no gnu [LWN.net]

no gnu

Posted Aug 21, 2009 3:28 UTC (Fri) by coriordan (guest, #7544)
In reply to: no gnu by sbergman27
Parent article: 40 years of Unix (BBC)

It just shows the journalist listened to PRs rather than actually researching the article. "Linux" or "open source" isn't the reason our favourite Unix-like OS exists today.

to post comments

no gnu

Posted Aug 21, 2009 3:59 UTC (Fri) by jordanb (guest, #45668) [Link] (8 responses)

It appears that the only source for the entire article was Dr. Peter Salus. Considering that he served as a Vice President of the FSF for a while[1], I'd imagine he'd have mentioned it if he thought it was apropos.

On the other hand, the article is atrocious from a purely factual analysis. In addition to the statement about Multics, it conflates the effect of Unix with that of Minicomputers (Unix rode the Minicomputer wave, not the other way around). Also, AT&T did *not* give Unix away "for free" they sold it for low cost to Universities and the government. While I'm at it, the caption under the stock photo of the fiber terminals is wrong: Unix didn't have networking "from the start." It was tacked on later.

According to his Wikipedia page, Dr. Salus is a linguist and not a historian. Maybe he doesn't he doesn't see history as something worthy of factual rigor. Or maybe the BBC writer is a pathetic hack in his profession who can't even get a direct quote right. Not knowing anything about Dr. Salus personally, I figure both are equally likely.

[1] http://en.wikipedia.org/wiki/Peter_H._Salus

no gnu

Posted Aug 21, 2009 4:59 UTC (Fri) by ncm (guest, #165) [Link] (4 responses)

networking ... was tacked on later.

And rather badly at that. I'd like to know what BBN's reference implementation system calls looked like.

no gnu

Posted Aug 21, 2009 7:29 UTC (Fri) by nix (subscriber, #2304) [Link] (3 responses)

If you can read past the ALL CAPS, RFC681 describes the rather peculiar
open()/close()-in-userpace /dev/net/$hostname scheme they used back then.

A shame it's not still in use.

links < /dev/net/com/google/www

no gnu

Posted Aug 21, 2009 9:29 UTC (Fri) by hppnq (guest, #14462) [Link] (2 responses)

echo "GET /search?q=/dev/tcp+ksh HTTP/1.1" > /dev/tcp/www.google.com/80

:-D

no gnu

Posted Aug 21, 2009 10:07 UTC (Fri) by nix (subscriber, #2304) [Link] (1 responses)

GNU awk can do this too. But it's not really in the filesystem, so it's cheating.

Of course, with FUSE and asynchronous name lookup, we can do this properly, without relying on a horrible userspace-NFS-server hack. (ls(1) would necessarily only be able to show you names you'd already looked up, though.)

Plan 9?

Posted Aug 27, 2009 8:00 UTC (Thu) by job (guest, #670) [Link]

Doesn't Plan 9 have a similar TCP-by-filesystem thing going on?

it's wrong to blame Salus for what a journalist wrote

Posted Aug 21, 2009 5:14 UTC (Fri) by JoeBuck (subscriber, #2330) [Link]

It's common for journalists to mangle the words of technically inclined interviewees beyond recognition, even when they put quote marks around those words.

no gnu

Posted Aug 21, 2009 7:25 UTC (Fri) by nix (subscriber, #2304) [Link] (1 responses)

Oh, blame the BBC. Its science/tech coverage has a track record of making
appalling messes of *everyone's* words. I suspect they spoke to Peter
Salus, but expect an error rate of a word or two per sentence, often
important words.

