Motorola not getting the deal for supplying the CPU for the IBM PC.
Rather than having a nice 68000 based system, with flat memory model from the start, we end up with a segmented, "640k should be enough for anybody" pile of crap 8086 cpu from intel, which stifles development for years:P
If anyone in the upper echelons of the Debian team is reading... its getting on near 1 year since slink was released, and the distribution is getting HUGE.
1 *year* in open source software development terms is an eon.. and I can't see that using the current model, Debian 2.3 (or 3.0, or whatever woody is going to be) will take any less time to stabilize.
Now for something hopefully constructive;)
What I would suggest, is that maybe the *core* distribution be shrunk somewhat, and the gigabytes of other additional software be part of maybe an "applications" distribution, or even broken down further - ala slackware almost;)
Example (pretty much in order of importance):
Core system: The kernel, C library, the base networking tools (PPP, telnet, ping, traceroute, etc), vi, pretty much all the stuff which is installed before dselect (or apt in future) is launched
Network tools: Ngrep, netcat, tcpdump, etc etc.
APPS: The various shell only tools... editors like Emacs, joe, pico.. mail readers etc.
X11: Maybe the base X system, and possibly include the various window managers, and desktop environments (KDE, gnome).
XAPPs: All the associated X cruft that people seem to love (x toys etc), and applications.
And maybe, a contrib section, where things can go without going into the main distribution straight away.
My point is that the main distribution is getting huge, and to get it stable with new things being added all the time is only going to get harder.
I know there is some sort of package voting system (i dont think i paid much attention to it seeing as i dont have a permanent internet connection), maybe the results from this could establish which parts of contrib go into the NEXT distribution as one of the base sections.
Thats pretty much it...
Just as a note, I am one of the people who happily runs unstable, and does not particularly care about the "stable" release. I have had no major problems even running unstable for the last 12 months, and will most likely continue to run "unstable":)
However, i can see that the "unstable" name is keeping people away... and that more frequent 'stable' releases would help to promote the distribution better;)
I used Redhat from versions 4.1 - 5.2, and had a period of time during which i used both quite a bit (Redhat 4.2 and Debian 1.3 I believe). I have switched to Debian exclusively since Hamm (v2.0).
For me, the benefits which made me switch were:
1. The package management system.
.deb packages are much more versatile than rpms. You don't just have dependancies, you can have packages which "suggest" other packages. This tends to lead to packages which REALLY only depend on what they need for core functionality, as opposed to blah.rpm being compiled with every option enabled and requiring 35 different rpms to go with it.
Also, the new apt package management tools just rock. You no longer have to search for packages to satisfy dependancies, or even select what to download and install. Simply do:
apt-get install blah
and apt will work out what packages are required, give you a list, give you the amount of disk space required, both for downloading and after installation, and then ask for confirmation. Once you hit yes, you can happily leave it to download whatever and then come back when it is finished to configure.
2. File system structure
This is hard to put a finger on, and switching from Redhat will take some getting used to, but in my mind, Debian just has a much 'cleaner' file system layout. If you are after docs for package X they will be in/usr/doc/x pretty much without fail. ALL config files are in/etc/blah. The init scripts are easier to understand by your average human.
3. Central location for pretty much all packages.
With the exception of KDE, every single package on my system was downloaded from one of 2 sites. Either ftp.debian.org, or ftp.au.debian.org (the aussie mirror;)
Also, a point to note is that to upgrade Redhat cleanly, I have had to reinstall for every new release.
I have installed debian ONCE on my machine in the last 18 months or so, and it is current, as of last night:)
Also interesting to note that Corel is based on Debian, which should prove interesting in the coming 12 months or so;)
When people choose Intel over competitors for non technical reasons it is not always marketting. Many people don't choose AMD over Intel for any reason at all. They just choose the processor with the fastest Mhz.
...
