Slashdot Mirror
Benjamin Herrenschmidt On PPC/Linux, Apple and OSS
MacBoy writes "The folks over at ResExcellence have a great interview with Benjamin Herrenschmidt, kernel guru for the PPC/Linux camp. He offers up some history of Linux on Apple and PPC hardware, and some discussion on Apple's current and past contributions to the open-source and Linux communities. He makes some interesting comparisons of Linux on PPC vs. Intel hardware, such as the ease of getting important patches into the kernel on PPC compared to Intel. It's an interesting read, especially if you are amoung the many who covet the new Dual-CPU GHz G4 Macs and want to know a little more about the PPC/Linux community."
PPC/Linux seems like an ambitious and technically interesting project, no doubt. Personally, though, I don't seem to understand the purpose. MacOs X is based on BSD, so you get all the nice Unix-like server features... and a decent GUI, something that Linux has never had. Are there some other advantages of Linux/PPC that I'm missing?
If guns kill people, then CmdrTaco's keyboard misspells words.
sorry, that was a test post for the accnt I just created. I love apple, and am looking forward to years of fun with my new imac
"someone should make a hot air balloon that is shaped like a giant vagina". --Bill Clinton
How does this get a score of 1 ?
why the fuck do you care?
Hey this is offtopic what point are you trying to prove? WE don't need that here please that that elsewhere.
--
WHO ATE MY BREAKFAST PANTS?
how do you "that" that?
"someone should make a hot air balloon that is shaped like a giant vagina". --Bill Clinton
I was going to ask if anyone has Linux running on the new dual G4's, but then I found this cool link
http://linuxppc.org/hardware/dualg4/
Anyone have benchmarks (under Linux) for the dual G4's and corresponding PCs for:
- compiling the kernal
- GIMP
- Quake1/2 benchmarks
I'd love to get the "other endian machine" and compile some game code on it. Anyone have any recommendations? How did id do the Quake3 port to Macs ?
"PC Expert - Mac newbie"
Sorry, sir. Didn't mean to make anyone mad.
"someone should make a hot air balloon that is shaped like a giant vagina". --Bill Clinton
sorry that that should be take that. and another thing we could use a guy like you.. ON MAC SLASH!
--
WHO ATE MY BREAKFAST PANTS?
Sence the site seems /.ed; Here's the article:
An Interview with Benjamin Herrenschmidt, PPC/Kernel Hacker.
by Michael Coyle
In the early days of PPC/Linux, it was extremely difficult to boot into the alternative OS. It required a three-finger keyboard command to boot into Open Firmware and then typing several lines of cryptic text.
Fortunately, thanks to work the of Benjamin Herrenschmidt and his free BootX Control panel, booting Linux in Old World Macs became much easier. It opened Linux to a large number of average Mac users. Today, several years later, BootX has been replaced by yaboot and Ben has moved on to become a major contributor to the PowerPC branch of the Linux kernel. Running Linux on a Powerbook or iBook with support for power management and accelerated graphics wouldn't be possible his efforts.
Born in Selestat, France twenty-nine years ago, Ben had his first computer at 8 years old (a ZX81) and by the age of 13 he was enrolled in his first "Technology Class". You wouldn't call him the "Teachers Pet", more likely a "Teachers Pest" since the instructor soon asked young Benjamin run the course for him.
His first exposure to the MacOS came when he ran a Mac emulator on Atari ST and Amiga computers. By college, Ben traded his Amiga computer for his first Mac. At that time, he had no interest in Linux, but enjoyed writing for the lower level bits and bytes. He adds,
"Unlike many other Linux hackers, Unix is a late discovery for me. After I sold my Amiga, I had the chance to have access to a Mac and started learning MacOS programming. I started doing that professionally after spending a couple of years having fun and drinking beer at the university. I quickly became interested specifically in device drivers and other kind of low level programming, and that's probably what led me to Linux."
Benjamin installed MkLinux on his Wallstreet PowerBook, but there was no Sleep support at that time. He contacted the maintainer of the PPC branch of the Linux kernel, Paul Mackerras, and asked permission to use his Linux/PPC PMU (Power Management) source code as the basis for an MkLinux version. Ben never finished the MkLinux version, but this introduction to Paul and PPC/Linux led to his switch from MkLinux to the PPC/Linux kernel where he rolled up his sleeves and dug in, fixing whatever low level drivers got in the way of enjoying Linux on his Powerbook. But what let to the development of BootX?
"At one point, I got fed up of floppy booting and used a code snipped I found on usenet that showed how to switch the CPU into supervisor mode. I used that to write BootX. That's how, slowly, I started using my knowledge of Macs and my MacOS hacking skills to work on improving Linux on Macs. But I've never been a Linux _user_, I'm still learning "newbie" tricks everyday. When it comes to userland, I believe I can't even write moderately complex shell scripts."