Where the BBC has a world-class reputation is in politics and current
affairs coverage, not in tech.

no gnu

Posted Aug 22, 2009 9:18 UTC (Sat) by lysse (guest, #3190) [Link]

And even then, don't you find yourself wondering about the stuff they get wrong in the fields you *don't* know about...? :)

no gnu

Posted Aug 21, 2009 5:22 UTC (Fri) by sbergman27 (guest, #10767) [Link]

Right. Because if Linux had been just a few months later in maturing, GNU/OS, riding atop Hurd, would have taken taken off like a rocket. It was admittedly a very close thing.

no gnu

Posted Aug 21, 2009 5:49 UTC (Fri) by rsidd (subscriber, #2582) [Link] (24 responses)

Um, yes, Linux (the kernel) is the main reason we have a free OS in widespread use today. If GNU hadn't existed, Linus would probably have borrowed from BSD and elsewhere (a huge part of a running Linux system has BSD roots even today). If Linux hadn't existed, the FSF would still not have a usable operating system, and we'd be running either some form of BSD or a proprietary system.

It is ridiculous for GNU to claim credit the way they consistently do. The GNU project contributed a compiler and toolchain, which was truly indispensible but does not constitute an OS; a C library which has not been under GNU control for much of its history; and a few utilities. Other than the toolchain, everything already existed or was soon reimplemented in BSD. The X window system, TeX, and much else was not invented by GNU. The continuing credit-grab is obnoxious. A working linux system today is the product of many minds, but its development since 1991 has been driven by the existence of a free kernel that made a usable OS possible. It is right that the system be named, in popular usage, after that kernel.

no gnu

Posted Aug 21, 2009 6:55 UTC (Fri) by drag (guest, #31333) [Link] (8 responses)

> Um, yes, Linux (the kernel) is the main reason we have a free OS in widespread use today. If GNU hadn't existed, Linus would probably have borrowed from BSD and elsewhere (a huge part of a running Linux system has BSD roots even today). If Linux hadn't existed, the FSF would still not have a usable operating system, and we'd be running either some form of BSD or a proprietary system.

Hrm.

I can run GNU/kFreeBSD or GNU/Solaris and have a much more Linux-like operating system then then if you took Linux and tried to run something else on top of it.

Face it. It's much much easier to take the Linux out of GNU/Linux and have a fully viable and familar operating system then it is to take to take the 'GNU' out of it and use Linux with something else.

Linux kernel is certainly very important and it is the poster child for opern source-done-right, but GNU portions are far more critical to the operation system I use everyday. I can replace Linux with at least 4 other kernels and still be able to do most everything I do on "linux", but I couldn't do the same with GNU software.

I think that it would been much more likely that the Linux kernel itself would of been rejected in favor of the already established BSD kernels and other things.

After all was Linux that started off with the 'You must not charge any money' license, not GNU. In fact I think that having Linux kernel under the GPL license is one of the reasons for it's success, and that is certainly due to RMS and the GNU folks (since they created the license in the first place)

I don't give a shit if people want to call it GNU/Linux or whatever. The way it looks to me now, considuring the portability of GNU and the ability to run the same environment I use every day on multiple different kernels that it would actually be more accurate to say 'I use GNU OS' rather then to say 'I use Linux OS'.

Either statement is incredibly inaccurate, of course.

no gnu

Posted Aug 21, 2009 8:38 UTC (Fri) by hppnq (guest, #14462) [Link]

It is far from a perfect metric, but checking the copyright notices for "Free Software Foundation" in binaries in /bin and /usr/bin using strings(1) turns up a total of less than 15% GNU programs on an Ubuntu 9.04 system.

Still, I agree with the sentiment that GNU is what made Linux popular -- and then Linux made GNU popular to the point that we now have a, well, herd of kernels to plug into GNU. ;-)

no gnu

Posted Aug 21, 2009 11:42 UTC (Fri) by stevenb (guest, #11536) [Link] (4 responses)

> It's much much easier to take the Linux out of GNU/Linux and have
> a fully viable and familar operating system then it is to take (...)
> the 'GNU' out of it and use Linux with something else.

Hmm, taking the 'GNU' out, isn't that what the Android folks did, more or less?

no gnu

Posted Aug 21, 2009 20:45 UTC (Fri) by eparis123 (guest, #59739) [Link]

They still use gcc ;), but your point is still true.

no gnu

Posted Aug 22, 2009 4:46 UTC (Sat) by drag (guest, #31333) [Link] (2 responses)

Yes. And it's a completely different personality.