And THIS, may just be Intel's downfall. Once they bring in IA-64, they will not be able to rely on Mhz alone. They will be forced to say "sure, this is 800mhz like your p3, but its a new architecture!", which may just make Compaq/AMD/whoever 's claims of being faster clock for clock "valid" to the layman.
hrm... maybe this is why they have been pushing IA-32 for so long without bringing in a new architecture. With the 386/486 etc it was easy, they just say "ahh well this is a *4* 86.. its much better, bigger number". With "Itanium" vs "Pentium" they kinda lose that..
So, what if it IS Microsoft's fault for making their software so "dangerous" to use?
They may well be to "blame", but that doesn't mean that they can be dragged into court over it.
Anyone who has actually READ the Microsoft EULA should know that they take no responsibilty for any use or misuse of their software, and do not guarantee it to be fit for any particular purpose.
Open source is in exactly the same position. There are NO guarantees on quality, and no-one (who has the sense to include a disclaimer) can be dragged into court over it failing to perform.
The best thing that could happen over this is that people become more aware of just how flaky the "license agreement" is that Microsoft offer for software that they have sold to you. Maybe some people will wake up and realize that Microsoft aren't the dependable solution that they make themselves out to be...
I dunno if you could do this with the current "client" versions of quake out there, but what the server could do (for future), is this:
When a client connects, do a checksum of the connecting binary of quake (1/2/3/whatever) and compare that with a database of "known good" binaries. I don't see that there can be *that many* different checksums to support...
If it doesn't match, don't allow the client to connect, until the version of the client they are using to connect with can be verified. It's up to the serveradmin as to what binaries he will allow.
This would still keep the source available, but any modified binaries could be excluded from online competition unless they have been verified as "legal" (ie they dont cheat).
regarding the comment that vi has a sucky interface...
I originally attempted to use vi without any sort of instructions, and without fully understanding the philosphy behind it. as a result I started using pico, and then joe, as my unix editors of choice. then, one day, I started doing sysadmin work, and from time to time vi was the ONLY editor available (other than ed, but that doesnt count:).
once you get the hang of it, vi has a VERY nice interface... most commands you can do without your hands hardly leaving the home row...
vi also shows you exactly what you have typed in quite nicely as well.. if you scroll the cursor over a tab for example, it will jump the whole tab space.. useful for editing makefiles etc where there is a difference in meaning between tab and space...
all i can recommend is to give vi a good chance... check out the vi manual (try search for it on google), and edit a couple of files... it grows on you.
your box is most likely stable because it is only a simple fileserver.
install a web server, sql server or proxy server on it and see how it fares...
i hate it. i had to go on site *4* (yep, count em) times yesterday because a particular client's machine kept dropping its connection, and of course MS-DNS server that they were using does not recover.. ever. ended up just disabling DNS on that machine.. was just running it as a cache.. its clients were using socks anyway..
ms proxy often has to be restarted, sometimes it locks up solid and you have to reboot the box... administrator logged in, try to kill it "sorry permission denied". wtf?:P
smash (i hate it, i hate it i hate it...)
ps.. i work in an isp. we do not personally use NT for anything important, other than firewalled office print/file server... our ISP network is 75% linux, 25% solaris:)
id like to see you write a significant program directly using a hex editor though.. i'm not saying its not possible...(ive done it with dos... program to call software interrupt 20 and reboot machine - 2 bytes:) ) but people went out and wrote assemblers (in binary:P) for a reason;)
smash (ie, it was a royal pain in the ass to write/debug:P)
Provide binary/source or binary only with the product, make the source available upon request, and (and this is the important bit) CHARGE source code SUPPORT.
if someone "breaks" it by messing with the source code, then charge them the appropriate fees for support, by the hour.
doesn't seem like a problem to me.. in the documentation state something along the lines of "we will provide free support only for binaries supplied with this product. support for user-compiled binaries is available upon request, at additional cost" or similar. or just state that you straight out don't support user compiled binaries (depends if you want to make money out of it).
this works with redhat.
im sure if you try to recompile your redhat distro with the supplied source CD, and ring up redhat saying something to the effect of "whats a makefile? how do i configure it???" they will not provide that sort of support for free;)
the whole idea about making money out of open source is charging for *consulting*, but not the actual product.