One advantage to being a developer on the PPC code base is how quickly your patches are accepted into the official Linux kernel. In the x86 world, there are substantially more developers trying to get their snippet of code added to the kernel, but in the PowerPC world, the pond is a lot smaller and it's easier to be a "Big Fish". While PPC is a smaller community, it's not limited solely to Apple hardware.
"The PPC tree has it's own complexities, mostly because we have to deal with a large set of very different machines and CPUs. From the PowerMacs, PRePs & CHRP, the 4xx and 8xx based embedded platforms, and even IBM pSeries and iSeries servers."
Because the source code is often available for Linux programs, the platform disparity that exists between MacOS and Windows is diminished. If a software developer compiles his project for the x86 architecture, the user of a PPC computer can simply compile the source on his own computer. Unfortunately, their are sometimes problems with the compilation. Are x86 developers reluctant to modify their code to accommodate PPC users, is Linux another OS where PowerPC users are given second class status?
"Well, except for some rare cases, I've never seen maintainers reluctant to merge in things like endian fixes. It's a bit more difficult when it comes to someone saying, "Hey, add that ugly hack that makes it work for me".
But Benjamin, do you make money writing Open Source software?
"So far, all my Linux work has been spare-time volunteer activity. I have a job in a nice small company where I mostly do MacOS software. Our company is evolving though, and we are now working on some embedded systems using Linux/PPC on IBM 405GP CPU. I do it (Linux/PPC Development) for fun and because I learned a lot of things doing it.
I quite like the concept of Free Software and am happy to put my work under the GPL umbrella. I believe at some point, computer science is like science in general, information and techniques have to be shared to benefit the most people and evolve.
I also feel these technologies have too important a position in our society, too many things depend on them, to safely leave them in the hands of an opaque monopoly. Free Software (or Open Source in this case) is a kind of guarantee that the technology you rely on can be trusted."
It seemed Apple would embrace Linux. there was even a ppclinux.apple.com website for awhile, but the site has disappeared as have most mentions of Linux from the Apple site. The boot process for MacOS X seems very similar to what happens in BootX and yaboot for Linux. Did you ever work for Apple? What is your opinion on how they view Linux?
"I did not work for Apple, but I used to (and still) know some Apple engineers. Apple use to have a "Linux technology manager", but I'm not sure I fully figured out what his role was, but he tried hard to get us hardware specifications (and mostly failed).
Apple's BootX and Linux BootX are two completely different things. yaboot is similar, but they were developed independently and share no code. There has been no collaboration here.
It's difficult to decrypt the reasons Apple does things. They didn't open source driver code for some third party hardware, but that is understandable. They might simply not have had the right to do so. Regarding drivers for Apple's own hardware, they did indeed "hide" a couple of them. What comes to my mind right now are the PMU driver (Power Management Unit) and the Firewire controllers driver (OHCI and Lynx chipsets)."
Still, by making Darwin, the core of MacOS X, more open, Apple must be helping the PPC/Linux group, even if inadvertently.
"Darwin code has been incredibly useful for some things. There is no documentation on the various Apple chipsets, so we have to "discover" them by either looking at the OF code (Open Firmware), disassembling MacOS bits, or reading Darwin drivers.
I have not used Darwin as a distro however. I do use MacOS X at work, and I quite like it. I have a machine running Linux 100% of the time, and one running MacOS (9 or X). In fact, I also run MacOS 9 in MacOnLinux on the Linux machine right now to write this email using CTM softwares excellent PowerMail"
PPC/Linux on a new Ti Powerbook
(Click for full image in a new window) It's possible Apple no longer actively supports Linux because the OS is multi-platform. In the final analysis, Apple is a hardware company. They leverage the MacOS to move CPUs. Since Linux runs on a variety of platforms, most costing less than Apple's hardware, support from Apple for Linux only increases the probability of migration to the x86 platform.
Still, it would be a mistake to think PPC/Linux lags behind the MacOS. It has always been fully multi-threaded and preemptive, features lacking in Classic MacOS, and Linux's hardware support exceeds MacOS X is some areas, PCMCIA and printing are two that come to mind.
Fortunately, the Mac/Linux community has volunteer coders, such as Ben Herrenschmidt, whose efforts keep an inexpensive and open source alternative to the MacOS available on PowerPC hardware.
Despite the early antagonism between the league for programming freedom and Apple, Macs and Unix went together pretty well. Long before samba, 680X0 Macs were sharing files quite nicely with unix servers. In many disciplines at universities, unix was used for heavy lifting, and macs were used for light word processing and making drawings.
Linuxppc has benefited from the generally better hardware on the PowerPC (higher quality and reliability, and less diversity). Also, for much of the '90s, the PPC chip was faster for many things than ix86. Recently, I had a 90 MHz powermac running KDE, xemacs, and Netscape without a hitch for months without any hiccups (a little slow but definitely useable). Under MacOS the thing would lock up several times a day. OS X is not an option for these older macs.
