.

Re:Biggest problem with Unix

[julesh]

There are a number of features of grep that don't work with utf8 encoded files, for instance the '-i' command line option will not correctly detect all case conversions for non-ASCII characters, there are whitespace characters that aren't recognised as such (thus confusing the -w command line option, and potentially -x also). As it has no idea of what constitutes or doesn't constitute a single character, the [^...] regexp pattern does not work correctly. Neither does [x-x] if the second character is not in ASCII. It'll also misinterpret some other similar expressions (eg 'x?') where the character specified is non-ASCII.

Sure, if all you're using it for is matching strings in a file, grep works with utf8. But not all of its features do.

Re:Biggest problem with Unix

[julesh]

Which would make his answer even more interesting, if you ask me.

Re:Biggest problem with Unix

[jovlinger]

aha!

So this is how to reboot the client when a mounted NFS server has gone down.

Re:Biggest problem with Unix

[spitzak]

You seem quite confused. The whole point of utf-8 is that things like -i and -w DO work. -i will still ignore case of Ascii characters. -w will still treat space and tab as white-space, and there are very good reasons to not assign "whitespace" meaning to any non-Ascii characters, so it really works perfectly.

Any encoding that is not in bytes would break grep completely, so it would be worse. Any encoding where Ascii letters do not represent themselves would break -i and -w completely, so this is better.

I mean really, I challenge you to suggest anything that would be better than utf-8!

Re:Biggest problem with Unix

[kahei]

It's quite telling that you pick two obscure, nitpicky issues while ignoring vastly bigger things such as ASCII dependency and the lack of a component system.

Re:Biggest problem with Unix

[Anonymous Coward]

Keep in mind that one of the great things about Linux is how well it scales to smaller systems as well as outdated ones. If you're going to support needlessly wasting memory or hard drive space, you mean to be in the Microsoft fanboy camp.

Re:Biggest problem with Unix

[hey]

The man page says this:

-f Force unmount (in case of an unreachable NFS system). (Requires kernel 2.1.116 or later.)

Re:Biggest problem with Unix

[ticktockticktock]

Not necessarily. On a mac, you could drag an app to a folder that is accessible to all but not writeable by all that normal users can access and run, but not modify. Also, being able to modify existing programs isn't the only way of being affected by malware. If you can send emails, so can malware that you may accidentally run. If you can delete your own documents, so can any malware that runs with the same privileges as you.

Re:Biggest problem with Unix

[defile]

Recently on the Google Labs Aptitude Test there was a question: "What's broken with Unix? How would you fix it?"

The moment I saw that question I said it must be a trick. UNIX develops by evolution, not by dictation. Whenever an individual change is dictated it almost never survives on its merits.

UNIX is beyond the comprehension of any one. One can introduce a change, but it is up to NATURAL LAW to ultimately decide if the change lives or dies.

That said, pttys, fifos and ioctls do in fact blow.

Re:Biggest problem with Unix

[Coniptor]

With regard to 1 and fuser not working with the kernel NFS daemon.
I don't know that it would work but you might try lsof.

As far as 2. Yeah.... Yeah, that is REALLY annoying.

Re:Biggest problem with Unix

[WWE-TicK]

We used to do this back in the MS-DOS days, minus the "drop and go" part since it didn't have a standard GUI.

Re:Biggest problem with Unix

[Pentagram]

What is needed is a decent error report. Often I've had a device refuse to umount for some non-obvious reason and has taken a bit of searching to work out why. "Can't unmount, device busy" is not very helpful -- it should tell me what the problem is, e.g. Can't unmount because pwd is /mnt/foo in terminal bar.

I do know about lsof but this is a very basic user interface problem.

Re:Biggest problem with Unix

[AuMatar]

Great, so is it cheap enough you'll pay for mine? No, then I'll stick with dynamic libraries please.

Not to mention the upgrade problems- if libc has a problem, I update 1 copy, problem solved. If you have a dozen versions floating around for each program, updating something for a security bug or the like is a nightmare.

Re:Biggest problem with Unix

[mikefe]

Before you start adding a lot of extra error reporting, it should be designed in such a way that it will be extendable in the future so that you don't end up with an unmanagable error reporting system.

Re:Biggest problem with Unix

[boots@work]

Ha ha.

But seriously, creat(2) should have been deleted altogether, as it's redundant with open(2). I'd prefer to have rmdir(2) go away too, instead having unlink(2) suceed on empty directories.

Of course it's too late now.

Re:Biggest problem with Unix

[stoborrobots]

Aren't enunuchs already fixed???!? :-)

Re:Biggest problem with Unix

[bani]

it doesnt work though. try it (kernel 2.6.8.1). it just hangs like everything else.

Re:Biggest problem with Unix

[bani]

Another thing that gets me is that a filesystem can be "in use" just because an app is using it as its current working directory. No open files, no uncommited writes, just because some program is sitting there it can't be unmounted, but again I can rm -rf it. (Yes, I do know that rm -rf only removes the linkcount by one, and the files might actually really still be there and not there at the same time.)

however this lets you do neat and useful things, like recovering files after they've been deleted (but while the process still has them open). cd to /proc, and cat out the fd's back into files. saved my ass a number of times.

Re:Biggest problem with Unix

[Anonymous Coward]

This one is very serious:

The story goes that when Dennis Ritchie was asked what he would do differently if he had to do Unix over, he said "I would spell 'creat' with a second 'e'."

OK, here's the obligatory

[w.p.richardson]

Looking back, what would you have done differently? Anything?

Re:OK, here's the obligatory

[Rosco P. Coltrane]

Hmm let's see: hiring a hitman to get a young hippy programmer working at MIT AI Lab in 1976 before he even started to type "vi /home/richard/great_new_projects_I_have/emacs/main .c" would have been nice... That would have saved 2 generations of programmers early onsets of carpal tunnel...

Re:OK, here's the obligatory

[Anonymous Coward]

Emacs keys are not that bad once you switch the caps lock and control keys. A more serious question is, what's the deal with the emacs haters on Slashdot?

Re:OK, here's the obligatory

[niteice]

You've never used emacs, have you? It is the most bloated, arcane, and BLOATED program this side of Microsoft.

Re:OK, here's the obligatory

[Anonymous Coward]

Tell that to the scores of coders who can no longer type becuase of the damge to their hands and wrists.

Re:OK, here's the obligatory

[Anonymous Coward]

Oh come on, emacs is a fantastic operating system. The only improvement would be if it came with a good editor.

Re:OK, here's the obligatory

[peawee03]

Has anybody ever tried to hack it so Emacs is the OS, as in "Emacs/Linux" vs. "GNU/Linux"?

Re:OK, here's the obligatory

[stoborrobots]

ummm.... perhaps

boot: linux init=/bin/emacs

resolv.conf

[Flashbck]

why was the 'e' ever removed from resolv.conf?!!?!?

Re:resolv.conf

[tchuladdiass]

Probably to match up with the "creat" system call

Re:resolv.conf

[JUSTONEMORELATTE]

Ahh, you see, UNIX was created back in the day when we couldn't afford vowels, and many consonants.
As such, common commands had to be shortened a little bit:
"Copy" becomes "cp"
"List" becomes "ls"
"Rename" becomes "mv" and so on.

Re:resolv.conf

[Anonymous Coward]

why was the 'n' ever removed from umount?

Re:resolv.conf

[roalt]

Why was the 'e' ever removed from resolv.conf?!!?!?

That was bcaus the -ky on my kyboard was not working as it should in thos days.

Re:resolv.conf

[Anonymous Coward]

That's probably be better posed to one of the BIND developers..

I think the resolv.conf file came from therre.

Re:resolv.conf

[GoofyBoy]

This must be the most awkward question you can give a programmer.

Can you justify why you named your objects/variables/files years ago?

Re:resolv.conf

[Anonymous Coward]

Is there some sort of 11 char limit? resolv.conf is exactly 11 chars.

Re:resolv.conf

[orangesquid]

Short commands are easier to type, especially on slow TTL hardware where several context switches have to happen for every single keystroke.

Early unix also had a 12-letter filename limit.
I don't know if that included the NUL or not; if it did, then resolv.conf makes sense, since you might want to make a backup copy named resolv.conf~ or such. Also, early fortran had a 6-letter symbol name limit; this might be the reason for creat (so _CREAT would fit within the maximum 6 letters?)

DISCLAIMED: Just some ideas, dunno if any of this is correct!

Re:resolv.conf

[Greyfox]

I usually just blame it on the drugs. It was the 70's after all...

Re:resolv.conf

[Trolling4Dollars]

Becauth thilly!!!! It'th not a vowel!!! ;P

Re:resolv.conf

[orangesquid]

see here -- 6 letter limit in early fortran, perhaps "UMOUNT" was preferred over "UNMUNT" or "UNMOUN"?

No clue if this is correct, but its a thought...

Re:resolv.conf

[AJWM]

That would explain the problem with the creat(2) system call too.

Oh wait, that would be crat(2).

Re:resolv.conf

[Mitchell Mebane]

That was bcaus the -ky on my kyboard was not working as it should in thos days.

No, I think it's just picky.

Re:resolv.conf

[metlin]

That was bcaus the -ky on my kyboard was not working as it should in thos days.

Looks to me like it's got a mind of it's own.

You do realize that the has an e, right? :-p

Re:resolv.conf

[peragrin]

>>Rename" becomes "mv" and so on.

Close.

mv = Move. to move files.

when you move files by changeing the name and not the directory you rename them.

Re:resolv.conf

[red floyd]

14 letter. Each directory entry was 16 bytes -- 2 bytes for inode, 14 for filename.

Re:resolv.conf

[pixelbeat]

For the same reason his email address
is r@google.com rather than rob@google.com
or God forbid rob.pike@google.com

Pity squatters already have ggle.com

Re:resolv.conf

[Anonymous Coward]

That would be the whole point of the joke.

Re:resolv.conf

[Anonymous Coward]

UNIX was created back in the day when we couldn't afford vowels

Even then, you could buy a vowel for $250.

Re:resolv.conf

[AJWM]

The early unix filename limit was 14 chars. That gave you two bytes for the inode number in a 16-byte directory entry.

Early C also had the 6 letter limit, not so much in the language itself as a limitation in early linkers/loaders, which only distinguished the first few characters of an identifier. (So for sanity's sake you limited the identifiers in your program to that so you would accidentally get a collision. This is one reason that old C code looks, well, old. (And why some old C coders still tend toward using short variable names.)

Re:resolv.conf

[Anonymous Coward]

Worst of them all: umount

Re:resolv.conf

[the chao goes mu]

UNMNT? Seems a little clearer. Especially as u- is vague, meaning both "un-" and "user-" (umount/umask v. ulimit)

Re:resolv.conf

[MarkGriz]

That was bcaus the -ky on my kyboard was not working as it should in thos days.

If you're getting KY on your keyboard, maybe you should look for a girlfriend instead :-)

Re:resolv.conf

[telstar]

"Ahh, you see, UNIX was created back in the day when we couldn't afford vowels, and many consonants.

• I blame that bitch Vanna...

Re:resolv.conf

[TheRaven64]

And those of us who come from a CP/M background still type cat when we mean ls on occasion...

Re:resolv.conf

[asdfghjklqwertyuiop]

resolv.conf is exactly 11 chars

exactly 11 characters? As opposed to filenames 11.0001 characters long?

Re:resolv.conf

[TALlama]

I never knew that Unix was designed on an episode of Wheel of Fortune.

Re:resolv.conf

[Kehvarl]

so does resolv.conf :]

Re:resolv.conf

[ohad_l]

It would seem that the developers of Unix had a lot of fun trying to skew the statistics so that e would not be the most common English letter.

Re:resolv.conf

[Alioth]

The lack of syntactic sugar in the Unix commands is still very useful, even if those old TTYs have long since disappeared.

I have a mobile phone with GPRS. It has an SSH client. The short Unix commands mean my phone is a useful tool if I have to try and fix a box when I'm on the move. Excessively verbose commands would suck on that tiny keyboard and display.

Re:resolv.conf

[akaina]

Actually "Move" becomes mv.

Even assembly programmers get MOV... Unix gurus are just saddists.

Re:resolv.conf

[chris_mahan]

you mean Vnn right?

Re:resolv.conf

[PhilipPeake]

Actually, C had much longer variable names, it was external symbols which were limited to 6 characters, so it only made sense for globals to be limited to 6 characters. this limit was not imposed by C, but by the DEC linker. DEC tried very hard to get the 6 character limit for global symbols integrated into the C standard to avoid having to "break" their linker. Fortunately, they failed.

Re:resolv.conf

[Anonymous Coward]

why was the 'e' ever removed from resolv.conf?!!?!?

Unix didn't have TCP/IP for a long, long time after it was created.

Re:resolv.conf

[Suppafly]

"Rename" becomes "mv" and so on.

rofl.

Re:resolv.conf

[noz]

Rob Pike's famous response to the obligatory "What would you have done differently" question (asked above) was: "I would have left the e on creat."

Re:resolv.conf

[JUSTONEMORELATTE]

Thank you. I was wondering if anyone was going to get the joke. I've already had two (very kind, helpfull) folks correct my "mistake."

Re:resolv.conf

[mmatloob]

The e was also removed from the creat() system call.

" Ken Thompson was once asked what he would do differently if he were redesigning the UNIX system. His reply: "I'd spell creat with an e." "
- The unix Programming Environment Page 204

Re:resolv.conf

[AJWM]

Yes, I believe I said that, although perhaps my phrasing was too subtle.

And it wasn't just the DEC linker. I had the problem rear its ugly head on other Version 7 based systems, on code that had run fine on a VAX (running BSD).

Re:resolv.conf

[dkf]

Ah, this reminds me of scientific FORTRAN use with the NAG library. To use an operation, you had to go to your (paper) documentation and look up what function you wanted to perform (according to what types you were manipulating, what numerical methods you wanted to use, how much memory you had available, what the phase of the moon was, etc.) and found out what its magic symbol was. That was the function name to use. They were utterly un-mnemonic IIRC.

For modern coders, just think on this: it was worse than the Win32 API. Well, most of Win32; there are some seriously scary corners in there...

Re:resolv.conf

[rsc9]

This is a stupid question. Rob had nothing to do with resolv.conf.

Re:resolv.conf

[Theatetus]

Unix didn't have TCP/IP for a long, long time after it was created.

I think you mean creatd

Re:resolv.conf

[Anonymous Coward]

This was done because you can pack 6 letters in 32 bits (and still have two full bits free)

Apple and Unix

[Anonymous Coward]

What are your thoughts on Apple's use of Unix.

Re:Apple and Unix

[Build6]

I would have modded this up except that it was at 5 already.

particularly considering apple's own long history in OS development being abandoned (well, of course, there's the corporate-maneuvering angle what with NeXT and Jobs)... .

Actually, the alternative to NeXT at that time was Be - and, I dunno, BeOS though not UNIX per se, *was* pretty UNIX-y - ls, ps, they were all there.

Re:Apple and Unix

[The Phantom Buffalo]

Apple doesn't use Unix.

Re:Apple and Unix

[sgant]

Apple doesn't use Unix

Care to enlighten us on what exactly it is you mean? Since OSX sits on top of a BSD varient...and BSD to the best of my knowledge is a true "Unix"...what do you mean by Apple not using Unix?

Just wondering.

Re:Apple and Unix

[Anonymous Coward]

I would have modded this up except that I'm an Anonymmous Coward.......doh!

Re:Apple and Unix

[The Phantom Buffalo]

Apple hasn't has OSX certified with The Open Group. To my knowledge, there is a lawsuit over this.

Re:Apple and Unix

[Brandybuck]

Legal nitpicking. To quote a recently departed master of the obvious, if it walks like a duck and quacks like a duck, it's probably a duck. In the case of OSX, not only does it strongly resemble a duck, it even has a sign around it's neck saying it is a duck! OSX has a Unix pedigree that many certified UNIX(tm) systems do not have.

Re:Apple and Unix

[The Phantom Buffalo]

Do you consider all trademark law to be legal nitpicking?

Re:Apple and Unix

[OmniVector]

unix is a generic term now a days. it means a lot of things, but in the general consensus most people agree that mac os x can be considered a unix based OS, despite whatever the technical definition of "unix" is.

Re:Apple and Unix

[Brandybuck]

Not all trademark law, but parts of it I do. When a word or symbol becomes associated in the public's mind with an entire class of product, then it should cease being a trademark. It's not just Unix, it's "aspirin", "kleenex", "jello", etc.

When a word ceases to be a proper name and becomes a noun, it should lose its trademark. Regardless of the cause. Regardles of how well you tried to protect it in the past. In the case of Unix, it has ceased to be the proper name for a specific operating system.

The problem with the Unix trademark was that it was allowed to be attached to too many products. Not only did it become a generic term, it became a generic term because AT&T/USL/Novell/OG allowed it to become generic.

Re:Apple and Unix

[The Phantom Buffalo]

The Open Group filed a lawsuit against Apple over use of the Unix trademark, which means they are defending it. You and I can debate it all day long, but it is up to the courts to decide.

Re:Apple and Unix

[socode]

The trademark is irrelevant, since this discussion is about lineage.

Re:Apple and Unix

[UnixSphere]

So according to your statement, a company who makes unofficial chevy parts can tell their customers that they are genuine chevy parts because "It walks like a duck".