I looked at the tools that I use and the environment that I deal with for various projects on a day to day basis and realized that the GNU stuff is either heavily used by myself I've realized that the a lot of the stuff I use couldn't be used without GNU.

Gnome, for example. GNU project.

Bash I use a lot. There are other shells, of course. But that would require me changing what I use. I use the GNU shell utilities a lot, also. Heavily. And I use and depend on the GNU-isms in there.. I like the colors and '-print0' and all that stuff. It makes things much easier.

So while you can go around and say 'oh there are other *nix environments', sure. But they are different.

And other things that I use.. like Vim or Python, are not GNU, but they are not the same as other platforms. They use different little GNU things all over the place for extra features and while they are not core to the system... they do make it nicer.

Look at the stuff your using right now. Given the choice between running GNU on Solaris or GNU on kFreeBSD vs any non-GNU Linux environment... (Android or Busybox or whatever..) which would be the less usefull? Which would require more effort to transition to and require the most sacrifices?

Now I use Linux kernel because I think that it's the best avialable, by a long shot, out of opensource/free software operating systems.

And I use open source software because of it's _Freedom_. If I cared about practicality over ethics or freedom then I would be using Windows or OS X. (and NO this attitude is not restricted to computers. I had my opinions and prejudices long before I ever came into contact with Linux or GNU. They just happen to match me.) Either of those OSes are much less effort to use and can be customized heavily if you know what your doing. Depsite being closed.. for a end user they are very flexible. I would probably still use Linux on the server, but whatever.

no gnu

Posted Aug 28, 2009 1:18 UTC (Fri) by JanC_ (subscriber, #34940) [Link] (1 responses)

I might be wrong, but AFAIK GNOME isn't a GNU project...?

no gnu

Posted Aug 28, 2009 19:14 UTC (Fri) by Darkmere (subscriber, #53695) [Link]

GNOME stands for the "GNU Network Object Model Environment", So I would say you are wrong.

no gnu

Posted Aug 21, 2009 18:40 UTC (Fri) by MarkWilliamson (guest, #30166) [Link] (1 responses)

I don't really disagree with your point. I do find the GNU/Linux naming thing somewhat irritating - but given GNU and Linux basically supply the most fundamental stuff most Linux distros are based on, I can understand the point the FSF are making.

I just thought I'd note that if we're judging it in terms of the effect on user experience, whilst for many users the GNU software has more influence on the "Linux-like" experience it is also the case that this software has very little influence on visible behaviour for many others.

For instance, many users of KDE-based Linux distros would probably not notice if you replaced all the GNU parts with BSD stuff - if you swapped some office worker's Mandriva install for PC-BSD, would they notice?

Again, I'm not really disagreeing with you - I think GNU software is very important to the functionality and behaviour of most Linux installs. But I think the low-level co-ordinating roles the GNU software takes (compiling, linking, scripting, libraries) is the main reason for its importance, not its user-visible characteristics.

no gnu

Posted Aug 27, 2009 10:59 UTC (Thu) by kafka (guest, #59620) [Link]

> For instance, many users of KDE-based Linux distros would probably not
> notice if you replaced all the GNU parts with BSD stuff - if you swapped
> some office worker's Mandriva install for PC-BSD, would they notice?

The BSD systems also rely on GNU stuff, so I guess that, yes, they would notice.

no gnu

Posted Aug 21, 2009 11:08 UTC (Fri) by coriordan (guest, #7544) [Link] (5 responses)

GNU Hurd was begun after a debate about whether to write Hurd or whether to use the BSD kernel. Had people not slotted Linux into the GNU system, the GNU project or someone external would have slotted the BSD kernel into the GNU system.

To write an OS, what the first step? You need a compiler. GNU wrote one. What's the next step? You need a libc. GNU wrote one. What's next? You need a shell, command line utilities, a debugger, etc. etc. GNU wrote all those things and even did the licence writing work. Then someone was spotted putting the finishing touch on it and got all the credit.

no gnu

Posted Aug 21, 2009 12:49 UTC (Fri) by Cyberax (✭ supporter ✭, #52523) [Link] (1 responses)

Yet right now GNU is not that essential. Most userspace utilities are not that important in GUI environment for most users, there are several non-GNU libcs, etc.