My old Mac died but I hope to get an ibook soon and put Linux on it. By the way, I wish there was something like fips for shrinking Mac partitions for people that want to dual boot.
The lmbench numbers show that linux is significantly faster for certain operations. I have a copy of the results that I pointed to in an old post over here. Linux spanks OS X. It's a reason.
Other reasons include access to all of the source of your OS and better support for certain things (pcmcia 802.11b card support? Better filesystems. More software already working).
I personally run Debian on my laptop 99% of the time because my environment is the same everywhere, and apt-get kicks ass (fink on OS X is cool, but there isn't as much stuff available).
Me too
Of course, the long-standing debate between Mac and Windows (I use the term generally) users has been that of their respective platform's performance. Sure, Max users complain that Windows is ugly, Windows user complain that Mac OS doesn't have enough apps...BUT, the big ticket is always when they start comparing Intel (or AMD) processors to their PPC competitors.
With Linux on both platforms, I would be interested to see some comprehensive real-world benchmarks comparing the two platforms. Really, I'm rather tired of the "Megahertz-myth" PPC touters and the "RISC sucks" x86 campions arguing which is better without any solid numbers.
-Jayde
What's a sig?
It's so hard to find a good article giving the history of an operating system and comparing it to another. Most of the time, when I find some sort of comparison article, it is either biased and lacking in information, or, comes to an indecisive conclusion.
One thing that has always bothered me though, is, how do two CPUs work together and not fight over whatever it sent to them? The idea of having two minds is more than painful to think about, so how does a computer with two processors work? Anyone have a good article on this?
"No manual entry for woman."
Why bother porting Linux to the PowerPC when
BSD already runs fine on it in the form of
NetBSD?
On one hand, you have them using and contributing to open source. On the other, you have them sending the lawyers after people who try to create a theme engine because they believe it is theoretically possible for someone to use it to violate Apple's trademarks.
You have them giving away developers tools a short while after they send their users firmware upgrades to prevent G3 users from upgrading (without telling the G3 users what the magical firmware upgrade did)
You have them sending firmware upgrades to block the use of RAM that doesn't meet Apple's strict standards, without telling users ahead of time the sticks they spent hundreds on are now utterly useless.
Apple can't decide wether it wants to be free and open, or continue to manifest its worst control-freak tendancies - ones that even rival Microsoft.
Anball, you asked a good question, which is why I have to reply to it, plus i am at work at 1:30 AM and am bored. Dual processing machines do constantly fight over resources, threads, processses and the such. Different SMP capable OS's handle the situation differently, with verying levels of success. However, no multi-processor system seamlessly handles the management of both (or even more) CPUs. That is why a dual Athlon MP whatever is not 200% faster than a single processor Athlon XP, same with PPC, SPARC, etc. This also raises interesting points in kernel design. OS's that are designed to handle all the threading disagreements between processors very well tend to be much less optimized at handling single processing machines. Take Solaris, often called "slowlaris" because of it's poor single processor performance. And take Windows 2k, which perfoms much much worse (comparative to other OSs) in mulitprocessor setups than in single processor setups. SO, to answer your question
"How do two CPUs work together and not fight over whatever it sent to them? The idea of having two minds is more than painful to think about, so how does a computer with two processors work?"
They don't, they are forced to get along and share, but are ultimately selfish and need to be watched at every point to make sure they are not trying to take all of the workload.
Skipp
It's interesting that the article claims better peripheral support in LinuxPPC. I don't know what their data are to support this, but it's certainly true that peripheral support is a huge problem for MacOS X.
Other than that, I don't see a lot of good reasons to mess with LinuxPPC. People are posting saying MacOS X is slow, but I think it would be more accurate to say that certain parts of MacOS X are slow, e.g. Sherlock. But so what? Just use ``locate'' instead of Sherlock.
As far as open-source application software, I find it easy enough to use X Windows when I need to run GIMP or whatever. Why bother with a dual-boot system?
Find free books.
While Mac OS X is very nice on recent machines(it's running quite well on my Lime 266MHz iMac with only 96MB of 66MHz RAM), it's certainly not meant for older machines. Yes, Darwin can run on even older PPCs, and some have even claimed it can outperform Linux in web and file server tasks. And it's even open source.
However, for those who wish to dabble in the *nix world without getting in too deep, a well-designed Linux distribution is probably a better option.
hah
you silly lil' trolls.
you got modded down and the other dude didnt
how GAY. musta been malda and neal
Is it just me, or is neal's obesity really disturbing???
XFree86 isn't 25 years old. "Broken" is relative -- for workstations like mine, there was no issue (except for 2.4.9, which I really would call broken). You had to be running a heavy-duty server with serious I/O load to see the problems (and you *wouldn't* be running OS X for a heavy-duty server), and the Linux VM is not borrowed from BSD.