The future

[johansalk]

How do see the future of Unix and its offspring OSes?

Re:The future

[DemoLiter3]

Will operating systems stick with current user/kernel/process/thread principles, or will there be some other concept to allow higher parallelization and/or different security model in the future?

The future?

[xenostar]

What do you see in the far future of operating systems, now that great advances in the way we think about computers, such as quantum computing, have been made.

Is Linux "unix"?

[Doc Ruby]

Is Linux "unix"? What did Unix get wrong, but was too late to change by the time that you realized it, that Linux can still get right, while it's still young?

Re:Is Linux "unix"?

[byolinux]

Given that Linux is the kernel used on GNU/Linux, and given that GNU's Not Unix, I'd say not.

Re:Is Linux "unix"?

[anonymous cowherd (m]

Followup to this: If Linux is not Unix, in what ways is it better or worse than Unix?

Re:Is Linux "unix"?

[Anonymous Coward]

Linux is not unix due to differences in some system calls and under the hood stuff that it does differently. Linux is not able to pass the certification process because of this, and there is generally no need to change it to fall in line just to receive the UNIX appelation.

Re:Is Linux "unix"?

[Doc Ruby]

You'd say that if you were Richard Stallman, but you might be kidding, or wrong. What does Rob Pike say?

Re:Is Linux "unix"?

[Doc Ruby]

That's why I made the subtle distinction in my post beween "unix" and "Unix" (and your "UNIX"). Especially with the SCO scams hanging above our necks, I'm interested in what Pike says is "unix". Perhaps it transcends the code, the trademark, the APIs, and has become a style, or something in between.

Re:Is Linux "unix"?

[Naikrovek]

Linux is definitely not UNIX, but it is very UNIX-like. Linux was written from scratch, and is not recognized as a UNIX by the open group (click).

My understanding is that unless someone wants to pony up money to have Linux reviewed, and potentially be certified as UNIX, it won't ever be a true UNIX. There are POSIX compliance tests, and lots of little standards that have to be implemented and be implemented correctly.

so, no. Linux is not UNIX.

Re:Is Linux "unix"?

[Naikrovek]

www.unix.org says what is and what is not unix. they own the trademark to "unix", "Unix", and "UNIX".

Re:Is Linux "unix"?

[Ithika]

So what you're saying is that:
a) Linux != GNU
b) GNU != Unix
therefore
c) Linux != Unix

What??

Re:Is Linux "unix"?

[Anonymous Coward]

Haven't they learned you in the Axiom of Transitivity of Inequalities? ;-) It renders a lot of math quite trivial once you start applying it.

Re:Is Linux "unix"?

[the chao goes mu]

But what about "Eunuchs(tm)"? (and EUNUCHS(tm) and eunuchs(tm))

Re:Is Linux "unix"?

[Jhan]

It is Unix if you can look at a vastly unfamiliar GUI shell (or a vastly strange file system), click around a bit and still see he common points and say...

It's a Unix system... I know this!

Re:Is Linux "unix"?

[Doc Ruby]

Does that make your post An Adventure 65 Million Years In The Making?

Re:Is Linux "unix"?

[JabberWokky]

No, they say what is and what is not legally considered Unix. The popular definition includes many other operating systems that are similar.

Tomatos are legally a vegetable[1], scientifically a fruit, and considered by many people to be either one. Champaign is technically only from a certain region in France, in American common usage it is just about any sparkling wine. Terms can often be approached legally or semantically. "Unix" in this case is short for "that OS that you created", not "an OS certified by the OpenGroup". If they dubbed, say, MS-DOS 4.01 Unix, it may be legally Unix but most people wouldn't consider it "real Unix".

[1] In America, for tariff reasons. Decision of the Supreme Court in late 1800s.

--
Evan

Re:Is Linux "unix"?

[byolinux]

GNU is a project to create a free Unix-like but non-Unix copy OS.

Linux is a kernel that powers GNU.

Therefore, Unix as an OS is different from Linux as a kernel.

Is anyone denying Linux is anything more than a kernel, now?

Re:Is Linux "unix"?

[stm2]

How much that the test actually cost?

Re:Is Linux "unix"?

[stoborrobots]

I wondered too... It took a fair bit of poking around on the Open Group's site to find this:
http://www.opengroup.org/openbrand/Brandfees.htm

It seems that "Linux" cannot get certified... or at least, not easily.

The tests themself are mostly available free - but the certification costs around $4000 + percentage of revenue.

Now I don't know about Linus himself, but there are a fair few companies (Redhat, SuSE, Debian) who would have to each shell out to sizeable chunks to be certified... Seems like a bit much to me...

Re:Is Linux "unix"?

[JoeBuck]

The US was not asked to sign the international treaty protecting the use of place names on wine, because it was during Prohibition. It really seems kind of funny that the US ignores naming rights in this way, though, considering how tough the laws are otherwise about trademarks.

See, "Champagne" should really be thought of as a trademark belonging to the wine-growers of the Champagne region of France, and almost everyone but the US sees it that way. It's kind of like letting a foreign company making knock-off jeans and calling them "Levi's".

Re:Is Linux "unix"?

[Doc Ruby]

How could my post be "Redundant", when no other post prior to it asked if Linux is "unix"? Or even the specific version of "what would you do differently?" that it includes. Spoilsport mod.

Are you suprised at the longevity of Unix?

[sgant]

When you were creating it, did you in your wildest dreams ever think that 30 years people would still be using it on a daily basis? Was it designed from the beginning to grow and be added onto?

Re:Are you suprised at the longevity of Unix?

[Anonymous Coward]

Hey man, all I really wanted was to play Spacewar...

Where do u think Linux is headed?

[Anonymous Coward]

Do u beleive after 10 years, there will be Liux & Windows only (like Bill Gates)?? Will Linux dominate & will Microsoft vanquish into the night!

Re:Where do u think Linux is headed?

[roalt]

Do u beleive after 10 years, there will be Liux & Windows only (like Bill Gates)?? Will Linux dominate & will Microsoft vanquish into the night!

Answer: no...

I think Liux and Widows will be the only 2...

Re:Where do u think Linux is headed?

[Anonymous Coward]

OK, who let the AOLer in?

Re:Where do u think Linux is headed?

[Anonymous Coward]

I don't know about Windows but Linux will be there

Question for Rob

[JUSTONEMORELATTE]

Hi Rob,
Based on your tremendous impact to the history of computing, I really can only think of one question for you.

Wouldn't you like to get a Totally Free iPod?!?!?!?

Re:Question for Rob

[Anonymous Coward]

This is off-topic not redundant. Who has some meta-mod points??

Unix co-creator?

[hruntrung]

Pardon my ignorance, but I was under the impression that Dennis Ritchie and Ken Thompson created Unix (Unics), on disused hardware at Bell Labs.

Am I incorrect in this belief? Someone, kindly, clarify the matter.

Re:Unix co-creator?

[Fred Foobar]

Rob Pike invented, or at least help develop the idea of, the pipe and command pipelines in Unix. It was a fairly major development, as it allows programs to act as filters.

He may have developed other stuff in Unix too, but that's all I know about him.

Would use use OO?

[esanbock]

Were you to start again today, would use use objects instead of an API to build the core OS components?

Re:Would use use OO?

[SewersOfRivendell]

I think you mean: Were you to start again today, would you use an object-oriented API instead of a procedural API?

Re:Would use use OO?

[CondeZer0]

Does this answer your question?

"object-oriented design is the roman numerals of computing." -- Rob Pike

Re:Would use use OO?

[generalphilips]

I have to agree that this question does not make sense. It indicates a total lack of understanding of what OO means. I would argue that much of Unix actually is object-oriented. Think, for example, about the concept in Unix of presenting many types of IO devices as files. This achieves encapsulation, polymorphism, and a very high degree of abstraction by hiding implementation details, and presenting a very clean, intuitive inteface to programmers. There are many different types (you might say "classes") of files - pipes, sockets, regular, etc. - all of which can be passed into read and write calls. I'm sure I and others could go on all day about the OO aspects of Unix.

Re:Would use use OO?

[Anonymous Coward]

You could go on about its aspects, but not its true nature. You idiots don't understand what OO means. having a

call( mystruct )

is completely different then

mystruct.call()
mystructinherited.virtualcall()

Slashdot is full of idiots.

Re:Would use use OO?

[generalphilips]

You will always be lost in your ignorance.

Languages

[btlzu2]

Hello!

Maybe this is an overly-asked question, but I still often ponder it. Does object-oriented design negate or diminish the future prospects of Unix's continuing popularity?

I've developed in C (which I still love), but lately, I've been doing a lot of purely object-oriented development in Java. Using things like delegation and reusable classes have made life so much easier in many respects. Since the *nixes are so dependent upon C, I was wondering what future you see in C combined with Unix. Like I said, I love C and still enjoy developing in Unix, but there has to be a point where you build on your progress and the object-oriented languages, in my opinion, seem to be doing that.

Thank you for all your contributions!!!

Re:Languages

[five18pm]

It still is difficult to leave C. Object-oriented languages go to great lengths to hide data (That object is off-limits boy!), but sometimes you need to access data, that too, immediately. That you can get only C (and in that ugly kludge, C++).

Of course, you can also build object orientation in C, it is difficult, not impossible. Just look at the Unix file system and drivers. It uses both delegation and reusable classes. I know atleast one professor who taught object orientation with Unix file system as example!

Re:Languages

[tchuladdiass]

Every time I see someone bring up the topic of object this or object that, I wonder what you can do with objects that you cant do with regular functions? You can make the data types used in a function generic so that it has some of the same benefits of "objects" -- for a good example see the qsort() standard library function.

Re:Languages

[the chao goes mu]

Not that hard, really. Structs stand in for classes, void pointers to function calls stand in for methods. Inheritance is not possible without a lot of ugly work-arounds (possiby a pointer to a struct of the parent class within the child class/struct), but the basic class/method bits of OO programming can be done. (Also void pointers allow for variable data types, provided you remember to cast them back when you need the data.)

Re:Languages

[btlzu2]

It takes a lot of time to understand the benefits. It isn't just "about objects" it's about a completely different design philosophy. Not "objects", but "object orientation" which contains a lot of different facets. I was the same way until I had a class in Object Oriented Design and saw how the structure of a properly designed object oriented program reduced rigidty and fragility in a design. Also, how it made the interfaces clear and made division of responsibilities simpler for multiple programmers working on a big project.

This is one paper (PDF document) that helped me start seeing the light. If you're interested, give that a read for starters.

Re:Languages

[CondeZer0]

Rob's opinion of OO is well known; one of my favorite quotes by him is: "object-oriented design is the roman numerals of computing."

As for building on Unix and C, his(and Bell Labs') answers are well known:

- Plan 9
- Limbo
- Inferno
- [New]Squeak

Re:Languages

[jgrahn]

It takes a lot of time to understand the benefits. It isn't just "about objects" it's about a completely different design philosophy. Not "objects", but "object orientation" which contains a lot of different facets. I was the same way until I had a class in Object Oriented Design and saw how the structure of a properly designed object oriented program reduced rigidty and fragility in a design.

Have you worked with it in real life though? You'll see numerous examples of OO becoming wanking excercises, producing huge blobs of overdesigned, underperforming hard-to-understand code. There's a backlash in there somewhere, and it's been going on for something like ten years ...

I like expressing things in OO, but I no longer try to apply it mindlessly to everything I see. Reading Strostrup and learing perl helped, and I'm sure reading Pike would have helped, too.

Re:Languages

[HyperChicken]

object-oriented design is the roman numerals of computing. - rob (Taken from the Plan 9 from Bell Labs fortune file)

Re:Languages

[btlzu2]

I have used it in real life and will be the first to admit that it's not for everything. I created a paging application that receives SNMP traps and makes decisions on when/who to page. I wrote the paging part and my co-workers worked on other parts. We designed interfaces and used what we had learned in some course we took. It's a BEAUTIFUL application! :) We've modified/added functionality and the oo promise worked much to our surprise. We didn't break anything adding new code and it was minimal effort all around.

However, my younger co-workers think it should be applied to everything and it just doesn't work. For example, a protocol parser we wrote is just more simply written linearly.

Re:Languages

[oexeo]

That's a problem with your object design, not C++.

Re:Languages

[wirde]

OO without inheritance and polymorphism is not much fun...

Re:Languages

[iggymanz]

oh? After programming with Ruby for 3 years, I find mixins & duck typing alot more fun than I & P, which I hardly use anymore

Re:Languages

[samberdoo]

I think you miss the point. Much of java is implemented using native libraries aka C. Java is really an OO 4gl for C.

Re:Languages

[btlzu2]

Erm...no. Unix and the first C compiler was built using assembler. Then, a new C compiler was written using C. Everything has a foundation. We don't use assembler much anymore--it's *fairly* abandoned when compared to higher level languages. The same might be said for the Java's of the world. That was the whole point!

Obvious (and probably redundant)

[Camel Pilot]

What do you think of the Evil called SCO?

Re:Obvious (and probably redundant)

[Rosco P. Coltrane]

Ovious and short-sighted: read up about the corporate history of today's SCO Inc. and you'll discover the history of a poorly managed, but essentially benevolent, pioneering Linux shop called Caldera Inc (which turned sour about the time it split in two around 1999 if memory serves), and also the history of a good Unix shop called SCO (the real SCO, now called Tarentella), that used to own a pretty fine Unix clone back in the days and that's been schlepping along, but essentially was honest.

Today's SCO is a combination of a Linux shop that didn't know how to compete with RedHat and ran out of cash, their financial bakers called the Canopy group, and its chief shitheat, millionaire Ray Noorda )who are hell-bent on (1) getting Microsoft and (2) making money off of lawsuits), and their buying SCO's (the real SCO's) IP assets.

In short, SCO hasn't been evil for a long time, and won't be for very long either, and I pity the poor employees who got stuck in the corporate quagmire...

my question

[matosinho]

Just like google does in their googler test:

What's broken with unix? How would you fix it?

Emacs or Vi?

[Neil Blender]

nt

Re:Emacs or Vi?

[Anonymous Coward]

Actually there is a funny answer to this question on the plan9 man pages. typing man emacs you get something that only says : see acme

acme is a really powerfull and interesting concept in term of editor. (you can get it on unix either using the spinoff called wily, or better by downloading the complete plan9port for unix)

Rob also wrote another great editor called sam (available on both unix and windows). Very nice to use, my favorite editor on windows.

the main characteritic of those 2 is innovation. I really think people should give them a try just to see that there is something else out there.
The man is a great original thinker.

a.c.

View on linux

[noselasd]

What are your views on the free/OpenSource Unix like operating systems, such as Linux and the *BSDs ?

Re:View on linux

[Anonymous Coward]

uh? why don't you read it by yourself, from "Systems Software Research is Irrelevant"(that is linked in the /. article):

Linux

Innovation? New? No, it's just another copy of the same old stuff.

OLD stuff. Compare program development on Linux with Microsoft Visual Studio or one of the IBM Java/Web toolkits.

Linux's success may indeed be the single strongest argument for my thesis: The excitement generated by a clone of a decades-old operating system demonstrates the void that the systems software research community has failed to fill.

Besides, Linux's cleverness is not in the software, but in the development model, hardly a triumph of academic CS (especially software engineering) by any measure.

Re:View on linux

[noselasd]

I've read that. It's (legitime) rants about software *research*, which
is rather specific and narrow. I'm sure he has other views as well than the "software research is dead" view.

Backtracking

[Antony-Kyre]

Is it possible to design a computer so you can backtrack in progress? A method where it records perhaps the last three minutes, and you could click on something, and rewind your computer just like a video? And the hard drive would be at that state? Perhaps with intervals of five seconds. This would go way beyond RAID. (Not sure if I explained this properly.)

Re:Backtracking

[Rallion]

They did it in a slightly more limited scope in the new Prince of Persia.

Re:Backtracking

[Mikkeles]

Smalltalk 80 did this by making periodic snapshots of the image.

Language for new OS's

[Anonymous Coward]

The C programming language was written for and spread by the Unix operating system. While it's still a useful tool, and far better than the wholly untyped BCPL that preceded it, C is really starting to show its age. Is there an existing programming language that you would recommend for the implementation of operating systems? Would you recommend creating a new language for a new OS, as was done with Unix? Would you recommend the creation of new OS's at all?

Re:Language for new OS's

[billimad]

very good question. please ask this.

Re:Language for new OS's

[faragon]

Well written C has no parangon when programming an OS. Why?

-It's a pretty simple and language (try compare the complete C BNF vs c. C++ BNF)

-Modern C, as rich C++ without classes, it's fairly flexible

-C compilers are, usually, safers than C++ ones (this applies also for gnu cc and gnu cppc)

-C is alive, being updated; growing in just the common sense, without absurd and unuseful features (what about C++ massive redundancy?)

-C fits entirely on your head, C++ don't

- * add your contribution, with love, here *

Re:Language for new OS's

[tunah]

Who said anything about C++?

Re:Language for new OS's

[HawkingMattress]

Yep, good question. When you think about it, conceving an OS, and the language which will build it at the same time must be something totally mind blowing.
Given the power computers have today, the omniprecense of networks, OO and so on, i'm pretty sure that if one conceived an OS that way today and tried to rethink everything from the beginning, he could end up with concepts we don't even dream of.After all we all still think inside concepts which were found 40 years ago, even in network terms.
Things like plan9 are cool, but there is probably *much* more to be done, and the same thing can be done on the windows side.