The only 'hard' dependency is GCC, and it's slowly getting replaced by LLVM+Clang.

no gnu

Posted Aug 23, 2009 1:33 UTC (Sun) by jlokier (guest, #52227) [Link]

The GPL is another hard dependency. I don't see that going away soon.

no gnu

Posted Aug 21, 2009 18:02 UTC (Fri) by rsidd (subscriber, #2582) [Link] (2 responses)

I grant GNU the compiler and the C-library in the early days. BSD made as much contribution to "utilities", or more, than GNU. By 1991, BSD had a nearly complete OS, minus the toolchain. That was also the year Linux was released. Why Linux "won" is still a matter of debate. But for the 7 years until 1991, GNU software development was motivated by FSF ideology. For most of the 18 years since 1991, GNU and other free software development has been motivated by the existence of Linux. Even GCC ended up being forked (into egcs) and then re-merged, while the maintainer of the GNU C library, Ulrich Drepper, is openly hostile to the GNU name-grab: "I find this [the GNU/Linux term] completely unacceptable and can assure everybody that I consider none of the code I contributed to glibc (which is quite a lot) to be as part of the GNU project and so a major part of what Stallman claims credit for is simply going away." And more and more embedded people are choosing other libc's (often based on BSD) anyway. The only significant continuing contribution of the FSF, for many years, has been emacs.

no gnu

Posted Aug 22, 2009 8:22 UTC (Sat) by dd9jn (✭ supporter ✭, #4459) [Link]

FWIW: A quick check of the GNU copyright list shows a wealth of "Assigns past and future changes" by Uli since 1994 - including GLIBC. Thus a statement "I consider none of the code I contributed to glibc ([...]) to be as part of the GNU project [...]" is not quite right.

Why did Linux "win"?

Posted Aug 26, 2009 3:11 UTC (Wed) by vonbrand (guest, #4458) [Link]

Because in the critical timeframe of the beginning 90ies the USL vs BSDI lawsuit threatened to make BSD go away, so hackers looked elsewhere. That the BSDs were less than friendly to outsiders contibuting stuff (and the FSF even less so) sure didn't help making them popular choices.

no gnu

Posted Aug 21, 2009 11:26 UTC (Fri) by dbruce (guest, #57948) [Link] (1 responses)

All these "If X hadn't existed, then..." are just uncontrolled thought
experiments - we have no way of knowing what would have happened.

But if GNU existed and Linux never got started, I think it is pretty safe to
assume that we would have some form of GNU system today - either with a BSD
kernel, or with a GNU kernel that might have gotten a lot more hacker
interest without Linux on the scene.

Reminds me of "without Microsoft Windows, we wouldn't have personal
computers".

no gnu

Posted Aug 22, 2009 23:03 UTC (Sat) by sethml (guest, #8471) [Link]

We had GNU tools without the Linux kernel for quite a few years. In '89 and '90, before
Linux was useful, I ran a cluster of HPUX machines. We had all the GNU tools compiled
and on the PATH before the HP tools because the GNU tools were much more fully-
featured and less buggy, plus we didn't have the funds to pay for HP's good compiler. Still, I
wouldn't have called them GNU/HPUX.

Without GNU I'm sure a perfectly reasonable compiler and userland would have been written
and found for Linux-based systems, but much of it would not have been as fully-featured or
as consistent.

no gnu

Posted Aug 21, 2009 17:25 UTC (Fri) by juriise (guest, #38305) [Link] (3 responses)

There are different perspectives. The Linux perspective: We have the kernel, lets take whatever is available to complete the system.

The GNU perspective is NOT the opposite. The GNU perspective is not "We have some compiler, where can we find the kernel to complete the system". If that were the case, you would be completely right in deciding that the GNU tools is the little part, the kernel is the bigger part.

On the contrary, the GNU perspective is: We want to build a completely free system out of parts from anywhere free software is available. The requirement for freedom was there from the beginning, in fact it was the whole point of starting. RMS began with what was natural for him, the compiler. It was never the point to do everything himself, rather to talk people into writing software that was free, so it could be a part of the system. Probably the most important part was defining free software through the GPL and otherwise.