Next time use lube dood!
The article is not Slashdotted at all. You're a karma whore and a fucking liar.
fuck you!!!!
FUCK!!
FUCK YOU!
you fucking FUCK!!!
hahaha
yes!!!
FUCKER!!!
YOU fucking FuCkEr!!!!!
When (and only when) each instruction takes one clock cycle to execute you can divide each instruction inte several sub-tasks, for example Fetch - Decode - Execute - Store. Now a pipeline can be introduced:
instruction 1 F D E S
instruction 2 . F D E S
instruction 3 . . F D E S
instruction 4 . . . F D E S
In this way work is parallellised. Four instructions are executed in parallell, during 4 clock cycles each, giving a throughput of 1 instruction per second.
CISC basically means that instructions take different number of clockcycles to execute. With such an instruction set you can not use Pipelining.
The i386 cpus today transform the CISC instructions into RISC instructions, and apply pipelining. For the price of a translator the major drawbacks with CISC are avoided.
It is probably true that Intel has to put more research and silicon into a CPU than Motorola to achieve the same performance. But this is not because Intel are stuck with building a CISC cpu - it is because Intel must emulate the i386 CISC instruction set on a RISC cpu core.
Thanks mate... that was useful
With OS X the usefulness of Linux on the PPC platform is indeed brought into question, but article is still a good read.
It's interesting, however, to note that Linux people still miss the whole point of finding out about Mac OS X. Don't like it? Don't use it. But if you're smart you'll develop for hit. Read the O'Reilly Network, or any any other article, for writing in cocoa for MacOS X.
If there's anything missing in the viability of the comments about PPCLinux and its "questionable viability" you could well say the same thing about Linux in general as a desktop platform. It's going nowhere. Genome and KDE will not only never be the finished GUI that OS X is, but they also won't have the installed (and growing) desktop users that OS X has either.
New games are shipping for OS X, not ports, but original games. The recent O'Reilly network article about writing for cocoa represents an idea which any geek interested in making a living doing something besides tweaking SQL databases may want to consider.
Sure, the Mac OS X kernal, Darwin, itself offers much to users. But any Linux geek who likes to write code should give it even more serious consideration, its portability to other platforms, including Linux.
If you're hung up on the idea of free software disregard this, but if you're looking for a good *nix OS check out OS X if you're smart.
As he poured a steaming bowl of hot grits down the front of his pants!
The one problem i personally can see and have experienced is that you cant find that much information concerning Linux on PPC online. Sure, there are some devoted sites line LinuxPPC.org and others, and groups.google.com is always a nice thing, but if you run into *specific* problems (like me, see further down that post) youre in for some serious pain.
I tried several times to put a usable linux distro on my old powermac 4400 (upgraded w/ a G3 procesor card, a voodoo3 2000 and a realtek-chip-based ethernet card).
Xfree will crash at startup, which can be only solved by going online, grabbing some drivers and compiling them into the kernel, which kind of sucks for a newbie - but thats not the point because I can't get online because the ethernet card is not recocnized even if I choose the realtek drivers.
So I did some google research, came up with the mentioned solution for XFree, but not for the ethernet problem.
Another point is that noone in your neighborhood can help because LinuxPPC is not as common as Linux for i386.
I'll try darwin next, but since my machine isn`t officially supported I`m not exactly optimistic about that.
If you're running Debian, you can post to the debian-powerpc mailing list and people will try to help you out. I don't know about the support for other distros, but I've found debian-powerpc to be more than adequate (especially considering the kernel hacker types also subscribe).
I always wondered why the people hacking away at Linux/PPC don't just fork the code base. Honestly, it seems like Linus has little or no interest anyway in the other architectures, and it could easily result in a much better end product.
--saint
Linux is often faster. Just like Intel Linux, it's possible to create a custom stripped down desktop with no fat. I've become accustomed to how things work on my Intel Linux machines and like having almost the same environment on my Pismo Powerbook. I also prefer the GNU utilities over the BSD ones (flame away...it's just a preference...get over it).
OSX is also next to useless on older hardware. We use 233Mhz Beige G3 desktop as an internal server. It's running ssh, Apache-SSL, NTP, mySQL, Apache, and Netatalk. The Apache/mySQL setup powers our troubleticket/inventory system. The Netatalk/Samba combo makes files available to both Mac and Windows clients. It does all of this with very acceptable speed and reliablity. The machine has zero need for a GUI...and doesn't have one. I suppose I could use Darwin but the machine wouldn't do it's work any better and I would have to mess with fussy ports of the daemons. It has full apt-get goodness....I forgot to mention that it is dead easy to admin.
So yeah, there are valid reasons to use Linux on PPC hardware.