Its probably true that flight sims are "worse" than first person shooters with regards to training people to be able to kill, however i dont see how this is relevant...
the thing is, your average kid can't just go and find an F-15 to run riot with. your average kid shouldn't be able to just go and find a handgun to shoot people with either..
there are terms for people who are unable to tell fantasy from reality - nutcase, lunatic, etc. they are usually locked up.
all the better to make a bigger cluster in a smaller space:P
seriously tho..
the rumors of emulating other CPUs in hardware sounds very cool. I only hope that if they are true, that you can emulate more than 1 instruction set concurrently...
imagine being able to run Linux (or whatever you decide) on it "natively", and then have an x86 application, a couple of Sparc applications, and a few old 16/8 bit emulators running at the same time, all doing the CPU emulation in hardware:)
Linux as it is, is fairly cross-architecture... with hardware emulation of different instruction sets, this would mean binary cross platform compatibility?
Netscape is kindof irrelevant these days, they may be dead (and as far as I am concerned, they have been dead for a good 12 months), but mozilla is alive and well.
I tried out one of the nightly builds a couple of days ago, and it really has come a long way... even since milestone 9:)
at this rate, we should have a decent usable browser in early 2000:)
Hercules were a big name back in the mid eighties, for their "Hercules graphics adapter.
Back then, the choices were MDA, CGA (EGA/VGA later), or Hercules.
MDA and CGA were 320x200 resolution, CGA could do this in 4 colors. In mono, CGA could do higher res (640x200? 640x480?).
Hercules was a mono only card, however it could do some huge resolution (at the time) such as 768x???, which made it THE choice for people who were working with grapical interfaces.
Unfortunately, EGA/VGA cards could not do hercules screenmodes, so it died a slow and lingering death..
how times change...
smash(some of the screen resolutions may me off... but i believe they are approximately correct:)
A note on the linux driver faq at creative mentions that the driver will have SMP support when it hits 1.0... my guess is that they just want to get a stable driver working (which would no doubt be easier without worrying about SMP) and then make it SMP safe...
Of course, when the kernel patches arrive, people will be free to make it SMP safe themselves:)
I DO happen to remember on the kernel mailing list.. when Creative was first talking about linux driver development (and they were looking for a couple of experienced kernel hackers), that they said the drivers would be open-sourced once there was a working example driver out there...
Slashdot Mirror
Motorola not getting the deal for supplying the CPU for the IBM PC.
:P
Rather than having a nice 68000 based system, with flat memory model from the start, we end up with a segmented, "640k should be enough for anybody" pile of crap 8086 cpu from intel, which stifles development for years
hrm...
smash
If anyone in the upper echelons of the Debian team is reading... its getting on near 1 year since slink was released, and the distribution is getting HUGE.
1 *year* in open source software development terms is an eon.. and I can't see that using the current model, Debian 2.3 (or 3.0, or whatever woody is going to be) will take any less time to stabilize.
Now for something hopefully constructive
What I would suggest, is that maybe the *core* distribution be shrunk somewhat, and the gigabytes of other additional software be part of maybe an "applications" distribution, or even broken down further - ala slackware almost
Example (pretty much in order of importance):
Core system:
The kernel, C library, the base networking tools (PPP, telnet, ping, traceroute, etc), vi, pretty much all the stuff which is installed before dselect (or apt in future) is launched
Network tools:
Ngrep, netcat, tcpdump, etc etc.
APPS:
The various shell only tools... editors like Emacs, joe, pico.. mail readers etc.