Re:Language for new OS's

[Larmal]

He more or less answers your questions regarding new OS's in one of the essays... it's short, but a good read.

Re:Language for new OS's

[serialhex]

yeah, i think this one should be asked...

Re:Language for new OS's

[faragon]

"Is there an existing programming language that you would recommend for the implementation of operating systems?"

Would you recommend creating a new language for a new OS, as was done with Unix?

Right, no one said anything about C++, but, as almost everything, new solutions come along with new needs. At 60-70's the only one real option for OS development was machine code, assembly in best case. The next step, for OS development was avoid assembly, using a "not so low level" language, C was a nice option, still codeveloped/coevolutioned with UNIX.

Why I said C++? Just because there is a lot of people who think that it is the next step. I disagreed in my previous case for the OS case; for applications, ok, It's how I survive, still better that do it in java. Sorry if you were speaking about a "new and superfashion" new language for a metrocool OS, or whatever. Regards.

Re:Language for new OS's

[Anonymous Coward]

The Bell Labs paradigm was always to write a language to address a problem.

When the problem was writing a multiuser operating system, C was created to address it.

Nowadays the languages developed to address specific problems are usually DSLs (domain-specific languages), which have a limited scope of applicability. However, if the problem is writing a better operating system than *nix, or writing a better system interface than POSIX, then it calls for a better language than C. You may assert that C++ isn't the answer (though IIRC both Windows and Syllable use it), but that doesn't mean that C _is_. More than likely, C is not an appropriate language for developing a modern operating system -- stronger typing and enforced array index/block size bounds come immediately to mind.

how do you spell 'creat'?

[Anonymous Coward]

How do you spell creat/create and do you agree with Ken that this is the biggest thing that needs fixing in Unix?

Dude:

[Anonymous Coward]

How many stock options did you get when you joined Google?

Simple question really...

[Gothmolly]

Why?

Re:Simple question really...

[booch]

Why not?

Re:Simple question really...

[T-Ranger]

SYNTAX ERROR: "Why?" is not a WFF.

simple question

[Anonymous Coward]

why?

Innovation and patents

[Zocalo]

With so many of your ideas being used with such ubiquity in modern operating systems, what is your stance on the issue of patenting of software and other "intellectual property" concepts? Assuming that business isn't going to let IP patents go away as they strive to build patent stockpiles reminiscent of the nuclear arms buildup during the cold war, how would you like to see the issue resolved?

Re:Innovation and patents

[Anonymous Coward]

You just hope they leave your sig out of the interview. ;-)

CLI

[Moby Cock]

Has the Command Line Interface become outdated? What are your thoughts on the CLI and if you had to 'do it all again' would the CLI be as prevalent?

Re:CLI

[Naikrovek]

The command line was the ONLY interface when unix was first developed. X came long after the CLI and even if you're running Windows you still use the command line from time to time. If you don't you're just not using the power of your system, in my opinion.

raw text is the only true inter-system communication protocol. my cats and my computer understand a lot of the same words. my cat can't type but i can, and my computer can't understand the noises my cat makes, but i do - command pipes! where would we be without those, and those ONLY work in the CLI! we would be in the stone ages of computing without the CLI.

Re:CLI

[cyngus]

Let me answer that for you, NO. Before OS X I thought the command line was a primitive and clunky tool. Boy, was I ever wrong, I wonder today how I lived without it.

Re:CLI

[Kurt Gray]

My two cents: command line interfaces are still invaluable for remote administation. Unlike a GUI a CLI is still usable over slow connections, requires far less system resources, and simple remote terminal apps are available on almost every platform.

Re:CLI

[hackstraw]

Has the Command Line Interface become outdated? What are your thoughts on the CLI and if you had to 'do it all again' would the CLI be as prevalent?

I'll answer that. NO!

Until you find an easier way to interface with files and a computer than the CLI, than the CLI will not be outdated. Granted there are a few things that are easier with a GUI like wierd file management. For example, selecting a bunch of files with a mouse and then "deselecting" a couple of them is kinda hard to do from a CLI, but most everything else is quicker and easier with a CLI. Examples include:

- command pipelining -- ps auxww | grep something

- command chaining -- test -f AFILE && do something with AFILE

- for loops -- this is a biggy, until a GUI can easily do multiple operations on a subset of files, then I will always jump to the command line to get stuff done

- file centric behavior vs GUI centric behavior -- I guess if your real diligent about keeping all your files in one place and never use more than one folder/directory for files, then this is not an issue. But for me, I like sitting at a prompt and seeing all of my files in front of me and I can edit them, compare them, move them, etc. In a GUI, I find that every app defaults to the wrong directory for where I want to be using at the time and I have to monkey around in the file dialog box to do something like save or open a file.

- tab completion

- !command and !!

- the list goes on, I like the CLI, especially on my Mac.

Re:CLI

[CondeZer0]

You obviously have never used Acme, the greatest UI ever created:

"Acme: A User Interface for Programmers" By Rob Pike

And now you can even run it on (l)Unix thanks to rsc's plan9port: http://swtch.com/plan9port/

I was just re-reading the Acme paper this afternoon, and this made me wonder, acme was supposed to eventually support non-textual graphics but this was never implemented, why? and what ideas did you have for such an implementation?

If done right that could allow acme to replace rio completely ;)

P.S.: For the parent, if you want to know more about Rob's view on user interfaces I recommend that you at least read his papers: 8½, the Plan 9 Window System and The Text Editor sam

Re:CLI

[Alioth]

Not in the slightest. Occasionally, I have to fix a machine remotely - when all I have at my disposal is a small GPRS enabled mobile phone.

Without the command line, this would be impossible. The command line (with the very brief Unix commands, lacking syntactic sugar) is really the only practical way I can do some emergency sysadministration with my little Nokia. Since I don't want a vast XDA-style mobile phone, a GUI will never be practical for this - not now, not in 30 years time.

Re:CLI

[Anonymous Coward]

"Has the Command Line Interface become outdated?"

"In the rush to personal workstations, though, some of their weaknesses were overlooked. First, the operating system they run, UNIX, is itself an old timesharing system and has had trouble adapting to ideas born after it. Graphics and networking were added to UNIX well into its lifetime and remain poorly integrated and difficult to administer." - Quote

Re:CLI

[Moby Cock]

I will read them. Much obliged. :)

Re:CLI

[ClosedSource]

"The command line was the ONLY interface when unix was first developed."

Yes, just like B&W was the only type of movies made in the early days. In both cases, it was the limitations of technology that was the deciding factor. It wasn't as if Ken Thompson considered using a GUI interface and then rejected it because he concluded a command line was better.

"raw text is the only true inter-system communication protocol."

Raw text is a very commonly used protocol that most systems implement, but it's still just a convention that only works because people designed systems to support it. It's like Java; Java is a programming language that allows you to write programs that will run on any system that supports Java.

Re:CLI

[Anonymous Coward]

My two cents: Pencil and Paper interfaces are still invaluable. Unlike a CLI a PP is still usable over very slow connections (i.e. snail mail), requires far less system resources, and simple stationary sets are available in almost any store.

Re:CLI

[revscat]

for loops -- this is a biggy, until a GUI can easily do multiple operations on a subset of files, then I will always jump to the command line to get stuff done

Have you checked out the preview release of Tiger yet? Automator may do what you describe here, although I haven't personally had the opportunity to check it out yet.

I agree completely, though. It is difficult to match the speed of a CLI in competent hands.

Re:CLI

[Anonymous Coward]

"In a GUI, I find that every app defaults to the wrong directory for where I want to be using at the time and I have to monkey around in the file dialog box to do something like save or open a file."

So I guess a CLI automatically changes directories for you? Why is cd yada/yada/yada/yada (oops it supposed to be yado/yada/yada/yada, I'll have to start over) easier then browsing to a directory?

The many Flavors of Unix

[pillageplunder]

Rob, do you see in the near or even far future, the many different flavors of Unix (Sun Solaris, IBM AIX, etc) morphing back together?
Second part of the question, do you think that the different flaovrs "should" morph back together, or continue to grow apart?

Systems research

[asyncster]

In your paper, systems software research is irrelevant, you claim that there is little room for innovation in systems programming, and that all energy is devoted to supporting existing standards. Do you still feel this way now that you're working at Google?

the old school

[Triumph The Insult C]

what modern OS reminds you the most of your old school OS hacking days? what OS do you think keeps closes to the *nix spirit?

One tool for one job?

[sczimme]

Given the nature of current operating systems and applications, do you think the idea of "one tool doing one job well" has been abandoned? If so, do you think a return to this model would help bring some innovation back to software development?

(It's easier to toss a small, single-purpose app and start over than it is to toss a large, feature-laden app and start over.)

Back in The Day

[Greyfox]

Were programmers treated as hot-pluggable resources as they are today? There seems to be a mystique to the programmer prior to about 1995. From reading the various netnews posts and recollections of older programmers, it seems like the programmer back then was viewed as something of a wizard without whom all the computers he was responsible for would immediately collapse. Has anything really changed or was it the same back then as it is now? I'm wondering how much of what I've read is simply nostalgia.

Re:Back in The Day

[Some Bitch]

Many 'programmers' these days really know nothing about computers. I know it's a scary concept but it's true. I work tech support and I'm a hobbyist who dabbles in a little programming from time to time (been doing it 20 years so I've picked up a thing or two) and yet I constantly find that, outside the areas where they work, many of these guys know next to nothing. They just aren't the geek Gods any more, they're business analysts who did a course on Java. That said, we have at least one Oracle God. He may not know computers in general but he knows the entire Oracle system inside out, back to front, and sideways. He's expensive but oh boy is he worth it, he never has to clean up when it hits the fan because he makes sure it never hits the fan in the first place but if it DID hit the fan he's someone I'd definitely want around to pick up the pieces.

Re:Back in The Day

[Burb]

1995 seems a rather arbitrary cut-off point for this question. There may well have been a demystifying of programmers since, say, the 1960s, as IT has become more and more prevalent. But I can't think of any reason why 95 should be seen as a watershed in this regard.

Re:Back in The Day

[Greyfox]

Well the thing is, even though programmers are no longer treated like the mystic wizards of days gone by, it seems like very frequently the systems they've implemented will still fall apart the moment they leave. Last contract I worked had a messaging system in place that had been implemented by a contracting company 3 years earlier. When I got on the project, no one really understood how the thing worked anymore, and configuring it was more a matter of luck than any knowledge, documented or otherwise.

1995 seems like it was the very end of the "golden age" of programmers. After that point, we'd had a few crops of freshly-graduated students who had just gotten in to the industry for the money, and companies were really starting to understand that they could get people from overseas for less money. Managing departments was more about filling a body count than insuring that any project would be particuarly successful.

I'm hoping that the industry has learned something from the IT implosion but I'd really be surprised if it has.

Re:Back in The Day

[mikael]

Were programmers treated as hot-pluggable resources as they are today? There seems to be a mystique to the programmer prior to about 1995.

Everything was more nuts and bolts back then, and there were a lot fewer graduates looking for work as well.

I started university in 1986 when our Computer Science department had just been given funds to expand; we had the latest technology - 4.77Mhz PC's with 14" CGA displays, connected by RS-232 to a 3/180 Sun server with a 32-port terminal connection - that was state of the art.

Programming languages were Prolog, Borland Turbo-Pascal, LISP, 8086/6502/6809/68000 assembly. By 1988, everything was connected by Ethernet; MAC address cards cost around $1000, still required the offical Ethernet cabling - thick yellow cables attached by "vampire taps" to blue cables). The skills in demand then were UNIX/X-Windows/C, X.25, ISDN, TCP/IP and Microsoft C.

Any GUI development was in X-windows/UNIX. Windows 3.1 was still seen as a toy; even solitaire ran slow.

By 1995, everything changed when Windows 95 came out, development shifted from UNIX/X-windows to Windows NT/MFC. Windows NT was going to replace Unix, as Microsoft kept telling everyone. The most obvious changes can be seen in Byte magazine. The
earliest magazines were hardware brewing, then home computers with assembler, then PC's hardware reviews, to pure hardware/software comparisons.

At this time, Sun fought back with Java (they had to find a work environment bigger than the corporate network). By 1996, the dot com boom really took off, and thousand of graduates did Computer Science to learn Java/web page design/XML/Shockwave/Flash/Realplayer etc..., without learning the basics like assembly, microcode, ASIC design, systems programming, and object orientated programming.

Questions from a disinterested third party..

[SeanTobin]

Mr. Pike, a few questions if you will...

There have been several quotes back in the era of "Big Unix" before the dilution of a certain company's intellectual property. Specifically, one relating to "Never underestimate the bandwidth of a station wagon full of backup tapes." My first question relates to this quote. To the best of your knowledge, was Unix source or object code ever backed up and transported via this method, possibly through Finland?

What are your feelings on Lucy's younger brother from the Peanuts(TM) cartoon?

Do you feel that the Communist Hippies in Berkley were involved in a mass conspiracy to doctor previously released copies of source code to attempt to dilute the value of the Unix operating system?

And finally, someone were to want to subpoena an individual very much like yourself, where and when would the best place to do so be?

Licenses

[Anonymous Coward]

To what extent do you believe the success of Unix can be attributed to the early use of the BSD license, which allows anybody to use and re-distribute the code without placing restrictions on the actual use or re-distribution?

Unix Versus Multics

[Anonymous Coward]

As a longtime Multics user (around 35 years), I've been hearing more and more about "Unix" over the past decades. Can you convince me to switch? In particular, I rely heavily on access control lists and PL/I code. What advantages can Unix offer me? My home computer is a GE-645, and at work we use Honeywell 6180-series machines.

Why does Roblimo think you're a Unix co-creator?

[hruntrung]

Unless I'm very much mistaken, Mr. Pike, you aren't a Unix co-creator. Dennis Ritchie and Ken Thompson are the co-creators of Unix. If my very quick Google research serves, you joined Bell Labs in 1980 and worked a lot on Plan 9 and the first Bitmap window system for Unix.

So why is Roblimo wrong?

ReiserFS and the future of file systems

[booch]

What do you think of the work Hans Reiser is doing with file systems? How does it differ from and/or improve upon Plan 9? What do you think of his theory that (nearly) all database functions should be done by the file system? What do you think about being able to treat files as directories in order to get to special (or not special) info? Is it useful to be able to treat a tarball as a file when you want to and as a directory when you want to? How about file metadata? Data forks? Do you think Linux, Windows, or Mac OS X will come up with the better database/search-enhanced file system?

Re:ReiserFS and the future of file systems

[Anonymous Coward]

Mod parent up! The free software community can use insight into this problem from a legend such as Rob Pike.

Re:ReiserFS and the future of file systems

[Eraser_]

Already commented or I would moderate this up. I would like to see the answer to this question, especially since Mac OS has been doing this for quite some time with their data/resource forking. OS X has furthered this with their .app "folders" which execute when double clicked, but expose their true contents when commanded. It's like resedit is built straight into the Finder, and thats pretty neat.

Re:ReiserFS and the future of file systems

[fossa]

I'd also like to see this question answered.

Recommended change to BSD's & Linux

[CFrankBernard]

What is your most recommended major programming change to the BSD's and Linux, especially for Theo de Raadt and Linus Travolds...anything from the Plan 9 OS?

Re:Recommended change to BSD's & Linux

[Anonymous Coward]

I would like to now what U think of the design of the BSD vs linux kernels? Is one approach prefereable to another?

Re:Recommended change to BSD's & Linux

[CFrankBernard]

I don't know enough to comment authoratatively. I currently agree with those who think OpenBSD/MirOS makes the best/most secure yet highly configurable firewall/router, FreeBSD (maybe DragonFly BSD in a year) the best/fastest server, and Libranet for an inexpensive easily upgradeable Debian desktop, but I look forward to the unique features of XandrOS/Samba, Suse/YaST/Novell/Evolution and something to replace Outlook for those companies that will keep Exchange for a few more years.

Since you now work for Google...

[anonymous cowherd (m]

How can search as a concept become better integrated into the desktop? Are projects like dashboard the next killer app?

The Duck Problem

[Anonymous Coward]

I have this problem that no-one takes seriously. I can't help it, it started when I was a kid.

But I like going to parks and touching the ducks. But in November a park keeper caught me fiddling with one, and called the police.
I had to lie and say I was rescuing it from drowning. They didn't believe me, I could tell, but they couldn't arrest me because it wasn't a crime. Anyway, some bastard told the local press, and I was in the paper. There was no picture but my name was in the same sentence as 'duck molesting.'

I'm so embarrassed, I'm going to lose my job because I'm a vet.

Once, someone bought a duck in that had been attacked by a dog. I asked everyone to leave the room so as not to shock it. Really I just wanted to touch it. It died not long after.

Anyway, I saw a shrink and told him but he laughed. He asked me why ducks? I said because they had chubby cheeks and looked cute. Also they are slippery when wet. I like that. I like squirrels too, but I've never tried to fiddle with one. I might soon if I don't get help.

Anyways, my question is: what should I do? This is going to be my ruin.

Is it really a practical joke?

[ari_j]

Unix and C a Hoax - is it true? ;)

Re:OFFTOPIC: New Virus/Exploit?

[gregarican]

Not since I patched my machines the first time to not be succeptible to the oldschool blaster/dcom styles.

Is systems research really dead?