As a group, BSD advocates are no more immune to licence zealotry and fanboyism than Linuxers. Anybody remember the picture of the Daemon doing Tux up the butt? How about the tiff the pf author got into with Theo and let's not forget his special "no GPL" license. I've also seen lots of BSD advocates get horked off over GPL forks of BSD code yet be delighted over proprietary forks. I even sometimes wonder if the entire universe would go up in a flash of gamma rays if Brett Glass were to ever shake hands with Richard Stallman.
If every instance of BSD and Linux were exchanged in your post it wouldn't come off any differently.
I had a little trouble getting all of the modules to work with one of his 2.4.17rc kernels. I figured what the hey and emailed him. He straightened me right out. I suppose the world of PPC Linux is small enough that a leading light can have a little time for an end user like me.
I have definitely fallen into this dilema quite a bit, but have yet to feel truly satisfied. I used Macs exclusively about 7 years ago and then found my way to Linux. I loved it immediately. I still use Linux or a *BSD on all my x86 hardware, but right now my iBook runs MacOS X only. I only came back to the Mac because of MacOS X.
;). However, I seem to be wishing for the speed that comes from true open source development as I want to see the changes more rapidly. These changes come quicker, but it is almost always obvious that no one has done any usability or QA testing.
I like the fact that though Apple takes a long time to release system updates, the updates work extremely well and are obviously QA tested (for the most part, ignore the iTunes fiasco
Also, we come to the desktop metephor. I used to use GNOME or KDE in Linux, but these days rely only on Windowmaker or some other trimmed down GUI. This is mostly due to the fact that no matter how good either of the major DE get, they are still not able to compete with the Mac OS as far as usability or design choices. A lot of people say they are just as good, but I suggest that these people don't really know much about UI design as everything Apple does it with concern to the user. i.e. Does a click here require more or less stress to the user? (Carpal tunnel). They examine all aspects and TRULY test usability. Now, this doesn't mean OS X is right yet. It isn't, but the improvements come quickly.
So, well anyways, I still find myself being pulled back and forth, but for right now OS X is winning...
Bill
This is NOT a flame, I repeat, not a flame.
It take some courage and some guts to wipe out a partition where MacOS X has been pre- installed with all the bells and whistles , all the goodies and all the simplicity in terms of usage and connectivity and then install Linux PPC instead.
After shelving about one grant more to get a slower Apple system vs a Pentium or better AMD box , and then trying to make it look like a PC running Linux, you've got to admit that either these people are nuts or heroes.
And all these technies know for sure that Apple's megahertz myth is pure brainwashing for selling embedding cpu at the price of desktops don't they?
So what's left beside the look of the system when you show it to your friends? If you hide the box and let the screen show thru a hole in a cardboard, would they know that you are on a Mac? If at least Linux was easier to use on an Apple hardware than a x86 machine, I would sorta understand. But clearly we are talking about the niche of a niche market here.
The only people I think these efforts bring joy are the hackers like ben who do it for fun. Apple gets more money out of it, they sell platforms at an outrageous price and don't have to provide support for the OS. Even Paul Mackeras, the main PPC kernel maintainer is on IBM payroll and gets big bucks from the patent company to maintain the kernel on PPC.
PPA, the girl next door.
-- I feel better now. Thanks for asking.
Lets forget about OS-X for a moment. What LinuxPPC does is fix the mac by providing a real operating system for it. There are macs where I work, and every time I'm stuck using one I start getting crazy notions in my head about taking a sledge hammer to them all, or finding some other way of destroying them. The reason has nothing to do with the hardware they are running. They're not super fast, but I'm not particularly picky in that area. What bugs me about them is the antique operating system they're running. If I wanted cooperative multi-tasking and no memory protection between apps, I'd run windows 3.1. When I sit down in front of a computer I expect the system to be responsive to me, regardless of what it is doing in the background.
One day about a year ago I loaded up the PPC version of SuSE 7.0 onto one of the macs. Lo and behold!!! A new computer was born! Or at least it seemed that way. The computer actually seemed to run FASTER as well as being infinitely more responsive. I was also a great relief to escape from the Macs GUI, which I find cumbersome. It was like I was trying to run with a cinder block tied to my leg and someone just cut the cord. Unfortunately I eventually had to wipe Linux and reinstall MacOS-9 on it.
Nowadays we have OS-X, which is about 6 years late, but better late than never. Its GUI isn't on par with KDE, but its far better than the previous MacOS versions. Its also responsive! Its great to actually be able to have my computer wait on me rather than the other way around. Its not terribly fast, but I'm not going to complain! Much better that it take longer to do things in the background but be more responsive in the foreground than it lock me out. Hopefully IBM/Motorola/Apple will be able to push the PPC architecture further in terms of clock rate.
Then of course there are the standard form-factor PPC motherboards that will soon be available. Whether they will ever outperform an Athlon based system is doubtful, but the fact that they are there and available is a good thing.
Lee
Muslim community leaders warn of backlash from tomorrow morning's terrorist attack.
Oh, come on! I had debian rocking & rolling on an old Centris with 8 MB ram.