X11:
Maybe the base X system, and possibly include the various window managers, and desktop environments (KDE, gnome).
XAPPs:
All the associated X cruft that people seem to love (x toys etc), and applications.
And maybe, a contrib section, where things can go without going into the main distribution straight away.
My point is that the main distribution is getting huge, and to get it stable with new things being added all the time is only going to get harder.
I know there is some sort of package voting system (i dont think i paid much attention to it seeing as i dont have a permanent internet connection), maybe the results from this could establish which parts of contrib go into the NEXT distribution as one of the base sections.
Thats pretty much it...
Just as a note, I am one of the people who happily runs unstable, and does not particularly care about the "stable" release. I have had no major problems even running unstable for the last 12 months, and will most likely continue to run "unstable"
However, i can see that the "unstable" name is keeping people away... and that more frequent 'stable' releases would help to promote the distribution better
smash
I used Redhat from versions 4.1 - 5.2, and had a period of time during which i used both quite a bit (Redhat 4.2 and Debian 1.3 I believe). I have switched to Debian exclusively since Hamm (v2.0).
/usr/doc/x pretty much without fail. ALL config files are in /etc/blah. The init scripts are easier to understand by your average human.
;)
:)
;)
For me, the benefits which made me switch were:
1. The package management system.
.deb packages are much more versatile than rpms. You don't just have dependancies, you can have packages which "suggest" other packages. This tends to lead to packages which REALLY only depend on what they need for core functionality, as opposed to blah.rpm being compiled with every option enabled and requiring 35 different rpms to go with it.
Also, the new apt package management tools just rock. You no longer have to search for packages to satisfy dependancies, or even select what to download and install. Simply do:
apt-get install blah
and apt will work out what packages are required, give you a list, give you the amount of disk space required, both for downloading and after installation, and then ask for confirmation. Once you hit yes, you can happily leave it to download whatever and then come back when it is finished to configure.
2. File system structure
This is hard to put a finger on, and switching from Redhat will take some getting used to, but in my mind, Debian just has a much 'cleaner' file system layout. If you are after docs for package X they will be in
3. Central location for pretty much all packages.
With the exception of KDE, every single package on my system was downloaded from one of 2 sites. Either ftp.debian.org, or ftp.au.debian.org (the aussie mirror
Also, a point to note is that to upgrade Redhat cleanly, I have had to reinstall for every new release.
I have installed debian ONCE on my machine in the last 18 months or so, and it is current, as of last night
Also interesting to note that Corel is based on Debian, which should prove interesting in the coming 12 months or so
smash
5 years?
the 386 came out in 1985 i believe (1986 at the latest), and 14 years later, windows 9x is still largely *16* bit.
i wouldnt guess that 32 bit desktop software will die until at least 2010...
smash
When people choose Intel over competitors for non technical reasons it is not always marketting. Many people don't choose AMD over Intel for any reason at all. They just choose the processor with the fastest Mhz.
.. its much better, bigger number". With "Itanium" vs "Pentium" they kinda lose that..
...
And THIS, may just be Intel's downfall. Once they bring in IA-64, they will not be able to rely on Mhz alone. They will be forced to say "sure, this is 800mhz like your p3, but its a new architecture!", which may just make Compaq/AMD/whoever 's claims of being faster clock for clock "valid" to the layman.
hrm... maybe this is why they have been pushing IA-32 for so long without bringing in a new architecture. With the 386/486 etc it was easy, they just say "ahh well this is a *4* 86
smash
im not sure where this originally came from, but:
:P maybe it would be because you can do a lot of damage with a car, and learning to drive is the most effective way of preventing that.
"A certain percentage of computer users will never truly understand computers. Why try to force
them to learn?"
hrm.. this is like saying "A certain percentage of automobile drivers will never know how to drive. Why try to force them to learn?"
gee i wonder
same thing with a computer. if you want to be ignorant, expect people to take advantage of you. applies to everything in life really..
smash
So, what if it IS Microsoft's fault for making their software so "dangerous" to use?