[Xpilot]

After reading your presentation on the death of systems research, I was rather disappointed at the dismal situation presented. Has anything changed since you presented that talk, or have your thoughts changed about the matter? As someone who is interested in systems research, what do you think is the most promising direction that is emerging today?

Re:Is systems research really dead?

[Anonymous Coward]

That has to be one of the best presentations of exactly what I think not just about the computer industry, but about reasearch institutions in general. There must be a larger divide between Commercial interests and research in order to spur creativity. There is a huge drain of people that can innovate in research and do more than measure things. There is an overwhelming amout of measuring things today and for what cause? To write papers that don't create anything new, but somehow get funding. I don't see the innovation in research today.

Why did it take Linux to popularise Open Source?

[RLiegh]

The GNU movement started on old Unix computers, and was aimed in part at them; so why do you think it is that the first wave of unix users were so resistent to the concept of Open Source?

Plan 9/OS insights from Google?

[jwjr]

Do you find any role for Plan9 at Google? Does Linux (or Linux with whatever customizations, extensions, and metamorphoses Google imposes on it) do everything Google needs or wants out of an OS platform? Does your experience with operating systems research pay off directly in contributing to the shape of the Google platform, whether for individual machine OS's, or for co-operation and clustered operation on the network?

Driving force of OS'

[cpfeifer]

What will drive the next crop of OS'? Is it in hardware innovations, new programming languages?

Re:Driving force of OS'

[Anonymous Coward]

um... what is going to drive the next generation of OS will be what has always driven OS development - user need/demand.

He's NOT a Unix co-creator

[hruntrung]

Jeez, someone, click on the fuckin link in the post with his name. He's not a Unix co-creator. He worked a lot on Plan 9, and wrote a bitmap windowing system for Unix. But he's not a Unix co-creator. The creators of Unix are Dennis Ritchie and Ken Thompson.

Re:He's NOT a Unix co-creator

[sgant]

Ah, so maybe you should email Roblimo and tell him this since the whole story is:

"Ask Unix Co-Creator Rob Pike"

Set Roblimo straight then...instead of just spouting off in the forums where no one but me will see it.

Re:He's NOT a Unix co-creator

[Anonymous Coward]

While you're correcting that, maybe someone should point out that UTF-8 is a character encoding, not a character set.

Article theft

[Zedrick]

The operating systems link goes to encyclopedia.thefreedictionary.com, which is a scam-site that steals articles from other sites, in this case from Wikipedia. The only thing they've added are ads. The original can be found at http://en.wikipedia.org/wiki/Plan_9

Re:Article theft

[sleepingsquirrel]

Why not link to the official plan9 site?

Re:Article theft

[Doktor Memory]

Although thefreedictionary.com leaves a kinda weird taste in my mouth as well, it's important to remember that what they're doing is (as I understand it) not only completely legal, but encouraged under the FDL.

So think of it as an advertising-supported mirror. If wikipedia ever runs into a major cash crunch, it might be helpful to have a second source for those articles around.

Re:Article theft

[Chalst]

Check out "http://en.wikipedia.org/wiki/Wikipedia:Send_in_th e_clones for the general Wikipedia attitude to such sites and the summary on thefreedictionary.com.

In short, thefreedictionary.com is using Wikipedia content more or less as intended, but is using Javascript to bypass the spirit, if not the word of the syndication license, and is in partial violation of the GFDL. As such, it is rated middling amongst violaters of the wikiepdia syndication license. Most wikipedians do not object to syndicated content getting higher rankings, if they are achieved fair and square.

Re:Article theft

[boots@work]

If wikipedia ever runs into a major cash crunch, it might be helpful to have a second source for those articles around. ... an out-of-date, non-editable mirror, run by people apparently without any interest other than making a quick buck. OK, better than losing the material altogether, but not much better.

Not everything that is legal is ethical.

Re:Article theft

[Anonymous Coward]

Theft? Copy infrigement you mean.
Oh. Not copy infrigment even.
PERFECTLY LEGAL.

Fucking Slashdot hypocrisis.

The question is in the story really...

[lxdbxr]

What exactly are you "Commander" of?

I'm not sure

[Anonymous Coward]

...but you've certainly come to the right place to ask.

The future of *NIX

[mbonig]

As *NIX and linux become increasingly popular in the business place people are looking to push Microsoft out of the field, replacing servers and workstations with free alternatives like Linux and some BSDs. This is causing kernels and OS designed for server performance to progress to desktop solutions. Do you feel that *NIX should stay in the server marketplace and focus solely on that market, or do you think moving the OS/kernel into a desktop role is "A Good Thing"??

ulink --- Come on did you just forget the 'n'????

[CyberSnyder]

subject says it all...

About Windows and UNIX

[mary_will_grow]

In your opinion, what are the key differences between the UNIX and Windows platforms from both a software developer's and an end-user's point of view?

Mod this down please -- technically incorrect

[HishamMuhammad]

Come on, I don't know how this was ever modded up (who am I kidding -- of course I know... just look at the buzzwords).

Anyways, it makes no sense whatsoever to ask "objects instead of an API". I won't explain what objects or APIs are here, but an API can obviously be object-oriented (need an example? the Java API).

FROM MY COLD DEAD HANDS

[Anonymous Coward]

The more worms et. al. get into a computer, the more a basic user needs to be familiar with tinkering with stuff behind the smiling computer icon. cf. the section "MORLOCKS AND ELOI AT THE KEYBOARD" of "In the Beginning Was The Command Line."

Re:FROM MY COLD DEAD HANDS

[Anonymous Coward]

I don't know what beginning that title refers to, certainly not the begining of computers. Computers predate the command line by many, many years.

You use a mac, don't you

[Anonymous Coward]

Only mac users call the command line the 'CLI'...

Re:You use a mac, don't you

[Weaselmancer]

Only mac users call the command line the 'CLI'...

Nope. The Amiga always called its shell the CLI.

Xanadu, Google, & Plan9

[Anonymous Coward]

Computing has advanced so much in the past ten years that it seems it's impossible to go back and correct any origional mistakes in the way we think about computers. With the amount of money behind the corporate machine, it really seems as if incredible ideas like Plan 9 and Xanadu are now things of the past--once good ideas that could have changed the way we approach computing, that simply missed their time. Do you see hope for these ideas and projects, or any other new or old computing paradigms that have potential for big change, but would require scrapping, say, the past several years of stagnant and unorigional work? -Ryan

Re:Why does Roblimo think you're a Unix co-creator

[AJWM]

He also worked on Version 8 Unix and (IIRC) came up with an early version of the /proc filesystem.

He's certainly a co-creator of what we now know as Unix.

Microkernel vs. Monolithic

[Ransak]

In the marketplace, monolithic OSes seem to be dominating, despite the advantages of microkernel OS design. I know this is straying into many other issues but from your point of view, why are monolithic OSes still viable in the marketplace - and why hasn't the public (ie, the 'programming public') demanded more?

Re:Microkernel vs. Monolithic

[Greyfox]

I seem to recall from the OS/2 days that the developers of the OS/2 PPC edition, which was to be microkernel based, discovered that the "obvious advantages" of the microkernel based OS were not so obvious. The canonical wisdom of the time was that microkernel based systems were smaller and faster than their monolithic counterparts. This proved not to be the case, at least for OS/2 PPC. It turned out to be bulky and slower than its intel-based brother, and was abandoned well before IBM abandoned OS/2 completely.

I suspect the "programming public" don't demand more because very few of them do kernel-level programming, and those that do could care less as long as the kernel they're working on isn't too hideous and runs well enough to meet their needs. The biggest advantage a microkernel gives you is better partitioning of functional responsibility, and if you look at a monolithic kernel (Like Linux) you'll see that in many cases it's pretty easy to determine what piece of code is responsible for the function you want to change.

Re:Microkernel vs. Monolithic

[eog]

"Because you need a brain the size of a planet to design a microkernel based system and we only have egos that big." -Dave Presotto

Re:Microkernel vs. Monolithic

[Ransak]

I vaguely remember there being complaints that the speed of microkernel OSes were not the same as their monolithic counterparts, but the stability was much greater by large factors (going on old memories here, so I might be wrong). Wasn't there a major advantage of micrkernel OSes with SMP systems in regards to processor overhead? At any rate, I can think of a few data warehouses that would trade clock cycles for stability ;)

If I remember, the mantra of the microkernel proponents was 'scalability, extensibility, and portability'. It just seems like this was almost abandoned at the kernel level, and it was one of the more impressive ideas I had seen come along (at least, I thought so!).

You might be right about the 'programming public' simply not working down on the kernel level. Ah well, guess I can dream.

Re:Microkernel vs. Monolithic

[Anonymous Coward]

"I touched some guy's underwear. Does that make me gay?" -

Depends, did that happen before or after you where sucking his dick?

Re:Microkernel vs. Monolithic

[argent]

The closest thing to a real microkernel in a widely used personal computer was AmigaOS. It was small, fast, efficient, and the ability to add services, networks, window management layers, and file systems simply by registering an application's message port in the appropriate list was widely used. I have never seen such wide and rapid development of useful and innovative OS-level components by hobbyists.

Unfortunately the system stagnated for several years while Atari's Jack Tramiel (formerly of Commodore) and Commodore's Irving Gould fueded over the Commodore Amiga (formerly a spinoff of Atari), and by the time Commodore foundered the Amiga was doomed.

The only other real Microkernel I know of that's widely used is QNX. At one point it was going to be released in a rebadged version under the Amiga name, but contract and financial problems (the folks at QNX actually wanted to be paid, fancy that) deep-sixed that. I believe there's atill a "Live CD" demo version of QNX available.

Re:Microkernel vs. Monolithic

[argent]

I vaguely remember there being complaints that the speed of microkernel OSes were not the same as their monolithic counterparts

It is harder to get good performance out of a microkernel for blocking operations, because where a monolithic kernel automatically gets a shepherding process to carry the operation to completion, the Microkernel comonents have to arrange for internal threading or queuing to avoid blocking on requests.

Other than that there's no inherent performance problem with Microkernels, unless you try and mindlessly force every component into its own protection boundary... which is fine for academic study, but a silly idea for production: a microkernel where components run in the same protection boundary when that's appropriate is just as much a microkernel design. Unfortunately, most of the big name microkernels are designed for academia and make precisely that mistake.

Re:Microkernel vs. Monolithic

[Anonymous Coward]

A lot of OS/2 was written in i286 and i386 assembly by speedfreak Microsoft hacks. It should be no suprise that a portable fully modern, kernel was slower than regular OS/2. See WinNT versus Win95 for example.

(Not to say that WorkplaceOS didn't have other problems, but in theory a psuedo-microkernel OS would have been fine for the target markets. For example, Digital UNIX or NeXTStep.)

Re:Microkernel vs. Monolithic

[Anonymous Coward]

Actually, the closest thing to a real microkernel in a widely used PC is Windows in its NT variant (NT, 2k, XP, 2k3). It is the only widely-used OS that can present different faces (POSIX, Win32, OS/2 1.3, DOS) all as user-mode subsystems. Yes, part of Win32 is run in kernel mode, but it's a driver, not a set of kernel system calls, and it's done for performance.

I think the main "advantage" for microkernels is the fact that everything runs as a separate process, and so is modular. This is convenient for stability if drivers all have their own address space, but it causes lots of wasted time swapping contexts and address spaces, and drivers can bring down the machine by screwing with hardware anyway! I seem to recall QNX won't crash if its filesystem module crashes -- it can just restart the process. But considering that the state of the disk would be rather unknown in the face of a crash, what is the point of running the system in an unknown state with a filesystem that crashes?

If you recognize that drivers can crash the machine anyway, so you might as well run them in the kernel's address space for performance, then having drivers be dynamically (un)loadable makes the system just as modular. WinNT has always had dynamic drivers, while having a monolithic kernel used to mean having to compile them in.

An actual comparison of NT's architecture with a "true microkernel" (whatever that is) would show it to be closer than any other widely-used OS. Note that OSX uses Mach, but not as a microkernel -- it's just runs a wrapper around FreeBSD.

aQazaQa

Re:Microkernel vs. Monolithic

[dotnth]

+1

Captain Pike...

[infinite9]

How does it feel driving around in that little buggy only being able to respond to questions with a simple yes or no?

What are you doing...

[Mark Wilkinson]

Google employees are apparently allowed to work on their own projects 20% of the time. Given that you probably can't comment on what you're doing for Google, what are you doing to fill the other 20%?

Re:What are you doing...

[Anonymous Coward]

Mod parent up! I would love to know what he is up to since he left Bell Labs

\k.

Re:Why does Roblimo think you're a Unix co-creator

[DylanQuixote]

Wikipedia article about Rob Pike...

The Year 2038 Bug

[WormholeFiend]

End of the World, or not?

Do you still have the cute harem outfit

[Anonymous Coward]

...that you wore at the Portland (Oregon) Usenix?

Which one is better?

[Anonymous Coward]

Which one is better? vi or emacs?

If 6 were 9

[not_hylas( )]

Mr. Pike,
I'm a fan of Plan 9 (along with the notion of what it could be), would you give us an idea of some creative uses for this OS.
Be as specific as possible.

Also, if you will, a link for (instructions) porting to Plan 9.

Re:If 6 were 9

[F2F]

Also, if you will, a link for (instructions) porting to Plan 9.

if you don't want to rewrite your code use this:

APE - The ANSI/POSIX environment...

rewriting your stuff is an eye opener though -- it mostly involves removing code that shouldn't be there in the first place but was forced upon you by unix' legacy crap.

My Questions

[Anonymous Coward]

Commander Pike, blink blink blink blink blink beep beep blink, blink blink beep?

Example of a reason why

[Antony-Kyre]

Imagine this. Accidentally installing adware onto your computer in a horrible way. Imagine a quick click of the mouse, and you can reverse the effects, instead of spending hours to remove it all. As for one second intervals for the last three minutes, I guess that could be compared to having 180 hard drives hooked up to your computer using a type of RAID, or something.

I just wanna know...

[halivar]

...did you get a proper SCO license before you started your work on UNIX? I have a feeling a lot of your code will show up as an exact duplicate SCO's IP, and I think it's important we get this issue resolved!

Ummm, there're reasons

[wiredog]

that people don't do systems programming in Java. Or C++, for that matter.

Re:Ummm, there're reasons

[btlzu2]

Well, I probably wasn't clear enough; however, I see these reasons (primarily performance) becoming more and more negligible as computers continue to become more powerful. There has to be an equilibrium point where the simplicity/ease of design (also an arguable point I'll grant) but less performance balances out the less simple, monolithic design with excellent performance.

Re:Ummm, there're reasons

[mikefe]

Though, it's harder to create good code from C++ than from C. It's simply more complex, and more to take into account, and fewer possible optimization points.

Re:Ummm, there're reasons

[Zordak]

There has to be an equilibrium point where the simplicity/ease of design (also an arguable point I'll grant) but less performance balances out the less simple, monolithic design with excellent performance.

You clearly have much to learn, grasshopper. Code should never be easy to design or maintain. It should be terse, ugly and preferably self-modifying. You may seek spiritual enlightenment here.

If you insist on using some of these new-fangled languages, at least learn how to use them properly.

Re:Ummm, there're reasons

[btlzu2]

LOL!

I was particularly fond of the Truth class!

class Truth {
boolean isTrue(boolean assertion) {
if (assertion != false) {
return assertion;
}
else {
return assertion;
}
}
}

Plan 9 relevancy

[sdcmk]

Do you feel that with the current usage in "traditional" operating systems such as Linux and Windows that there is no place, or need, for a Distributed Operating System in the industry?

The Google Operating System

[andyfaeglasgow]

Can you comment on the speculation about a new Operating System being created by Google?

Re:The Google Operating System

[Naikrovek]

there is no way he would comment on this. no way Google would allow the first employee-mentioned browser or operating system or messaging system to be via a slashdot interview... slashdot is not a powerful group of folks. we're a bunch of nerds that push our glasses up our noses while we vehemently debate the practicalities of single-line and multi-line comments in source code.

in other words, we bicker too much to be taken seriously.

Re:The Google Operating System

[Anonymous Coward]

in other words, we bicker too much to be taken seriously.

No we don't!

Why...

[codergeek42]

,,,did you decide to develop C and write Unix using that rather than using an existing language such as Pascal, etc.?

Spread UNIX

[TechDogg]

What would it take, in your opinion, to make UNIX more known to the personal computing public? Is it a good idea to bring UNIX to them?

Many different design strategies

[jd]

Plan 9, from my understanding, is essentially a microkernel with a blend of ideas from Unix and MULTICS. However, it follows the same core concept that most Operating Systems use, which is that the machine has a central processing system, plus devices to which instructions/data should be farmed.

In today's computing environment, this isn't always strictly the case. Multi-CPU boards are becoming more common, and clusters are inexpensive and powerful. Devices are also (returning) to a more "intelligent" state, with higher-end peripherals having comparable computing power to the main processor.

It would seem logical, then, to have a kernel which was more evenly distributed over the system, rather than hogging one specific resource. However, many attempts to do this are crude. Beowulf and MOSIX clusters, for example, run the whole kernel on all the nodes in the cluster, rather than just the bits of the kernel that are actually needed. This eats resources and limits the scalability of such solutions.

Do you have any plans for a distributed Plan 9? And, if so, what would you do differently to the other solutions people have adopted?