The only downside was the 20 MB hard disk - by the time I got everything I wanted installed, there wasn't any room left to do anything.
Obviously you want a Sun laptop instead of a Mac, so why not just buy one?
Moron. Apple hasn't included an ADB keyboard with it's systems for almost five years now.
Nothing against OSX, but in all honesty even a fully loaded Linux desktop will take less RAM than OSX. I'm running a fully loaded KDE right now, with Mozilla (rumor says it's bloated), StarOffice 6.0, and Kmail, and my used (not cache) memory is only at 64MB. Note that only KDE and Kmail share many libraries.
I don't know where people get this idea that you need a ton of memory to run these programs. Granted, 128MB is little on the high side if you're talking real low end, but these are desktop workstations, as you said. OSX requires that 128 at a bare minimum, and even that maxed out when I was using 10.0 (don't know if they decreased the RAM requirements as of 10.1, but I doubt it). Linux apps just don't take as much memory as people say. KDE gets you a solid UI with a lot of bells and whistles for relatively low memory cost. Plus, the inclined desktop user could simply run fluxbox or twm, which isn't possible without Darwin, in which case you can't easily switch to Aqua. OSX is great, but Linux has its advantages too.
"I may not have morals, but I have standards."
How a 1st-generation PowerMac boots:
1. 680x0 emulator starts
2. run MacOS in ROM (partly 680x0 code)
3. read patch partitions to fix drivers
4. mount root filesystem
5. continue binary-patching the kernel
It's still pretty bad today. The firmware
reads HFS and HFS+ filesystems, expecting to
find a kernel with a strange Mac filesystem
attribute on it.
Enrique S. Reymundo quiero ANAL COX.
--
Miguel de Icaza
Strange, no mention of simply fixing the struct definition to instead use types from include/linux/asm/types.h (or equivalent) of arch-dependent "short" and "long", appropiate use of __BIG_ENDIAN et al in ordering the fields, and qualifying with __attribute((packed)).
Also, the macros in include/linux/byteorder/*.h handle data of both endianness (and provides the more limited ntohl() quartet).
"Merging into heavy traffic at near light speed!"
"Our inertial mass ever increasing!"
What is it about accesss to lower level operating system procedures that I should care about?
I can see where if you are running a server that a decent GUI is pretty irrelevant, but most people don't operate servers. There seems to be a feeling in the unix user community that if you can't cluster or if the server capacity is even slightly degraded that the computer isn't worth a damn.
If there is ever a time when there are more servers than there a people who actually access those servers for real world applications, the computer industry will die.
The continued growth of the computer industry depends on getting computers into the hands of the mass market. As it stands now even a Mac is too complicated to use for most people. The goal of computer scientists should be to make the computer easier to use, not more powerful as a server.
The Macintosh is designed for people who's job is something other than using a computer. They are a tool for people doing other things. A cake mixer is not a tool for furthing the manufacture of cake mixers, it is for someone who wants to make cakes and is designed accordingly. The same should apply to computers. The real use of computers, is to do non-computer related work. This is what the Mac is good at.
Linux people seem to be in the business of designing a better milling machine just so they can make a better milling machine. They don't seem to consider what is going to be made with that machine when it is made perfect.
Linux sounds great if you want to make a better computer for running a better version of Linux, but what exactly is Linux trying to do that matters to anyone who wants to do something else entirely?
"Are there some other advantages of Linux/PPC that I'm missing?"
Yes. Linux is Free software. Free as in, "Give me liberty or give me death." Free as in, this is ours and nobody can ever take it away. Mac OS X is not. Darwin is not.
I remember in '96 everybody telling me how slow Linux was on the Mac. Of course, that's because MkLinux was all that we knew about.
/etc files as well. Also, the Fink project has ported Apt and "dpkg" to Darwin, for your apt-get goodness needs. The best reason, which admittedly invalidates my last point, is that Darwin comes pre-installed. While this may not seem like that big of an issue for many slashdotters, I find that repartitioninging a Mac is not as easy as the standard Windows based PC:
Native Linux is almost certainly faster than Darwin, a BSD system hosted on the Mach Subsystem. It would be interesting, and possibly valuable to see a proper and thorough attempt at comparing performance of MkLinux vs Darwin. (It seems that MkLinux is only available with Kernel 2.2, though somebody hacked a 2.4 version.)
This might also reveal issues relating to Mach that are distinct from LinuxPPC or NetBSD, which would also make good data points. If time weren't an issue, OpenBSD, Debian and the other Linux Distros would be nice too.
A less technical but still valid issue is that of supporting Free Software. While Darwin is Open Source, it is not "Free", so it is less than ideal for development unless you are specifically targeting Mac OS X. While Linux and the GPL mean that you are forced to support the Free Foftware community if you distribute a derived product, the APSL effectively makes you an unpaid Apple developer, supporting the Apple community, whom in most cases already bought their support from Apple Computer Corp. If you ever tried to made your derived product commercial, you would have to take great pains not to attract the Ire of the big Apple, or your license might be pulled out from under you, and they would have full rights to your product to boot.