Linux was free, and it was chosen. Probably because it had a GPL license, and the Hurd could not be completed. The objective of GNU is nearly fulfilled. Therefore it can be called the GNU system.

Seen from GNU, the complete free system is important. Seen from Linux, the kernel is important.

no gnu

Posted Aug 21, 2009 18:40 UTC (Fri) by bronson (subscriber, #4806) [Link] (2 responses)

Not sure what you're arguing for or against. I've never met a Linux dev that thinks "only the kernel is important." Have you?

Also, love the way you spin the exact same idea two different ways.

"the GNU perspective is: We want to build a completely free system out of parts from anywhere free software is available"

[the Linux perspective is:] "Lets take whatever is available to complete the system."

Yes, both projects adopted large amounts of existing software to help them complete the big picture. Why use pure rhetoric to make one look evil and the other look benevolent?

No accident

Posted Aug 21, 2009 19:04 UTC (Fri) by alex (subscriber, #1355) [Link]

I think the point was (as I've heard RMS say many times) is that when Linus looked around for a compiler and user-space to complete his OS it wasn't by accident he found GNU. The FSF's route map towards system freedom started by re-implementing free versions of libraries on proprietary systems as they realised starting from the kernel would be a fairly tricky task.

Who knows what would have happened if Linus hadn't written his terminal emulator/kernel. Most probably something like BSD/Debian, but we'll never really know.

no gnu

Posted Aug 21, 2009 21:49 UTC (Fri) by juriise (guest, #38305) [Link]

"Not sure what you're arguing for or against."

I am arguing for the importance of GNU, and that it is OK to call it the GNU system, when I want to point to the freedom aspect of the system.

"I've never met a Linux dev that thinks "only the kernel is important." Have you?"

I have heard that Linux is the most important part, that calling it the GNU/Linux system is taking credit wrongly, that free software is the old name for open source, they provided the compiler and the toolchain (thats it)...

I think that the biggest contribution from the GNU side is the definition of free software, the licenses, the persistent marketing. They also made some software.

no gnu

Posted Aug 21, 2009 21:09 UTC (Fri) by AJWM (guest, #15888) [Link]

Compiler aside, the core userland tools were re-implemented over and over in many different languages. Kernighan & Plauger's Software Tools (1976) described many of them with source code in Ratfor, a C-like preprocessor for FORTRAN, and the later Software Tools in Pascal (1981) did the same in Pascal. Pascal compilers were almost ubiquitous at the time, and pretty easy to implement (I had a version of Wirth's that was about 3000 lines of Pascal).

I had a reasonably complete (for day-to-day use) 'nix-like userland on a 60-bit CDC Cyber 70 running NOS back in the day (circa 1980), based on Software Tools, several years before FSF and GNU. The real contribution of the latter was the GPL.

yes gnu

Posted Aug 23, 2009 1:33 UTC (Sun) by jlokier (guest, #52227) [Link]

The GNU project contributed two other things you forget to mention:

The Idea, and the Legal Instrument which helped bring it about.

Specifically the idea of a full-featured unix-like operating system that was not encumbered by commercial restrictions, as the Unix source code was at the time.

Actually more than that: the inspiring idea of an operating system that makes the world a better place; the idea of sharing code, encouraging sharing, and resisting attempts to prevent sharing.

The legal instrument was the GPL of course.

You can't deny that idea is quite pervasive throughout our favourite OS, and so is the legal instrument.

If you just count lines of code owned by the FSF, you have missed their most important contributions.

Nowadays you can say Net/Open/FreeBSD have the same idea, which is true, and they have succeeded in their way.

But at the time, BSD wasn't an option; it was still encumbered by commercial constraints. The GNU project ended up taking some code from BSD because it became possible later: the legal situation changed. But GNU would have written everything from scratch if they'd had to.

GPL

Posted Aug 27, 2009 8:03 UTC (Thu) by job (guest, #670) [Link]

To be fair, the GPL itself has played an important part of Linux' success.