Re:Many different design strategies

[F2F]

you take the cake... no, really!

"any plans for a distributed Plan 9?"... you rock!

Re:Many different design strategies

[jd]

I've gotta ask this... HOW can that post possibly be interpreted as a troll? The meta-moderators will probably catch it, but that's not the point. The best possible interpretation I can put on it is that the moderator had some spare points, didn't understand the post, and decided that modding was easier than looking stuff up.

On that assumption, I'm including a dictionary of terms, here. If you don't understand something, read or ask. Leave the criticism to those who do.

First off, intelligent devices/peripherals. This term refers to devices that have their own processing capability, independent of the main processor. Most SCSI devices are of this type, as are some of the higher-end graphics cards. Any printer capable of handling Postscript is also of this type.

Second, distributed processing. This refers to splitting up a task into parallelizable components and running those components on (otherwise) independent systems. Anyone familiar with PVM, MPI or even RPC or CORBA is familiar with the idea. Beowulf and MOSIX distribute applications. distributed.net, BOINC (SETI@Home's new underlying system) and COSM all distribute data.

Distributed kernels are somewhat thinner on the ground. ANTS is about the closest I've seen. Normally, clustering systems such as MOSIX and Beowulf assume an identical (or nearly-so) kernel is running on all machines. But that's not particularly useful and results in the kernel consuming more memory and CPU cycles than it needs to. It also means that you can't always put everything you want into the kernel, in order to keep it from becoming too big.

There are other problems with this autonomous node philosophy. If a node crashes, whatever was running on it is lost. You need additional software (high availability) in order to recover. However, high availability generally knows nothing about clustering. It's independent software. As a result, when it tries to create a duplicate setup elsewhere, it is impossible to be sure it really is elsewhere. The clustering software could just as easily migrate it to the same machine the first copy is running on.

A distributed kernel has two parts - a "nanokernel" which handles everything that is absolutely critical in order to run the machine as a networked device (ie: it's got to be able to communicate with the network card and with memory, and it's got to be able to time-share between running the primary kernel and reading/writing to the network, but that's it), and a "visible kernel" which actually handles the processing, resource management, etc. that you expect a real OS to do.

The nanokernel simply provides a mechanism by which data and software can be passed around. It needs to do nothing else. It would be easy enough to implement High Availability at this level, simply by multicasting all I/O. This means that neither the "visible kernel" nor the applications need to know anything about HA - which is a good thing, as that's one of the biggest problems with many HA solutions. Because it can carry software, it can carry kernel modules for the visible kernel, or even a whole new kernel. This means that the OS on any given node would depend on what that node was trying to do, rather than having what you could do on the node determined by the OS that was running.

The visible kernel needs to know nothing about memory management or networking, as these are handled for it. It just needs to know that those tasks involve a call to the underlying nanokernel. Nothing needs to be built-in, as it can be provided on an as-needed basis. Because there is a lower level, the visible kernel can be hot-swapped at any time. The nanokernel just has to populate the new kernel's process table(s) so that the new kernel can continue where the old one left off.

What would this mean for Plan 9? Well, it means that you'd be able to cluster the computers. Plan 9 doesn't (as yet) have anything comparable to Beowulf or MOSIX, AFAIK. I

Re:Many different design strategies

[jd]

I'm not sure how to take your comment. MOSIX and Beowulf allow you to distribute applications across a network. distributed.net, BOINC (the system SETI@Home uses) and COSM allow you to distribute data the same way.

A kernel doesn't absolutely have to have everything stuffed into it. Well, we know that - you can have kernel modules for Linux, for example. If you can ferry around code and data, you can ferry around kernel modules too.

However, there's proably more you can rip out. Do you need an advanced multi-ring scheduler, if you know in advance what the node is going to be running? You can simply install the scheduler that is right for that workload.

On the other side, a distributed kernel would treat a cluster as a single computer. In other words, all devices/memory are visible to all nodes using a generic variant of NUMA. This being the case, common (but infrequenly used) kernel resources can be placed centrally. They don't need to be on every node. All the kernels will see those drivers or whatever anyway, and the latency doesn't matter if they're not used much.

Kernels would be able to use spinlocks, semaphores, etc, and have those visible to other kernels. It takes a lot less bandwidth to transmit a one bit semaphore than a huge data packet if the recipient is not able to take the data right then.

Re:Many different design strategies

[Anonymous Coward]

Plan 9, from my understanding, is essentially a microkernel

Your understanding is wrong. I suggest you study the work done in the 1970s and 1980s.

Do you have any plans for a distributed Plan 9?

Plan 9 is a distributed operating system. Do some reading.

Re:Many different design strategies

[jd]

First, my understanding comes from the Plan 9 website. I trust them to know what Plan 9 is, better than someone who won't even post a pseudonym.

Secondly, Plan 9 is not a distributed kernel. It is a client/server system, nothing more. Client/server is not distributed, except in the most primitive of senses. Plan 9 assumes that specific machines perform specific tasks. This is a rigid structure, not a distributed structure.

Further, the kernel is not a distributed component. Each machine runs its own kernel. A distributed kernel is a single kernel object that is split over and runs over multiple machines.

I am getting increasingly depressed over the apparent total lack of understanding of what a distributed system actually is. I'm not the greatest coder on the planet, but I am beginning to wonder if I'm the only one who understands that if you sub-divide the work between many machines, you can also sub-divide the components needed to carry out that work.

If you have N machines and N threads, then you don't need each of those N machines to be capable of handling N threads each. You're wasting CPU cycles in context-switching and task-switching without gaining anything from doing so.

If you have N machines and M threads, where not all threads are performing all possible operations the kernel is technically capable of, then you can reduce the kernel on all of those machines to only perform those functions essential for carrying out those tasks and also be capable of talking to the other machines.

I know of NO operating system that meets this description - that can self-modify the kernel in these ways, that can even pass bits of kernel around, or can even distribute threads. (MOSIX and Beowulf can't farm out threads, only whole applications.)

Re:Many different design strategies

[HyperChicken]

No, that's a clustered kernel, not a distributed kernel. Plan 9 does provides nice methods for distributing resources. Making everything seem like one big computer isn't the goal of a distributed system.

Re:Many different design strategies

[Anonymous Coward]

Ah.... MOSIX does process migration. It doesn't do anything similar to distributed.net, SETI, etc. Unless, of course, the application running on the MOSIX cluster already has parallel processing capabilities. Where then you'd start a bunch of the processes up. But you can do that in a more sane manner on Plan 9.

Re:Many different design strategies

[Anonymous Coward]

Man, you are histerically funny... keep the good work.

Has ever occurred to you that maybe _someone_ actually uses Plan 9 and that they might have a clue what they are talking about? f2f certainly does... but don't let reality get in the way of your comedy...

UNIX on the Desktop

[Ignatius_VI]

Do you see UNIX as a viable operating system for the home user on the desktop?

Lack of variety in operating systems

[Junks Jerzey]

Most people graduating with computer science degrees are only familiar with Windows and a handful of UNIX-variants: Linux, BSD, OS X. Is it good that stable technology is becoming the standard, thus allowing developers to focus their attentions elsewhere, or was this tremendous reduction in variety premature?

Sam Text Editor

[lfelipe]

What's the status of the Sam text editor ? Do you still use it ? Have you tried to apply some of its concepts to other editors ? I tried to use it , but the linux port did not work for me.

wiki

[spoonyfork]

WhyCan'tYouHaveSpacesInWikiKeyWords?

Revolution Needed?

[Anonymous Coward]

When Unix came out, it was written in the highest level language of any operating system: C. Why do you feel that operating systems are still implemented using the oldest, lowest-level languages?

With recent advances in high-level application languages like Java (low-latency garbage collection, dynamic inlining, etc), it seems to me that an operating system based on such a language would offer far more opportinuty for a vastly different kind of operating system more akin to an operating environment. Haven't attempts to add object-oriented features, such as in Plan 9 or to a limited extend Mach, failed due to the choice of implementation language?

It's just painful to see all the disgusting machinations necessary to implement a filesystem, network stack, scheduler, etc in C/C++...

Re:Revolution Needed?

[argent]

Java isn't significantly different from C++. It's a much more conservative language than 25-year-old languages like Smalltalk and Scheme.

Plan 9's native language is Aleph, which is a concurrent programming language descended from C.

Re:Revolution Needed?

[Anonymous Coward]

Java is vastly different from C in everything but appearance.

Can you imagine an operating system where there is no memory protection, interrupts, system calls, needed since these are requirements of having pointers? Where drivers run more safely than in a microkernel with less overhead than in a monolithic one? This question is important not just to get his take on it, but as a reflection on how and whether old hats can change with the times.

Aleph has pointers and is basically C with some syntactic sugar, just like Plan 9 is a warmed over version of Unix.

Re:Revolution Needed?

[Anonymous Coward]

Oldest, lowest level?

Oldest and lowest is machine code aka assembly language. Did a little of this in college had a Prof who wanted us to really appreciate what C makes so easy for us. Talk about painful. 1st program we were not even allowed to use a compiler that would let us use mnemonice names for the commands. We literally strung a bunch of numbers together to be compiled into assembly. Thank god he didn't make us just type in a bunch of 1's and 0's. We wrote a few more simple programs where we used the compiler. Compared to assembly C is FUCKTASTIC. I am a firm supporter of using the right tool for the right job and advocate picking the langauge to use based on the requirements. There are times when C is a bad choice and others where anything else is a nightmare

Here is a link to a programming language timeline that shows the ancetory of the languages http://www.levenez.com/lang/

Some Snippets:

Fortran:1954 Lisp:1958 Cobol:1959 Basic:1964 Pascal:1970 C:1971 AWK:1978 C++:1983 Perl:1987 TCL:1988 Python:1991 Visual Basic:1991 Java:1995 JavaScript:1995 PHP:1995 C#:2000

Re:Revolution Needed?

[Anonymous Coward]

Java isn't significantly different than C++? C++ has templates, typedefs, and unsigned integers. I'd say that's significantly different.

Re:Revolution Needed?

[Empty Threats]

Correction, when UNIX "came out," it was written in the lowest level language of any operating system. Everything else was either written in straight assembly or a sane language, like Lisp or PL/I. Furthermore, Lisp had all those lovely features in 1975.

It's also worth noting that many the problems we have today are based in the concept of the "operating environment" -- today's software and hardware design paradigms are rooted in C. Implementing any language other than Algol-derivatives on top of C certainly fits into the category of "disgusting machinations."

In short, as people have recognized for twenty years or more, UNIX succeeded because it was the lowest common denominator, not because it was any good.

Re:Revolution Needed?

[Anonymous Coward]

Plan 9's native language is Aleph,

No it isn't. Aleph isn't even developed anymore.

Re:Revolution Needed?

[Black Acid]

Java isn't significantly different than C++? C++ has templates, typedefs, and unsigned integers. I'd say that's significantly different.

Don't forget operator overloading, multiple inheritance, and the ability to access a superclass's superclass.

Re:Revolution Needed?

[sleepingsquirrel]

Check out the operating system project written in haskell.

Re:Revolution Needed?

[argent]

Java is vastly different from C in everything but appearance.

Java is crippled Smalltalk with half the important features missing: a hybrid object system, no dynamic objects, and virtually no ability to use reflection even if the other problems were fixed because it's jammed into a C-syntax straightjacket.

Can you imagine an operating system where there is no memory protection, interrupts, system calls, needed since these are requirements of having pointers?

Yes, Burroughs was using it getting on forty years ago now.

It did it better than any Java OS could, because it didn't restrict you to a single language.

Aleph has pointers and is basically C with some syntactic sugar, just like Plan 9 is a warmed over version of Unix.

Ah, you want to ask him about Inferno and Limbo then.

Thoughts on Bell Labs.

[geeber]

Plan 9, Unix and so many other great things came out of Bell Labs. Since the crash of the internet bubble, telecom companies have suffered immensely. One of the results of this is that Lucent has systematically dismantled one of the worlds greatest industrial research facilities. You spent a great part of your career at Bell Labs. What are your thoughts about the history and future (if any) of Bell Labs, and how did the culture of the Labs influence the growth of Unix?

Re:Thoughts on Bell Labs.

[Anonymous Coward]

Is this a question or your own little biography of the guy?

Re:Thoughts on Bell Labs.

[Anonymous Coward]

dumbass. what part of '?' do you not understand?

Re:Thoughts on Bell Labs.

[Cragen]

Wasn't Bell Labs sold in ~2001 to SAIC? Wonder how they are doing now? SAIC - where VP's cash-in and programmers just crash until the end of the present contract.

probably redundant by this point...

[Guano_Jim]

emacs or vi?

Re:probably redundant by this point...

[HyperChicken]

His response would probably be "Neither: acme".

Re:probably redundant by this point...

[Anonymous Coward]

Given that Rob has written a number of editors (sam, acme) and has been very critical of emacs and vi I'd say the answer would be "Neither."

Silver medal in Archery at the 1980 olympics?

[Anonymous Coward]

Your bio at bell labs and most other bios writen about you, mention that you won the silver medal in archery at the 1980 Olympics.

First, the US and Canada boycotted those olympics.
Second, Boris Isachenko, URS(BLR) won the silver medal at those olympics.

Is this an example of a joke that now has become folklore? Is it a way to "prove" to people that they should check their sources? Or is it just puffing up one's resume?

It seems strange in an era of quick and dirty research that you would still post this on your bio at bell labs. It only took a quick "I'm feeling lucky" google search on "1980 Olympics archery" to pull that info up.

So my question is, why do you keep that on your bio?

Re:Silver medal in Archery at the 1980 olympics?

[Black Acid]

For anyone curious, this is the bio in question, with commentary from http://c2.com/cgi/wiki?RobPike:

RobPike wrote the following mostly-serious but slightly tongue-in-cheek bio for his 1994 Usenix paper presentation on Acme: "RobPike, well known for his appearances on ``Late Night with David Letterman'', was also a Member of Technical Staff at BellLabs, where he has been since 1980, the same year he won the Olympic silver medal in Archery. In 1981 he wrote the first bitmap window system for Unix systems, and has since written ten more. With Bart Locanthi he designed the Blit terminal; with BrianKernighan he wrote TheUnixProgrammingEnvironment?. A shuttle mission nearly launched a gamma-ray telescope he designed. He is a Canadian citizen and has never written a program that uses cursor addressing." I believe he did appear at least once on Letterman as assistant to Penn and Teller, but he's joking to say he's well-known for that. The comment about archery is also a joke of some sort; Pike is a Canadian citizen who has worked in the U.S. for decades, and both Canada and the U.S. boycotted the 1980 Olympics, so you needn't go looking for the records of that year's Olympics to figure this out. The rest of his bio appears to be true. He now (2004) works for GoogleSearch.

Re:Silver medal in Archery at the 1980 olympics?

[karvalo]

It is a joke. The web page clearly mentions that.
From the page:http://c2.com/cgi/wiki?RobPike/

"I believe he did appear at least once on Letterman as assistant to Penn and Teller, but he's joking to say he's well-known for that. The comment about archery is also a joke of some sort; Pike is a Canadian citizen who has worked in the U.S. for decades, and both Canada and the U.S. boycotted the 1980 Olympics, so you needn't go looking for the records of that year's Olympics to figure this out".

Re:Silver medal in Archery at the 1980 olympics?

[valintin]

This is very interesting. It is strange that you did not consider it could be a typo. That would have been my first thought.

Re:Silver medal in Archery at the 1980 olympics?

[Anonymous Coward]

The problem with this, is that secondary bios about him quote this as fact. Often leaving out the 1980 part, which is the key to ever figuring out that it was a joke. Besides searching all winners of olympic medals.

Bio #1 http://www.lysator.liu.se/c/rob/

Wiki Page: http://en.wikipedia.org/wiki/Rob_Pike (no 1980)

Heros' page: http://www27.brinkster.com/sparksite/gallery/heroe s1.htm

Tribute to linux people: http://www.unixica.com/html/bell2.html

In fact if you search google groups, you will see Rob Pike used as an example of someone who is both athletic and scholarly.

If it is a joke, its not that good of one.

Maybe when Pike wasn't very well known it was an inside joke to people.

Bitterness about Linux from older developers

[Junks Jerzey]

Among a certain crowd, Linux is viewed at the savior of computing--a young, hip operating system for the new century. But at the same time, there have been definite twinges of bitterness from a more old-school crowd, including people like Brian Kernighan, Jaron Lanier, and possibly even you. This bitterness appears to stem from the horror of a 25 year old operating system returning to the forefront of computing (for anyone vehemently disagreeing, consider if clones of VMS or OS/360 were suddenly all the rage). Who is right? What's your take?

Re:Bitterness about Linux from older developers

[OnanTheBarbarian]

I think there might, just possibly, be a little bit of bitterness over the fact that Linux is a highly celebrated _clone_ of previous original work, like many other major - and over-celebrated - Open Source projects.

I think a lot of people out there were hoping to see something new. That might be unrealistic; perhaps OS progress is something that now must be done incrementally from now on. I certainly understand how disappointing the latter prospect would be to people who experienced a more exciting period of developments in computer science.

Inturruptless Realtime Operating System

[Anonymous Coward]

I read somewhere that the space shuttle uses an intruptless realtime os... do you think this type of design will eventually replace the operating systems of today? will we ever see the end to dialog boxes halting progams waiting for user input, will the flow of data from devices actually flow and not be timesliced? also please explain to us the differences between a true realtime os and and interupt based os.

thank you for your time

fu-kung.