Don't get me wrong; I would love to run Mac OS X, but I am not yet ready for the prerequisite hardware investment, my 604 still works. Unless I am trying to hack OS 10 support for my "Mac" (clone), I would not use Darwin as a platform for development without remuneration, preferably from Apple. If I were _really_ determined and talented, I might even attempt to add Darwin interfaces to a Free Software system like NetBSD. (Not that I could... but for the sake of curiosity, does anybody know if Mac OS X makes direct use of the Mach subsystem?)
Of course, the moral issues are more substantial than any technical reasons, right? Theoretically, all technical issues are solved over time. But at the moment, Linux is much better for interoperability, especially in the department of mounting non-native filesystems. Darwin, and AFAIK all of the Free Software BSDs have a paucity of available filesystem drivers.(FFS[nee UFS], Ext2[non-anynchrounously], cdromfs[iso9660], msdos[FAT], LFS[Log structured], and FFS+softupdates) Darwin might only support UFS(really FFS?), HFS+ and cdromfs.
Just to play the devils advocate, Darwin isn't without merit. Debian is dead easy to admin once you know the Debian tools, but before that it might as well be Slackware. Darwin has NetInfo, which is supposed to be very easy to use for administration, and a very powerful centralized tool which may be less of a learning curve than Debian. It is supposed to respect standard
You must be aware of the various system related partitions; pdisk is not as pleasant as cfdisk or even fdisk; there is no lossless repartitioning utility like Partition Magic or even FIPS - you must use a defragmenter to clear a contiguous area from the end of the partiton, alter the partiton table, then repair the "resized" filesystem. The best answer is usually just to buy another hard drive, or just the extant Darwin.
Oh yeah, MOL (Mac On Linux) is worth considering if your Mac (clone) isn't up to running Mac OS X. Does anybody know what ever happened to Sheepshaver? It runs Mac OS under BeOS, and was supposedly ported to LinuxPPC.
-castlan
Just to defend against your attacks on the character of Mac OS/System x... I do not see how you can question its validity as "a system at all." The irony of your stance is especially poignant when you consider the famous debate between Linus and Tannenbaum, who similarly argued that Linux is a bad system design." The point here is that both Linux and Mac System x are Monolithic, as opposed to the Microkernels that Tannenbaum and the academic community supported.
.o binaries, then Linux would be an unworkable sytem - recompilation is the only saving grace of Linux... the GPL is essential to the survival of Linux in a dynamic hardware platform.
Bear with me on this next point for a moment, if you will. Classic Mac OS, like Linux, tend to be stable systems in general, but due to their monolithic nature, they are can only be as stable as their weakest link. Early Macintoshes were very stable, but they were closed systems. They were very good PCs for their time, but what you call bad system design might better be considered eggregious hacks to support an archaic system. In the long run they hurt the Mac OS, but the useful life of Macintoshes were far greater than Intel peers. Nevermind Windows, I know of Macs that were useful (and used) almost as long as Linux has existed.
The problem is that being a well defined (Mac ROM) and closed system, the Macintosh couldn't really adapt when new ideas in computing were becoming the norm for general consumers. You know, newfangled things like CD-ROMS, 802.2 Ethernet and TCP/IP (for a Personal Computer? No Way!). The crude hack was "extensions", after the fact drivers that played monkey in the middle with the system folders and extended the Mac in ways that it was never designed for. Considering the constraints they placed on themselves, Apple did a fine job, and Macintoshes outlasted generations of Intel PCs, up utill about the time Windows 95 brought half decent attempts at memory protection and a WIMP GUI, finally making 68K Macs less than viable as a primary computer.
As for Apple having almost no experience with OS X, I have to strongly disagree. Apple was founded by Steve Jobs. Steve Jobs founded NeXT, and NextStep was a stable Unix longer than Linux existed. Steve Jobs is back at Apple, and he brought his Unix with him. Just about all of Mac OS X technologies are stable and well understood systems, from the Mach Microkernel foundation, the BSD based single server and Unix Security model, the Free/NetBSD userland, the Carbon MacOS toolkit, the Mach-o binary, ObjectiveC source, up to the DisplayPDF GUI descended from the NeXT DisplayPostScript system.