They may well be to "blame", but that doesn't mean that they can be dragged into court over it.
Anyone who has actually READ the Microsoft EULA should know that they take no responsibilty for any use or misuse of their software, and do not guarantee it to be fit for any particular purpose.
Open source is in exactly the same position. There are NO guarantees on quality, and no-one (who has the sense to include a disclaimer) can be dragged into court over it failing to perform.
The best thing that could happen over this is that people become more aware of just how flaky the "license agreement" is that Microsoft offer for software that they have sold to you. Maybe some people will wake up and realize that Microsoft aren't the dependable solution that they make themselves out to be...
just me
smash
I dunno if you could do this with the current "client" versions of quake out there, but what the server could do (for future), is this:
When a client connects, do a checksum of the connecting binary of quake (1/2/3/whatever) and compare that with a database of "known good" binaries. I don't see that there can be *that many* different checksums to support...
If it doesn't match, don't allow the client to connect, until the version of the client they are using to connect with can be verified. It's up to the serveradmin as to what binaries he will allow.
This would still keep the source available, but any modified binaries could be excluded from online competition unless they have been verified as "legal" (ie they dont cheat).
smash
oooh... so i could make a stack of them and have a beowulf cluster? :)
hehe.. sorry...
smash
maybe they should look at:
punched cards.
heh.
smash
regarding the comment that vi has a sucky interface...
:).
I originally attempted to use vi without any sort of instructions, and without fully understanding the philosphy behind it. as a result I started using pico, and then joe, as my unix editors of choice. then, one day, I started doing sysadmin work, and from time to time vi was the ONLY editor available (other than ed, but that doesnt count
once you get the hang of it, vi has a VERY nice interface... most commands you can do without your hands hardly leaving the home row...
vi also shows you exactly what you have typed in quite nicely as well.. if you scroll the cursor over a tab for example, it will jump the whole tab space.. useful for editing makefiles etc where there is a difference in meaning between tab and space...
all i can recommend is to give vi a good chance... check out the vi manual (try search for it on google), and edit a couple of files... it grows on you.
smash
So the license states that I am not allowed to download it if i am under 18?
:)
If i am under 18, I am not legally bound by that license...
So... i just download it anyway, seeing as, legally, there is not a thing they can do about it.
I am sure the board if film/tv/etc ratings has not classified this distribution as 18+ so..
Where's the problem?
smash
your box is most likely stable because it is only a simple fileserver.
:P
:)
install a web server, sql server or proxy server on it and see how it fares...
i hate it. i had to go on site *4* (yep, count em) times yesterday because a particular client's machine kept dropping its connection, and of course MS-DNS server that they were using does not recover.. ever. ended up just disabling DNS on that machine.. was just running it as a cache.. its clients were using socks anyway..
ms proxy often has to be restarted, sometimes it locks up solid and you have to reboot the box... administrator logged in, try to kill it "sorry permission denied". wtf?
smash (i hate it, i hate it i hate it...)
ps.. i work in an isp. we do not personally use NT for anything important, other than firewalled office print/file server... our ISP network is 75% linux, 25% solaris
.. something that compiles :)
beta = something which is somewhat functional and can be used fi you dont mind the odd crash or misfeature..
smash
if you do that, the binary IS the source ;)
:) ) but people went out and wrote assemblers (in binary :P) for a reason ;)
:P)
id like to see you write a significant program directly using a hex editor though.. i'm not saying its not possible...(ive done it with dos... program to call software interrupt 20 and reboot machine - 2 bytes
smash (ie, it was a royal pain in the ass to write/debug
Provide binary/source or binary only with the product, make the source available upon request, and (and this is the important bit) CHARGE source code SUPPORT.