Not Co-Creator

[kjs3]

Rob Pike wasn't a "co-creator" of Unix. He came to the Labs in the early '80s, more than a decade after Unix was created. He's very important and influential, but not for that...

fork/exec and other techical decisions

[Anonymous Coward]

fork()||exec() was a great idea for its time, but recently the overhead of page table duplication has meant that supporting that model slows down or adds much complexity to operating systems. What are your feelings on this problem, and what other significant decisions made in UNIX do you think have aged particularly well or particularly poorly?

I'd also like to know particularly what you think is the current and future usefulness for treating devices, sockets and named pipes like files, for the UNIX threads/ processes model, and for UNIX signals, and of course anything that you think is interesting, or decisions that you would like to revisit based on current knowledge and current hardware.

Re:fork/exec and other techical decisions

[boots@work]

fork()||exec() was a great idea for its time, but recently the overhead of page table duplication has meant that supporting that model slows down or adds much complexity to operating systems.

Do you really have numbers for performance? I had thought that the relative performance of fork was going up over time, not down, because less and less data is copied. In particular for the case where the child execs without doing anything else then very little will be copied.

And this copying is relatively small compared to the work of starting a new process: loading the program and any libraries it needs, dynamic linking, program startup...

As far as complexity: well, it makes a (highly-tuned) kernel more complex than just having spawn(), but it makes the applications a lot simpler. Imagine how many parameters spawn() would need to have to support the flexibility of doing things in the child process before exec'ing.

Anyhow, it's not like you could remove much complexity of fork() -- assuming that you want multiprocessing or multithreading there is always going to be a need to copy an mm.

An old grudge, an new liscense?

[emil]

Two questions:

1. Dave Cutler, mastermind of the Windows NT kernel, once described UNIX as a "junk OS designed by a committee of Ph.D.s.".

   Given two important facts:

   • Windows NT is mostly written in C and C++, both Bell Labs innovations, and
   • IE/Mediaplayer integration has turned the Windows NT codebase into a security disaster

   How would you respond to Cutler's assertions, and how would you rate the code quality of the NT kernel (assuming that you might have perused the recent leaked NT4 source)?

2. While UNIX-like operating systems are growing in popularity, actual Bell Labs code is rarely encountered in free operating systems because of licensing issues (with a few notible exceptions).

   This is a frustrating situation for all of us. Do you see any possibility that major portions of UNIX and Plan 9 source being released under licensing that major distributions would find acceptable?

Please also accept my personal thanks for your work in the field of computer science. The influence of the community of researchers at Bell Labs will be felt for many generations to come.

You're probably right...

[andyfaeglasgow]

but I like dreaming :-)

Re:You're probably right...

[Anonymous Coward]

Of a world ruled by one crappy OS -- The Google OS?

CORBA(com) scriptable interface

[sasoon]

If you are to design a new OS(redesign Unix), would you implement a some sort of CORBA(COM etc) interface to the kernel and core system tools.

starting from scratch today?

[Nerkles]

If you were just starting out with today's computers, what would you do differently or the same?

Commander Pike, eh?

[poot_rootbeer]

Are all of his answers going to take the form of a sequence of flashing lights mounted to his life-support washing machine?

Freeness

[identity0]

How much do you think UNIX's success has been shaped by the relatively light restrictions placed on its use, distribution and modification?

The original UNIX, BSD, and now Linux seem to have been 'freeer' than other OSes of the time, do you think they would have been successful without this?

Finally... vi or Emacs? ; )

Re:Freeness

[identity0]

Oh, and an addendum - Do you think Plan 9, or any other OS with a relatively restrictive license can succeed now against traditional UNIXes and Microsoft?

Re:Freeness

[retiarius]

FYI 'vi' vs. Emacs

although we've seen proof (usenix, 1983) that some of its features (e.g. "-v") are harmful, 'vi' originator WNJ himself
is on record (somewhere) preferring 'cat' as a line editor.

sorry for lack of proper citations -- we'll just have to "a9" for them!

Hindsight

[stuffduff]

If you had to do it all over again, what would you do differently? Why?

Hardware

[SwansonMarpalum]

Rob, When you were engineering UNIX, processors weren't as beefy, memory was grotesquely expensive, and storage was a premium. These days all of these resources have largely become commodities and can be frittered away wastefully by neglectful programmers. Do you think that in an alternate world where UNIX hadn't been conceived as early in the progression of hardware as 1970, rather had come along at this stage in the timeline where hardware vastly outpaces all but the most glaringly negligent software, it would have been as compact, fast and efficient? Thanks! -Alex R.

Re:Hardware

[Anonymous Coward]

UNIX was not terribly compact fast and efficient for it's day.

UNIX was written in C. C was a high-level language then, and most OSes were written in machine code.

The reason for C was to make the OS portable.

And a lot of people probably felt Unix was boated back then too.

What's your name?

[andyfaeglasgow]

Captain Manwaring: "Dont tell him Pike!"

Just reminded me of one of the funniest lines ever written. For non-brits, check out the show "Dads Army"

Re:What's your name?

[cranos]

And of course there was "Don't Panic, Don't Panic!!" and "Put that light out"

Sigh I'ld be showing my age if I wasn't 27.

More obligatory questions

[sleepingsquirrel]

What operating system to you use the most for your personal and/or work-related needs?

Re:More obligatory questions

[eog]

As far as I know Rob Pike runs the unix port of the Plan 9 userland on Mac OS X.

Re:More obligatory questions

[Anonymous Coward]

I think an interesting followup might be instructions for n00bs on how to setup whatever it might be.

Re:More obligatory questions

[Project2501a]

He is using Plan 9. I saw it in the Plan 9 mailing list somewhere

Please explain your alleged Olympic medal

[Anonymous Coward]

Your online biography claims that you won the Olympic silver medal in archery in 1980 (the year that several western nations boycotted the games). However, the official FITA web page for this event says that the silver medal was won by Boris Isachenko of the USSR (Tomi Poikolainen of Finlans won gold; Giancarlo Ferrari of Italy won bronze).

Please explain!

HURD

[Anonymous Coward]

Someone please put together a good question about The HURD and his views on the project, if I do it, it won't get modded up, I'm bad at expressing myself in english.

Re:HURD

[Anonymous Coward]

uh, yeah, I could hardly make this one out.

Comment removed

[account_deleted]

Comment removed based on user account deletion

Is Unix fixable in the *real* world?

[Anonymous Coward]

From your own words, Plan9 wasn't a "unix evolution", you didn't get a small subset of unix features and then start from there, you just started Plan9 from scratch.

While I agree that "starting from scratch" was much easier than "fixing unix", I think that fixing Unix is possible. Just much harder, but not impossible. I also agree with your "system software research" paper, although I'm not so pessimistic. Linux/BSD's may not be the perfect OS, but they're not subject to comercial companies, and making them evolute is not impossible.

I know, your paper claims that systems software research is dead, and the "right step" would be constructing a new operative system, not making linux/BSD's evolute. But code reuse is a research area, isn't it? "Evoluting" a operative system is also a valid research field IMHO, and switching a operative system like linux/BSD to new design concepts like those in Plan9 is much more interesting than writing Plan9 from scratch. Do you think it's possible to make linux/BSD's evolute? (take in the account that in your own paper you admit that Linux's strengh is the development model, so I don't think "it's a lot of work" is a valid excuse)

Plan9 advantages over Linux and *BSD

[Florian]

In which areas would you say is Plan9 (still) ahead of Linux/GNU and *BSD, the two operating systems which represent the most contemporary iteration of the original Unix design?

Workplace OS: a legendary IBM failure

[Anonymous Coward]

The failure was not due to being a microkernel so much as being everything to everyone. The project, called Workplace OS was supposed to replace Windows, OS/2 and AIX and run on PowerPC and x86 machines. The teams woking on supporting different operating systems didn't talk to each other much, and the end product ended up being late, huge and slow, and not running any of the target systems very well. It was then cut back in scope to replace OS/2 only, and wasn't even very good at that. *two billion* dollars later, having shipped only two versions to a lukewarm (at best) reception, IBM decided to cancel Workplace OS and stop developing new operating systems. That project is the biggest single reason that IBM has adopted Windows and Linux so enthusiastically.

If you google for "workplace OS failure" you will find some papers analyzing the failure and describing how poorly the project was managed.

Ease of Use/Shortening the Learning Curve

[SlashMaster]

1. What open source community initiatives do you think are desirable/necessary to make Linux appeal more to 1st time computer users/buyers and also to 'shorten the learning curve' for these people in-terms of desktop help, application-driver integration, etc.

2. I've noticed that linux/unix software that is or was sold (before becoming open-source or free) generally is easier to install and configure probably because the company wished to keep support costs low. Additionally, 1st-time users don't care about the underlying OS and software, they just want to get their work done. For instance, I haven't run-across a GUI interface for setting up LDAP yet(Its just an example of many apps are typically setup manually or via scripting). It just seems like a GUI desktop version of www.tldp.org combined with applicable man-pages is in-order. ... something like a Unix Expert System. For those of us who enjoy 'tinkering' and customizing manually is great, however, what initiatives would help that 'lowest common denomiator' or newbie.

...Perhaps an automatic tarball package expander/installer that parses the respective configuration file for application/library dependencies, similar to Debian's apt-get packager, (only for tarballs) and installs/verifies them as well? ...providing reasonable automated assistance in the process as needed?

3a. According to your research in the original post, OS Research was pretty much dead before Windows. I would then take Windows to be a paradigm shift in the OS world. Do you see any other paradigms on the horizon that can/will change OSs? For instance, neural-learning algorithms that make users more productive, recommending breaks and/or alerting them to tasks and alternative methods, providing file/application links based on the user's current tasks similar to a 'virtual assistant/secretary'(analogous to the Linux Expert System)?

3b. How about integrating speech recognition (Similar to IBM's implementation on OS/2), and/or hand-writing recognition into the kernel for broadened appeal and useage of the OS and its applications?

4. How much more 'critical mass' (in terms of # of PCs running Linux) does the open source community need to entice mainstream manufacturers into providing better driver support for new products?

A Jerq at Every Desk

[SimHacker]

Has your vision of "A Jerq at Every Desk" been realized, or do we still have a long way to go before fully exploiting the potential of personal computers?

-Don

Re:A Jerq at Every Desk

[SimHacker]

(For those of you who aren't Rob Pike or Dennis Ritchie:)

The Jerq was the original name for the Blit -- a name that management didn't approve of, for some reason.

More info: AT&T 5620 (and Related Terminals) Frequently Asked Questions

And also: Can someone advise me regarding a gui for UNIX

From: Dennis Ritchie <dmr@plan9.bell-labs.com>

Norman Wilson's account of the Jerq/Blit etc. is quite complete and correct, though there was some recycling of names. 'Jerq' actually was used quite early, when Pike got interested in bitmap graphics. The name was a takeoff on the Three Rivers Perq, which he (and I) saw at Lucasfilm Ltd. while attending an early Usenix. Blit was the slightly more PC version (suggested either as part of BitBlt or "Bell Labs Interactive Terminal). The originals used the Motorola 68000, and part of the development messup was AT&T Computer systems' decision to switch to the WE32000 processor with consequent delay for porting and reworking.

The earliest versions were not quite as wonderful in practice as Norman suggests for the later ones. They were built by the Teletype corp. model shop (in quantity of a few hundred) and downloading the OS took several minutes at 1200bps--necessary at startup, since they didn't have a ROM for the whole thing, just enough for doing a download. They were also static electricity antennas! Many is the time that I would shift in my chair, then touch the keyboard, only to have the terminal reset itself. I developed the habit of putting my hand on the heavy steel case before moving around.

On the other hand, the basic idea was architecturally right (and the later commercial versions were not so subject to static, and had ROM for the OS). They were even nicer at 9600bps.

It's good to know that Norman is still using his.

Dennis

Would you consider Mac OS X a version of Unix?

[the_webmaestro]

What do you think of Mac OS X? Have you used it? Would you consider Mac OS X a 'version' of Unix? Would you consider using it as your main operating system? What do you loave about it? What do you hate about it?

Re:Would you consider Mac OS X a version of Unix?

[andrew71]

and/or:

you sure heard about Bundles. What do you think of them?

Re:Would you consider Mac OS X a version of Unix?

[the_webmaestro]

A nice (albeit geeky--which fits the bill!) thing to add. If using my original question, please add this to Andrew's addendum to it! Thanks!

Don't ask this question

[generalphilips]

This is very similar to the another post about OO. Clearly you don't understand OO. There's nothing stopping a C programmer from using OOAD. Language is only incidental. Now I understand that OO languages make OOAD easier, but there's nothing stopping you from using those languages on any platform. If you like Java, well you can develop Java ON UNIX. What exactly is your point about C? C is closer to the metal than most OO languages, which is why OS developers write in C, and many libraries are presented directly to C. But you don't understand the way languages work if you're asking the question, "why C?". Ultimately everything comes down to machine-level instructions, so language is almost completely irrelevant to the OS. Perhaps you mean to ask, "Does the rise of virtual execution environments like Java's or .NET's diminish the importance of the OS?"

Re:Don't ask this question

[Anonymous Coward]

But you don't understand the way languages work if you're asking the question, "why C?".

No, you do not understand the way languages work. An OO operating system doesn't need memory protection, lowest common denominator one-way calls, lack off callback from kernel->app, freakin' reference counting on vnodes and sockets (did you know there's a garbage collector in the kernel just for sockets? why not for lots of other things?).

Re:Don't ask this question

[generalphilips]

Well, the things you mention do have something to do with language, so I see your point now. C isn't type safe. I believe that's your point. Why not make it, then?
However, I don't agree it's as simple as you think it is. I didn't know that sockets were garbage collected. I suspect (as my first guess) this is because a socket may have to stay around longer even after all the processes that were using it are gone. Reference counting works fine, and is pretty fast for things that can be destroyed as soon as they are no longer in use by any process. Now, I know that garbage collected systems can have very good amortized performance. However, in the kernel, it may simply be unacceptable to occasionally pay the price of walking through memory to collect garbage. The assumptions that work well at the application level might break down in the kernel. Remember that many complex algorithms simly aren't used in kernel development because of the fact that they are complex. The kernel must operate in all sorts of adverse conditions, like when there is very little memory, and might have all sorts of timing constraints on it, like when its running an embedded system.
By the way, what do you mean by:
An OO operating system doesn't need memory protection
I think you mean that memory protection is simpler in a type safe language. Memory protection is always going to be a very real problem for an OS. I wonder if you know anything about concurrency. There isn't any language that can automatically solve the problems that the OS developers must solve.
There are lots of smart people that do kernel development, I think you should give them a little credit. They don't just do things because "that's the way it's always been."

You

[Anonymous Coward]

Fail It

Talos IV

[Farce Pest]

Do we have to ask yes/no questions so Commander Pike can answer them (one blink for yes, two for no)?

(mod as -1: Star Trek reference)

Is UNIX worth it?

[jlrobins_uncc]

Is all of the current effort maintaining implementations of the UNIX way worth it? More specifically, when should we finally cut our losses and stop living with design desicions which were made decades ago for hardware and user constraints which are not nearly as relevant now as they were when they were made?

Re:Is UNIX worth it?

[argent]

What's the alternative?

Win32? You have got to be kidding.

NT kernel, new subsystem? Microsoft hasn't and won't release the API, so the only way into NT is through Win32 or Interix. Interix is UNIX.

VMS? Maybe.

Plan 9? Great design for largish networks, but you really need a bunch of machines to bootstrap it.

MacOS? Well, that's UNIX now.

BeOS? On life support in Germany.

What else would you do?

Re:Is UNIX worth it?

[jlrobins_uncc]

Well, Rob did something different -- took what they learned from UNIX and wrote something new. And in his eyes, Plan 9 probably has way-fewer warts than POSIX.

I'm surprised this was modded troll. I'm a UNIX user since '92 or thereabouts, and have my FreeBSD 1.0 CD to show for it. Rob is textbook hard core old school, yet he decided to develop something decidedly *different* from UNIX. Therefore, he must have felt that something more radical was warranted than tacking on new substructures to the old warhorse. Cleaner and more interesting solutions were only possible through starting fresh.

Anyway, I was interested in his opinion of if UNIX deserves incremental change and updates, or if it, in his opinion, is ultimately a dead-end -- that our time would be better spent working on something that takes what was good from UNIX yet leaves the bad behind, just as UNIX did to MULTICS.

For example, Plan 9's per-process mount tables are definitely interesting, making more general a concept found in UNIX, but a much bolder change than what one would expect on a UNIX. Similar things can be said of Hurd's abandonment of the 'process running as root has all priviledges' concept. Likewise for languages whose system runtimes perform array bounds checking automatically.

Re:Is UNIX worth it?

[argent]

First, I don't think Plan 9 is radically different from UNIX. If you follow the development through the 10th Edition before Plan 9, Bell Labs was already heading that way.

The per-process mount points are great, but they're not that radical, and not enough to consider Plan 9 "not UNIX". Microsoft Xenix had per-process and per-user-ID network mounts in 1984. Finely grained privileges have been found in a variety of UNIX systems, they're pretty much a necessity for any system that's DOD class B or better, and there's class B UNIXes.