Apple hardly had "weak support of their own hardware platform". Rather, their support was so solid, that such concepts of "plug and play" weren't even an issue, that was and still is an Ideal that Microsoft aspired to reach, as all hardware supported by Apple was seamless if you could manage to keep SCSI IDs unique - can you count to 6? Unfortunately, 3rd party hardware, and lazy non-Apple software started to require Extensions to be functional, and that is the straw that broke the camel's back. This is the strongest concrete argument against binary only drivers for Linux... in a mololithic system, drivers step on each others' toes, and instability results. If all Linux drivers were only
As for MS's quality... is it better or isn't it? If my graphics card is poorly "supported", then the fact that "they support a much broader base of hardware devices" is meaningless. Either the hardware you need is supported, or it isn't. PCI is an open standard, and not x86 specific in any way. NuBus devices tended to be more expensive than ISA, but were far better supported (i.e. What is an IRQ? Eeew, why?) The Linux "zoo" environment consisted of Linus' PC. Early users added support for their PCs. This eventually added up to a very populated zoo, but by sheer number of interested users (those frustrated by AST's lack of public leadership for a Free Minix.). This eventually built up a large base of hardware support, but not in the way that you seemed to imply with your zoo metaphor - for the longest time, weak hardware support was considered a very significant Linux flaw. Commercial companies tend not to have individual systems that they can use as a target to port to... they rather port to a class of systems, which seems to be what you refer to with your "zoo" analogy. In this aspect, BSD is much stronger in "zoo" support, as embodied by NetBSD. NetBSD is designed to be very portable, and that is maintained... once hardware is supported for any of the large stable of NetBSD platforms, is it usually supported for all NetBSD platforms. Even in desktop-oriented hardware - NetBSD was the first Free OS supporting USB for eample.
That is not to say that Microsoft is necessarily bad with system design. Windows NT was a very good design, a MicroKernel that IIRC was written on MIPS hardware and then ported to IA-32, just to prove its portability. When NT supports hardware, I have found that the support is usually quite robust. Unfortunately, if graphics hardware support is at all flaky, then the system suffers. As NT is a thorougly graphical platform, MS chose to integrate the Graphics subsystem into the kernel level for performance reasons. This defeats most of the advantages to a microkernel based design. They have also discontinues support for non Intel platforms, which was another Win for Win NT. While these decisions don't significantly impact marketshare, they have severely hurt the integrity of the system. Similarly, Mac OS X and Darwin are based on the Mach MicroKernel, but only implemented as a single server, which is effectively just a monolithic kernel with message passing overhead. And while Mac OS X is limited to PPC hardware, there is at least the potential for Darwin to run on other Mach supported Platforms, especially IA-32.
As for Apple and RISC... the PPC is arguably the weaskest implementation of RISC. More significantly, the point is moot - The RISC/CISC argument is archaic in the face of superscalar architectures. Curent Intel processors don't use the X86 instructions internally, they run microcode similar to RISC style instructions. And the G4 with Altivec is very un-RISC-like. Apple has positioned itself as a hardware company, but their software was always what made their systems valuable. Just because the Mac OS was designed as a single-threaded single-user OS doesn't mean it was a poorly designed system. Remember that it was a full windowing system when Bill Gates was advocating the 640KiB ceiling. Unix running X11 was not feasible without a massive cash investment, and if you had massive cash, you could have bought an Apple Lisa. The Apple Lisa was designed as a multitasking OS from the start, but it was too pricy to have any commercial succes, so Apple pared back the design to create the original Macintosh. In the 80's the limit was not software design but production costs.
Wow, it really looks like I an against every single paragraph you wrote, but at this point I might as well go for broke! Apple realised that their hack upon hack of cutting edge early 80's PC design was unmaintainable, so they tried to maintain backwards compatibility while integrating technology that had become more readily available. You say that they were afraid of Linux as a competetor, but it sems that you are forgetting about their previous Unix on Mach, MkLinux. They didn't fear Linux as a competetor, they actually allocated significant resources into Linux on Mac development for a few years, before BSD was even considered. The issue had nothing to do with any of your imagined "desktop dominance" by Linux, but with the GNU GPL. BSD wasn't chosen so much as NeXTStep. BSD happened to be a foundation for that platform, but more importantly, NeXT BSD was hosted on Mach, so that the MkLinux development effort wasn't a complete loss, they recovered much of the Mach related work. After purchasing NeXT wholesale, they leveraged the Linux inspired hype of "Open Source" without the cost of truly embracing Free Software with their APSL. Marketing Genious building upon many years of Next's technical wizardry.
Your last paragraph missed the point of Linux. It is just an implementaion of POSIX standards. Especially with Open Source and Free Software, anything that runs on Linux runs on Mac OS X by virtue of the POSIX compliance in BSD as well. The only exception is the few binary-only applications which run on Linux while running against the Free Software spirit of Linux. In contrast, in addition to Free GNU software, Mac OS X draws on the large base of existing NeXTStep/OpenStep software and Mac OS software. If Darwin runs on your Mac, then RAM seems to be the only limit or OS X running on your Mac as well. If Mac OS X won't run on your Mac, then LinuxPPC or MkLinux are fine choices. But despite my preference for Free Software, I have to admit that if you can run Mac OS X, then it currently beats Linux hands-down.