;)
if someone "breaks" it by messing with the source code, then charge them the appropriate fees for support, by the hour.
doesn't seem like a problem to me.. in the documentation state something along the lines of "we will provide free support only for binaries supplied with this product. support for user-compiled binaries is available upon request, at additional cost" or similar. or just state that you straight out don't support user compiled binaries (depends if you want to make money out of it).
this works with redhat.
im sure if you try to recompile your redhat distro with the supplied source CD, and ring up redhat saying something to the effect of "whats a makefile? how do i configure it???" they will not provide that sort of support for free
the whole idea about making money out of open source is charging for *consulting*, but not the actual product.
just me..
smash
Its probably true that flight sims are "worse" than first person shooters with regards to training people to be able to kill, however i dont see how this is relevant...
the thing is, your average kid can't just go and find an F-15 to run riot with. your average kid shouldn't be able to just go and find a handgun to shoot people with either..
there are terms for people who are unable to tell fantasy from reality - nutcase, lunatic, etc. they are usually locked up.
cool, mobile CPUs...
:P
:)
all the better to make a bigger cluster in a smaller space
seriously tho..
the rumors of emulating other CPUs in hardware sounds very cool. I only hope that if they are true, that you can emulate more than 1 instruction set concurrently...
imagine being able to run Linux (or whatever you decide) on it "natively", and then have an x86 application, a couple of Sparc applications, and a few old 16/8 bit emulators running at the same time, all doing the CPU emulation in hardware
Linux as it is, is fairly cross-architecture... with hardware emulation of different instruction sets, this would mean binary cross platform compatibility?
smash
.... that Netscape is not Mozilla :)
:)
:)
Netscape is kindof irrelevant these days, they may be dead (and as far as I am concerned, they have been dead for a good 12 months), but mozilla is alive and well.
I tried out one of the nightly builds a couple of days ago, and it really has come a long way... even since milestone 9
at this rate, we should have a decent usable browser in early 2000
smash
you dont necessarily need all the stuff than init runs to use your machine ;)
/bin/sh will do ;)
if all I want to use is echo to write and read text files, then the kernel +
smash
the java virtual machine is not an OS.
it runs on your OS to simulate a machine which does not exist. in theory you could run another OS by using the java virtual machine.
smash(thats my take)
if we go to one extreme, we could say that the operating system is the one thing which is required before any applications will run.
/bin/sh
:)
this basically narrows down to the kernel.
i personally count "operating system" to mean the kernel + a shell.
in the case of windows 95/98, this means dos.
in the case of windows nt 4.0, this counts the gui i guess.
in the case of linux/unix, i would say a minimalistic operating system is kernel +
smash (operating system = software required to "operate" your system this just mean device drivers + kernel?
ahh young grasshopper...
:)
Hercules were a big name back in the mid eighties, for their "Hercules graphics adapter.
Back then, the choices were MDA, CGA (EGA/VGA later), or Hercules.
MDA and CGA were 320x200 resolution, CGA could do this in 4 colors. In mono, CGA could do higher res (640x200? 640x480?).
Hercules was a mono only card, however it could do some huge resolution (at the time) such as 768x???, which made it THE choice for people who were working with grapical interfaces.
Unfortunately, EGA/VGA cards could not do hercules screenmodes, so it died a slow and lingering death..
how times change...
smash(some of the screen resolutions may me off... but i believe they are approximately correct
A note on the linux driver faq at creative mentions that the driver will have SMP support when it hits 1.0 ... my guess is that they just want to get a stable driver working (which would no doubt be easier without worrying about SMP) and then make it SMP safe...
:)
Of course, when the kernel patches arrive, people will be free to make it SMP safe themselves
smash
I DO happen to remember on the kernel mailing list.. when Creative was first talking about linux driver development (and they were looking for a couple of experienced kernel hackers), that they said the drivers would be open-sourced once there was a working example driver out there...
:)
that was the gist of it anyway...
smash(just bought an SB-Live last week too...