BeOS is much more radical, but it's also clearly a collateral descendent of UNIX, as is QNX. Really, today, if you want to find an operating system that isn't clearly derived from UNIX, one way or another, you have to look at VMS or some of the really aggressive research operating systems like Amoeba: and UNIX emulation is de-riguer for any of them. Microsoft even ships a free hosted UNIX subsystem to run under the NT kernel.

I guess what I'm saying is that not only have the popular operating systems whittled themselves down to UNIX, UNIX, UNIX, VMS, UNIX, UNIX, UNIX, OS/360, or Win32... but even if your basic design isn't UNIX-like you have to provide a good UNIX environment one way or another.

Sure, there's plenty of room for nifty stuff under the covers, but the opportunities for a new OS that can't run the UNIX software library well... even if you're Microsoft... are pretty negligable. We are all UNIX.

Re:Is UNIX worth it?

[IntlHarvester]

if your basic design isn't UNIX-like you have to provide a good UNIX environment one way or another

I would say this has less to do with the design merits of UNIX itself; and more to do with the historical accident that UNIX is the only operating environment to ever become a vendor-independent standard. And because of that, free/cheap implementations became available. Be and Microsoft only support UNIX-like environments because it was cheap to do so.

Put another way, if DEC would have allowed a "Berkeley VMS", UNIX might not have become the consensus choice.

Re:Is UNIX worth it?

[argent]

I would say this has less to do with the design merits of UNIX itself; and more to do with the historical accident that UNIX is the only operating environment to ever become a vendor-independent standard.

There are design elements in UNIX that made that happen, regardless of the cost of the system. Back in the '70s and early '80s there were a lot of people who spent a lot of time and energy and, once it became available, money... on UNIX-like environments for mainframe and mini operating systems. There was no "Free BSD" back then, there was Idris and Regulus and Cromix and Lanetix and the Software Tools Virtual OS and Phoenix and Eunice (which was a commercial UNIX emulator, similar to Cygwin in concept, that ran on top of VMS... and people paid real money for it!). People bought VAXes, paid the VMS license, then spent tens of thousands of dollars MORE for a UNIX license and ported UNIX to it. None of this was free or cheap... most of these systems had to start from bare assembly code and build up to a UNIX-compatible environment. And sell it. And people bought it... not because it was cheap, but because they wanted UNIX, the UNIX design, and that's what it took to get it.

UNIX wasn't "cheap" for licensed commercial UNIX vendors either. It certainly wasn't cheap for Microsoft. They were the biggest UNIX vendor back in the early '80s, and they even advertised that MS-DOS and Xenix would eventually merge back before Bill Gates saw the Xerox Star and the Macintosh and got a new religion.

And all this was before BSD was anything but an academic curiosity. Most of it was before anyone outside academia had heard of BSD. Much of it was before BSD existed.

if DEC would have allowed a "Berkeley VMS", UNIX might not have become the consensus choice

Couldn't have happened. VMS depended so heavily on the VAX memory management they never released a VMS for their first generation RISC boxes, just Ultrix. And even if it could... UNIX was already competing with VMS on the VAX that DEC was forced (reluctantly) to implement their own VAX UNIX. And it wasn't just because of BSD: there were multiple independent UNIX-like environments for VMS... and one of them predated BSD.

No, UNIX didn't take off because of BSD. BSD exists because UNIX was already taking off, and it was taking off despite AT&Ts inability to sell it, not because AT&T was "allowing Berkeley UNIX"... there wasn't any Berkeley UNIX back when Idris and Regulus were developed because the ONLY way to get UNIX outside academia or AT&T was to clone it from scratch. You have no idea how exciting UNIX was back then, it was like the Macintosh, except it all started underground... there were just these Bell System Technical Journal articles, no "1984" superbowl ads, and people who were tired of the operating systems like VMS they had to deal with daily went "hot damn, I gotta get me some of that!"

Be and Microsoft only support UNIX-like environments because it was cheap to do so.

BeOS? BeOS has a non-traditional kernel, but it's not just providing a UNIX emulator... it boots right into a single-user UNIX environment, with shell scripts forking and execing away to bring the GUI up. Microsoft bought a whole company, Softway Systems, to get Interix. I'm not sure that counts as cheap, even "cheap for Microsoft".

Re:Is UNIX worth it?

[jlrobins_uncc]

Well, as far as vendor-independent operating system standards, there was also the MS-DOS, PC-DOS, Concurrent-DOS family (any others in there offhand w/o consulting the wikipedia?). And from one perspective, the vedor lawsuit perspective, it was just as successful as the UNIX family.

But it was killed off by Windows 95 (or was it axed by dirty tricks in the Windows 3.X family? I dunno).

But perhaps it was because the UNIX family of vendor neutral (and hardware neutral) operating systems always targeted larger, multiuser machines, and end-user hardware at the PC level ended up being 'large' (hardware resource-wise) and multi-user (mom, dad, kids, maybe not concurrent as in the good old days of the university vax running 4.3BSD, but hey!) that let the initial design of UNIX still be pertinent today. And that the source code written on top of UNIX APIs is still relevant today.

Re:Is UNIX worth it?

[IntlHarvester]

All good points, but still it goes back to the "historical accident" of AT&T providing cheap source licenses to all comers.

As far as I know, until Windows NT*, there was no other third party operating system a hardware vendor could license and customize for their particular hardware.

* Back in the RISC chip days, the press had figures like $1Million for the right to port NT to your platform, plus of course the per-copy royalties on top.

Re:Is UNIX worth it?

[argent]

but still it goes back to the "historical accident" of AT&T providing cheap source licenses to all comers

They didn't. They provided source licenses to universities, if you weren't a university you didn't get in on the deal. The later commercial UNIX licences were a LOT more expensive. That's why people were writing UNIX clones in the first place: they wanted to get UNIX... not just any operating system.

until Windows NT*, there was no other third party operating system a hardware vendor could license and customize for their particular hardware

There were hundreds, of all degrees of sophistication from simple program loaders like CP/M and MS-DOS through high-performance real-time and time-sharing operating systems. If the only operating systems available for hardware vendors to customize had been UNIX and NT then things would have been pretty dire back in the '70s and early '80s. Anyway, personal computers that used these customised third-party operating systems included the Apple-II (UCSD p-System, CP/M), TRS-80 (CP/M, ...), IBM-PC (CP/M-86 and MS-DOS), TRS-80 Color Computer (OS/9), Amiga (Tripos), Palm (AMX), Heath H-11 (RT-11). If you wanted UNIX, before AT&T was legally permitted to sell it, you had your choice of close clones like Idris and Regulus, and UNIX workalikes like OS/9.

Palm and Amiga are interesting cases: they used a third party OS for only part of their system. AmigaDOS hosted Tripos over the Amiga Exec, and PalmOS is a hosted OS running under AMX.

No, the historical accident had nothing to do with it. People whose only exposure to UNIX was reading about the design went out and implemented parts of it. People created UNIX workalike environments under operating systems (including VMS) because it was more productive. My first job, I spent several weekends typing in the listings from Software Tools and creating a runtime for RSX-11 so I didn't have to deal with MCR or DCL, the two comman line interfaces available for RSX. DCL, by the way, was the new improved user-friendly DEC Command Language they developed for VMS.

No cheap UNIX source licenses there, and no urge at all to run "Berkeley VMS" even if it had existed.

Re:Is UNIX worth it?

[Anonymous Coward]

For example, Plan 9's per-process mount tables [bell-labs.com] are definitely interesting, making more general a concept found in UNIX, but a much bolder change than what one would expect on a UNIX.

I've got to ask about this one: On my Linux 2.6 kernel, why is /proc/mounts a symlink to /proc/self/mounts?

Funny story....

[Anonymous Coward]

I was talking to my brother a few months back - we hadn't been in contact for several years, and were catching up. He's been in the Unix world for a very long time, and knows most of the old Unix guys personally.

He was really pleased when I told him I hate windows, and run Linux on my desktop, and that I admin Linux boxes for a living - when I told him I do most of my programming in Pike, he said "Oh god, that wasn't written by Rob was it? It would figure for that egotistical bastard to name it after himself."

He seemed a little relieved when I told him Pike was from Sweden, and started out as LPC.

Lucent / Bell Labs

[Anonymous Coward]

How do you rate the chances of survival for this once proud company / organization?

As a current Lucent employee I see the family silver being sold off every day (closing the Holmdel library was a real kick in the teeth for many). Can the worlds best know research lab recover from the bubble and it's aftermath?

Sincerely

Pat (Posting as AC so not to annoy "the street")

Schemas for UNIX

[generalphilips]

Unix suffers today from a proliferation of file and output formats that makes integration between the CLI/config files and the GUI awkward at best. For example, a common idiom for Unix GUI tools is to parse output from a CLI program and present it visually. This would be greatly simplified and much smoother if those programs produced structured output rather than raw text. The same holds for programs that read configuration files, like resolv.conf. Do you think UNIX would benefit from standardization of formats that coalesce around XML? What do you think of the idea of developing schemas for OS objects? What about schemas for common application-level objects - the idea behind WinFS?

I realize the question needs work, but I hope you get the idea.

Re:Schemas for UNIX

[Anonymous Coward]

> This would be greatly simplified and much smoother if those programs produced structured output rather than raw text.

That's a straw man argument because all of those text files are structured, generally as name-value pairs terminated by newline. Some files also have a scope based upon a keyword name, like X config files. That's *exactly* what XML gives you. There is no semantical difference. Just because you could write
<xml hosts schema url reference blah blah blah>
<host>
<ip address>127.0.0.1</ip address>
<name>localhost</name>
</host> ....
</xml>
doesn't mean you're being better, just more verbose.

Re:Schemas for UNIX

[generalphilips]

Like I said, the question needs work. You're right - that statement about "structured output" is not very good.
My emphasis was on STANDARDIZED formats. Yes, I realize that current config files have structure. However, it's not a straw man argument because I wasn't just talking about config files. I was talking about the bigger picture, which I think you missed.
Okay, X config files are pretty good. What about all the other config files? What about output produced by ps, ls, etc? One of the things that's extremely valuable in UNIX is a programs are reusable, and not special purpose like in Windows. If something is in an hard-to-access format, however, that's going to limit how I can use it.
Imagine for a second that XML was the universal format used in UNIX. Imagine what kinds of programs you could write. Imagine how easy it would be to format the display of information produced by a command-line program, and, at the same time, how easy it would be to write a GUI app that used that same command-line program as a back-end. Imagine how easily your app could read and write config files, without ever worrying that you would break formatting rules. Imagine a shell that understood XML natively, and what you could do in that shell. Imagine everything you could do from the command line, you could do from a GUI (even though you may not want to). Imagine how stark the comparison between Windows and Unix would be then.
I think you are grossly underestimating the value of XML. Why are there such things as XPath, XSLT, XML Schema, etc? Information model, information model, information model.
Thanks for pointing out, though, a problematic statement in my question.

Power

[schon]

True. CLI is the equivalent of spoken or written language, and the GUI is the equivalent of pointing at something and grunting.

Spoken/written communication is much more powerful (easier, faster, more effective) when both parties understand the language, and the idea is a complex one ("I would like a job at your pie shop.")

Rudimentary communication is easier with point-and-grunt (answering the question "which pie would you like to purchase?" - you point to the one you want)

If the parties don't understand the same language, complex concepts are *much* harder. Learning to communicate by pointing is easier, but the true power of communication comes from spoken/written languages.

Think I'm wrong? Write a detailed response *without* using your keyboard.

Re:Power

[Anonymous Coward]

Try reading this with a keyboard.

Re:Power

[Anonymous Coward]

Well, I can reply without needing a GUI. What else you got?

Re:Power

[Anonymous Coward]

Dollars to donuts says that you replied by "pointing and grunting" at your full screen console app and did not reply using curl or other commandline app.

Re:Power

[ClosedSource]

"Learning to communicate by pointing is easier, but the true power of communication comes from spoken/written languages."

People can and do communicate quite effectively in sign language as well and it isn't written. By the way, what does a spoken language have to do with the CLI?

Re:Power

[synotia]

I'm not sure that the differences in phylosophy between CLI and GUI can be so easily identified as this. As usual, things are a lot more complicated.

It all comes down to how best to represent the underlying concepts. An extremely flexible way to represent concepts is to describe them using a natural language (like English), but almost always English is overkill, and can often be ambiguous and cumbersome. This is where Mathematicians start to invent notations for certain mathematical concepts like algebra, matrices and so on. These notations can concisely represent instances of these concepts without being ambiguous or verbose. Many other languages and notations have been invented for the many other concepts and calculuses used with computers. One such language is C, another Haskell, and another Bash or TCSH.

Graphical User Interfaces might be viewed as the odd man out here, but they also have cousins like UML (which is also graphical), paint programs like GIMP and Photoshop, and WYSIWIG document editing, among many others. Each of these were designed to best represent their respective underlying concepts, and each have a place and a job to do.

The problem with the standard GUI notation & widgetry is that it is highly overused. This is probably because it was such a good idea in the first place - it's just that some movers and shakers in the IT industry have used it as a silver bullet for all UIs. This is a short sighted, but not surprising. Many of us have become so accusomed to the mighty GUI that it becomes very difficult to think of UI without it. To think what's outside of the box is difficult when you've been living in it for so long.

Re:Darl McBride

[schon]

My money would be on McBride. Not only is he too stupid to know when he's lost, his head is so thick it would be impossible to knock him out.

Hmm maybe the two are related? :o)

1980 Olympics

[Anonymous Coward]

How did it feel to win an Olympic medal in Archery? Could you describe your experience at the 1980's Olympics?

Re:1980 Olympics

[Anonymous Coward]

it felt pretty ok. except i wish i could've gotten some chicks because of it. all i got was this stupid medal.

What does this mean?

[VonGuard]

Failure to do so will result in Dom!

Paste job

[oexeo]

>> Axiom of Transitivity of Inequalities (Rough) translation: The principle that a comparison of X objects can be concluded by observing the inequalities of their methods. Also, why is each word's first character uppercase, did you contrive that using an online thesaurus?

Re:Paste job

[oexeo]

Damn it, that was my first post; I sent it through as HTML formatted, so NL !=

Should have looked like this:

>> Axiom of Transitivity of Inequalities

(Rough) translation: The principle that a comparison of X objects can be concluded by observing the inequalities of their methods.

Also, why is each word's first character uppercase, did you contrive that using an online thesaurus?

Re:Paste job

[oexeo]

Fuck sake! I sent the last one as plain text, but Slashdot clearly don't convert > and < to HTML entities, the post should have read:

ahh, forget it.

Re:Paste job

[Anonymous Coward]

Use the "Preview" button dumbass.

Re:Paste job

[oexeo]

Use the "grammar check", dumbass.

>> Use the "Preview" button, dumbass.

Comma added by me.

Let's get personal!

[TheOtherChimeraTwin]

Boxers or briefs? Er, I mean Emacs or vi? And why?

Yes!

[zogger]

I brought this up recently on another thread, because I am in favor of it also. One of the things I really liked about my previous mac experience. Are you (or anyone) aware of any linux distros that do packaging like this?

Re:Yes!

[mikefe]

The only one who could practically create and maintain a staticly linked Linux distro would be Gentoo.

Ask them, maybe there is an option for that.

Though, every time a library is updated, it will require any program that depends on that library to be recompiled.

You lose a lot of security possibilities when you us static linking also. With run-time linking and dependencies, you only have to change one package to fix a security vulnerability. With static linking, you need to get an update for each of your applications or you have a hole.

If you just had compatible package repositories for everything this problem would be avoided.

Immagine of sourceforge had deb and (rh, suse, mdk, etc) rpm repositories...

Re:Yes!

[glitchvern]

Gobolinux has directories named after programs and keeps all the program's files in a subdirectory named after the version of the program. Various symbolic link tricks are used to allow programs to see other programs' libraries and such. You can just rm -rf a program directory to remove the program as long as no other program depends on it. They include a script to determine a programs dependencies so running that script over all the programs and grepping the output to see if anything depends on a particular program is pretty easy. You can not move programs around like you can on the Mac because linux programs are not really designed that way. They internally refer to the location of other programs and even themselves in all sorts of ways. The only program I am aware of that can be moved like that is OpenOffice.org. There is another one, but I can not think of it right now.

Home PC

[5m477m4n]

What operating system do you use on your home PC and why?

Enough with the complicated questions.

[higgenbottom]

For pity sake try to remember, the guy can only blink once for yes and twice for no.

LLAP

Intellectual Property

[lordvdr]

The US (and to some extent, the EU) are facing mounting issues from Software Patents (The idea of patenting an idea opposed to an implementation). What do you think about the current state of Intellectual Property laws?

What limits should be placed on Software Patents? Should they be eliminated entirely? Should all patents be moved to a trademark like system where if they are not enforced, the holder loses the trademark?

What is the fix and what is needed to make it happen? Will it ever be fixed?

Desktop UNIX

[tty21]

Rob, The media and UNIX/Linux boosters have been developing/promoting UNIX variants as a competitor to MS desktop OS's. Are UNIX and it's variants destined to have a significant share of the commercial desktop, or is it a compromise and will always be a small share (but doing most of the work!) OS compared to the star-studded MS lineup? What further steps do OS developers have to take to bring it to every desktop?

If you could write a new OS from scratch

[LWATCDR]

What would it be like. Would it still be unix like?
What would you write it in? I mean if you had the time, money , and a mandate to create the best OS ever and you did not have to care about backward compatability what would you come up with.

So he's created more than I ever will ...

[WCityMike]

... involved in the development of many modern programming concepts, GUI advances, character sets, operating systems, and life as we know it to exist.

... but what have you done for me lately?

(Alert for clueless mods: joking.)

Re:Why did it take Linux to popularise Open Source

[CondeZer0]

You are either ignorant or trolling, the original Unix developers already developed in the "open source" way long before RMS came along with his GNU crap.

Here is a very good article about the Lions book and how code was shared during the old days of Unix:
http://www.salon.com/tech/feature/1999/11/3 0/lions /

Before there was an Open Source Initiative, before the Free Software Foundation was even a twinkle in St. iGNUcius' eye, Unix hackers were fighting lawyers and commercial interests for the right to copy and distribute source code.

Current Interests

[eog]

You have been active in many areas of computer science from computational physics to user environments to operating systems. What are your current interests?

Does Mark V Chaney write your responses?

[World_Leader]

I heard all your interviews are actually responded to by Mark V Chaney and you only lip-synch. Is this true? And how will we know for sure?

Unix IS object-oriented

[spitzak]

Notice that a lot of calls take an "fd" as the first argument. This is the "object", and the call is the "method".

In fact Unix is far, far ahead of many other systems even today in being truly object-oriented. OO to be useful requires a "common base class", meaning the same method is useful on more than one type of object. In Unix a vast number (for the time) of different objects could be controlled with the read, write, and ioctl methods.

Contrast this to current systems like COM or Corba where the main claim for OO is that it uses C++ syntax: it is much less likely that you can take an arbitrary object and reuse it with code written for a different object. Your code to update the display object cannot be used to write to the file object. In Unix, back in 1970, the code to update the display *could* be used to write to a file!

Re:Unix IS object-oriented

[Anonymous Coward]

Can you inherit from those calls? NO. Can you

Your Biggest Mistake we're still suffering from?

[bsdnazz]

What's the biggest mistake (design, paradigm, API) you've made that we're still suffering from. And I don't mean leaving the e off creat()!

WHAT "recent leaked NT4 source?"

[Anonymous Coward]

Uh, what the frisk are you talking about?

Re:WHAT "recent leaked NT4 source?"

[emil]

some discussion

Re:Why did it take Linux to popularise Open Source

[RLiegh]

Microsoft also shares source; but they're a long, long way from being open source. One notable point that differentiates GNU and modern OSS from the Unix/Microsoft models is that the former explicitly give you the right to modify and distribute the source. The original Unix licences most certainly did not give you the right to redistribute them.

X11 in the future

[moath]

Do you see X-Windows (in whatever form) as a viable platform for GUI technologies in the future or is it approaching the point of diminishing returns?

Re:The answer from the lab

[Anonymous Coward]

it's in the plan9 faq
http://www.ecf.toronto.edu/plan9/plan9faq.htm l#pla n9design

Appology to C-Shell users

[stripyd]

Don't you feel you owe an appology to a decade of 80s UNIX novices for wasted hours before discovering you weren't talking about the shell *they* were using in "The UNIX programming environment"?

Not that I'm still bitter 20 years on...

Re:Appology to C-Shell users

[MGS Hartman]

'cos no-one in their right mind uses csh.

This is more the source of Windows-hatred

[spitzak]

Really, despite how old Unix is, old timers certainly remember Unix very fondly as a huge breath of fresh air: a fast, SIMPLE, complete, powerful, and SIMPLE (!!!!) operating system. The rules for file names were trivial with only the characters '/' and nul having meaning. There were no "file types" or various character encodings. You used the SAME calls to write files as to draw on the screen or punch a paper tape or read a mag tape! You could name a file anywhere in the system with a single string! You wrote a newline to a file by adding a single byte using the same call you used to write letters! And hundreds of other things that seem painfully obvious today, but at the time, compared to what else existed, this was absolutely amazing.

CP/M is mostly based on RSX-11M and RSTS/E from Dec, vastly more complex and painful systems to program. Windows is, despite all claims to the contrary, mostly based on CP/M. That is where drive letters and case-independent filenames and lack of names for devices comes from. Windows is based also on VMS, but VMS is the best-known source of record-based files where the program to copy one file to another was bigger than the entire operating system! This was jettisoned for Windows, but the bad taste remains...

It sure is too bad nothing better has become popular. I had huge hopes for Plan-9. But a good deal of Linux's success can be based on absolute disgust that Microsoft would bring back the horrors of pre-Unix operating systems that we had all hoped were dead and gone. So any hate for Linux is being directed many times more at Windows.

Re:This is more the source of Windows-hatred

[Anonymous Coward]

You are making a ridiculous over-generalization. Quite a few of the "Unix Haters" of the old days are still Unix Haters to this day.

This is largely based on the complete and utter resistance of the Unix Technology Taliban to fix longstanding problems. (For example, there's been many intelligent criticisms of the Unix shell and X11, but the Unix community responded nothing more than more of the same.)

Re:This is more the source of Windows-hatred

[spitzak]

I'm talking about long before X11 existed. Back when Unix was an alternative to VMS. I can state absolutely that EVERYBODY liked Unix much much better. And this was at DEC!

Fess up

[carcosa30]

You stole the source code for UNIX from SCO, didn't cha?

Language based operating systems

[stevedekorte]

Do you see a future for language-based operating systems like the old Smalltalk and LISP machines or the Newton?

Reliability and Security

[starseeker]

I appreciate and largely agree with your presentation on the state of systems research, but one area where there are perhaps both academic and commercial points of interest is the security/reliability question.

(I should point out I consider these two things to be closely related, since a compromised system cannot be considered a reliable one.)

I agree people have largely decided how they want to use computers, but I don't think anyone can dispute that while computers and software today might provide people with the functionality they want they do not do it in a robust manner. I would think there are a great many interesting problems surrounding figuring out how to impliment a proven, uncrackable software system that provides the functionality people want today.

I had hopes that the EROS system might prove to be the first step in this direction, but research on it has slowed to virtually nothing. But in addition to designing from the ground up in a truly paranoid fashion, they used proof logic to verify key mechanisms of their security structure. I know there are distinct limits to this approach, but I have always wondered if, working within these limits, a system that provides today's expected functionality could be created.

(Note when I say functionality I refer primarily to end user applications like the spreadsheet, database, and word processor. Other types of change in the user environment, so long as they are changed for a reason where benefits are real to the end user, are fine.)

Could you comment on this as a possible future direction for computer research?

Re:Why did it take Linux to popularise Open Source

[Anonymous Coward]

Again you are wrong, the original Unix license allowed redistribution of source modifications among Unix licensees.

Not that it matters, because 1) the Unix developers had nothing to do with setting the license 2) the license was mostly ignored and code was shared freely by all Unix hackers.

Even when the lawyers tried to stop it people kept sharing code around anyway, there are various accounts of Bell Labs Unix developers(wont mention names ;)) "forgetting" tapes full of source code after visiting universities; or "accidentally" dropping a case full of tapes at certain places.

If what you want to know is why the legal department of AT&T were such jerks(just like the legal department of any other big company), you probably are better of asking someone else...(or reading about Unix history and informing yourself, there are plenty of speculations about the matter, including by DMR and some BSD guys).

Moving the cursor

[Anonymous Coward]

What the hell is wrong with cursor control using the keyboard. I don't have three hands man!

Re:Your Biggest Mistake we're still suffering from

[Anonymous Coward]

That's easy: UNIX print services.

Re:Why did it take Linux to popularise Open Source

[RLiegh]

No; YOU are wrong, and you even say so yourself:

the original Unix license allowed redistribution of source modifications among Unix licensees.

That is a restricted group; GNU software was the first to specifically allow access to anyone.

That's what that whole "free as in speech" bit was about, in contrast to the "free as in amongst-a-closed-group-of-licensees" culture of the time.

Unix is legion

[DragonHawk]

"unmounting filesystems 'that are in use'

This is mainly an implementation issue; there are many *nix implementations which let you "force" an umount. That will usually nuke whatever process(es) had inodes on that filesystem open, but presumably that is considered an acceptable loss if one is using the "force" option in the first place.

"why the hell there is a 'D' run state that is completely uninterruptable"

D has been recon'ed to mean "uninterruptable sleep". It means that a process has made a syscall, and the scheduler is waiting for that syscall to return before it can wake the process back up. Processes sleep on syscalls all the time. The "uninterruptible" aspect generally indicates part of the system (such as an interrupt handler or bus master DMA hardware) needs to finish something before the syscall can return. For example, maybe a buffer in memory is going to be written to. You can't just kill the process off, then -- it's being used to track that resource.

The details of this are an implementation issue. Traditionally, *nix implementations have tended to schedule nearly all system processing as part of the user process which triggered the need for the system processing in the first place. This has a number of benefits (e.g.: keeps complexity down, uses fewer resources, makes resource tracking and accounting a lot easier). There's no userland API reason for this, though; you could handle everything using separate kernel threads and a lot more interfacing. Whether or not that is a Good Thing depends on the beholder.

"Another thing that I see as "broken" in UNIX is that there is no normal/standardized/sane way of installing software."

These days, when people say "Unix", they generally don't mean the AT&T(TM) Unix(TM) System (or whatever it's called today). They mean systems that derive code from AT&T or BSD Unix, or implement the same set of userland interfaces, or implement POSIX, or otherwise resemble all that. (I use the term "*nix*" for this meaning, avoiding the ambiguous "Unix".)

It is true that there is no standard package system that works over all *nix systems, but you have to look at what a huge field that is. There is also no standard package system that works on both Mac OS X and my Palm Pilot. You're asking a lot, more then human engineering is possible of providing, I think.

*nix actually does remarkably well. Within a particular implementation (say, Red Hat Linux, Debian GNU/Linux, OpenBSD, Sun Solaris), one generally finds there is a "native" package system. Quality varies, of course, but most work well.

Outside of the "native environment", where just about anything goes, the "five-step install" is surprising portable:

tar -xzf package-x.y.z.tar.gz
cd package-x.y.z.tar.gz
./configure
make
make install

Sure, it's not point-and-click, but compare that to, say, Windows or Mac OS, where every version of every programs uses its own install routine, shared system files are routinely trampled, and things are generally messy. Microsoft defines "portable" as "runs on more then one version of Windows". That's a fair bit less then what we have in the *nix world.

Re:Unix is legion

[thinkfat]

Actually, "D" state means that a process sleeps on a wait queue or semaphore and will not be woken up by a signal (signal as in 'kill'). It's not really related to DMA or other interrupt activity. You could also wait 'interruptible' on a queue, meaning that it sleeps (in state "S"), but will be woken up if you 'kill' it (or via alarm(2)), it just means that you need to take care of the cause of your wakeup - it might not be what you actually expected.

Plus, you must be prepared to immediately return to userspace so that the signal can be handled. But, there are situations where you just cannot do that safely. You might be just in the middle of something and cannot roll back the changes you have already done. That's where god invented the 'D' state :)

sleeping in 'D' state is however fatal if the event you're waiting for will never come because of a system failure somewhere else. Imagine you scheduled some I/O and wait for its completion, but the hardware went into some strange state and the I/O cannot be completed. Or the driver is plain buggy, or else...

"why not add a timeout" you might ask, but if you chose to D-sleep there are reasons for it. One of them being that you cannot just cannot return to the calling userspace application unless you completed *something*. A timeout would not be feasible...

More command

[DragonHawk]

I once saw a quote on the origin of the more(1) command, which went something like this:

I took a big risk when I created 'more', as not only was it longer then the traditional two characters, but it was a real English word.

Alas, I am unable to find the exact quote after endless searching with Google.

X?? Pike&Thompson's 8.5 Windowing System

[billstewart]

Rob gave a talk at ?Nashville Usenix showing off the 8.5 Windowing System. One of his comments was that "Ken and I spent 10 years learning what things windowing system shouldn't do and we've written one that doesn't do them." 8.5 (the ".5" is properly the Unicode "1/2" glyph...) is small and blazingly fast, firing up a windowing system in about the time you'd expect a normal carriage return to give you a shell $ prompt back on a 680x0 machine, and the Acme windowing user interface on it was also lean, mean, and very fast.

Re:Why did it take Linux to popularise Open Source

[Anonymous Coward]

Whatever you say, my dear stupid ignorant zealot...

Ever heard about this thing called BSD?

And then there is this thing called GPL that does put _restrictions_ on redistribution, read the god damn license...

Got news for you, the Unix users didn't give a fuck about "Open Source" because that was the way they worked already and they weren't religious zealots and didn't care about licenses, they cared about code... and the development model was infinitely more open than anything comming from the FSF ever was or ever will be.

All in the FSF is controlled by RMS that is a control freak(just ask the GNU/TURD(sic) main developer that was kicked out just for commenting that the GFDL was not free at all, which is basically pointing out the obvious).

BTW, I'd like to see RMS hear you say that he has anything to do with "Open Source".

Get your facts straight before going around lecturing people with nonsense.

How to Return to Small, Clean Kernels?

[billstewart]

I remember Rob and Ken ranting against microkernels back in the day, but Plan 9 was smaller and cleaner than most of its competition, and things that were viewed as "hopelessly bloated" a decade ago are now comparatively small and under-featured. How do we get back to relatively small kernels and fast reliable operating systems without ditching all the useful things that have been added since then? Hardware support for PCs and connected devices including accelerated video cards and USB plugins are critical issues.

Re:Why did it take Linux to popularise Open Source

[Anonymous Coward]

If I ever want to lecture people with nonsense, I'll get you to do it instead. I can tell by your post you're much more adept at spewing nonsense and vitriol than I am.

Database filesystems

[defile]

The buzz around filesystems research nowadays is making the UNIX filesystem more database-ish. The buzz around database research nowadays is making the relational database more OOP-ish.

This research to me sounds like the original designers growing tired of the limitations of their "creations" now that they're commodities and going back to the drawing board to "do things right this time". I predict the reinvented versions will never catch on because they'll be too complex and inaccessible.

Of course, this second system syndrome isn't just limited to systems. It happens to bands, directors, probably in every creative art.

I think what we've got in the modern filesystem and RDBMS is about as good as it gets and we should move on. What do you think?

Re:Database filesystems

[TheInternet]

I predict the reinvented versions will never catch on because they'll be too complex and inaccessible.

Complex from the perspective of the user or developer?

- Scott

Re:Darl McBride

[Forbman]

I think Rob would probably "spot him a 'J'", to which Darl would reply, "I knew that..."

Re:Why did it take Linux to popularise Open Source

[Anonymous Coward]

> That is a restricted group;

Not that restricted. If your organization could afford a minicomputer, they could afford the quite reasonable fee for the AT&T tapes. For many years it was a flat fee for an entire site (like $10K for all of UC Berkeley).

GNU really didn't get going until there was per-CPU and per-user licences for UNIX.

Additions to Unix after it left Bell Labs.

[Anonymous Coward]

What do you think of the various additions to Unix
after it left the Bell Labs- specifically Sys V IPC, STREAMS, Networking(TLI) etc.

Do you think Unix was become better or worse after it
left the Bell Labs?

Re:Additions to Unix after it left Bell Labs.

[MGS Hartman]

it already had streams and networking and NO sysVile IPC.

Standardized interfaces

[ewe2]

Rob, given the glacial POSIX process to standardize system/user administration or even shells, what do you want to see prioritized in this area?

The major vendors are using such interfaces as selling-points, while the open OS's can't agree on such apparently simple matters as the arguments to cat let alone generalized administration.

Carol Vorderman, surely?

[Burb]

Ask any UK TV viewer about the "W-A-N-K-E-R-S" incident on countdown...

Mr Pike,

[Liquid Len]

your campaign seems to have the momentum of a runaway freight train. Why are you so popular ?

Redundant Array of Inexpensive CPUs?

[Cardbox]

Mainstream operating systems were designed when electronics were expensive and programs had to treat the computer system as a shared resource. Hence timesharing, multitasking, shared filesystems, and the rest, with all the combinatorial problems of N programs interacting with N other programs.

Now that CPU-plus-memory is so much cheaper, do you see a phase change coming where it is better/more secure/simpler to have one CPU per application? What impact would this have on operating system design?

fortune cookies

[MGS Hartman]

rob: what's the phrase that used to be used after reading a fortune cookie?

what have we LEARNED? and NOT?

[brre]

What are the top three lessons that the community has learned from Unix? What are the top three lessons that the community has by and large FAILED to learn from Unix?

one of the original Unix gurus said that

[JoeBuck]

One of the Bell Labs Unix pioneers, I forget if it was Dennis Ritchie or someone else, when asked what he would do differently if he redid Unix, said "I'd spell creat() with an e at the end."

Would you consider an OO approach?

[Anonymous Coward]

If you were to re-do the system today, would you base it on C++ instead of C? I see lots of opportunity for inheritance and virtual functions in the base code. There's plenty of copy-paste going on just to create a driver that's only slightly different from another existing driver.

Glenda

[airwick]

Shouldn't Glenda (http://plan9.bell-labs.com/plan9dist/glenda.html) be an Angora rabbit (http://www.angorarabbit.com/)?

Bell Labs.

[moljnar]

...was the birthplace of a lot of things fundamental (transistors, lasers, operating systems worth a darn). where are these things going to come from now that no single entity seems to think it's worth while spending money on activities not directly related to a bottom line?