Re: Linux isn't an operating system

Andreas Kostyrka (andreas@medman.ag.or.at)
Sat, 9 Mar 1996 11:59:28 +0100 (MET)


On Thu, 7 Mar 1996, Richard Stallman wrote:

> You've proposed that we agree on the term "Linux system" for the
> complete systems based on Linux, and give credit to GNU developers "in
> the second breath". That is a reasonable way of giving credit, but it
> has two drawbacks:
>
> * It is what most people have been doing, and that is what has led to
> our present problems.
What problems? I mean, the last thing I've heard FSF was, that Linux is
- not really interesting and important, because there is HURD,
- just one of additional third party systems like SunOS, Solaris, HP-UX, ...

So what let you change your mind? (The success of Linux? Which is partly
a success of the GNU tools, but it is much more a success of the development
style of Linus, HJ Lu, ...)

You speak about cooperation, and still you have complained loudly about
usage of GCC snapshots in libc development. The problem is not a problem
you will solve by renaming LINUX. You could help it tremendously by
admitting that Linux is not just one of systems that can compile GNU
software, but one of core systems.

Another thing would be a change in development style: Let the GNU
developer take part in the discussions on the linux lists. If they have
good points, they will be taken into account, if not, then they should
not try to make HJ Lu take the wrong decision. Remember: FSF stands for
FREE software foundation. And the GPL has somewhere the sentence that
starts like ``By contrast, the GNU GPL is intended to garantee your FREEDOM ...''.

It's a hell job for HJ Lu already now. And he does a mission critical job:
Linux libc is mission critical to Linux. I wouldn't ask him to use his
precious time on cooperation with the glibc folks, especially as the GNU
folks have proven often enough not to have very compatible mind set.
(The ``mission'' for HJ Lu is a working SYSTEM libc, and just some
REPLACEMENT for another system libc.)

(Kind of Billy's tactics: Cairo (oops, not on time), ..., Windows 95
(lucky: We have removed enough features, so it's ready.) -> HURD (oops,
no working system around), HURD (oops, no press releases) -> Linux (it
works, starts to have the press coverage, ...) )

> * The real issue is not giving credit, but encouraging the cooperation
> that we all need.
Correct. The problem here is, that Linux success comes also from it's open
developement cycle. So changing it is rather not practible. And it's much
more flexible, because it takes a long time for a paper document to travel
by snail mail -> this makes the linux development cycle quite impossible:
1.) there is fatal bug in the currect release of libc.
2.) someone submits a patch that cures this.
3.) We wait a week till the paper signing away the rights arrives, and
then include the patch into the next release.
(in Linux cycles a week can be two or three releases!)
summa summarum: In this one week 967 harddisks in systems around the
world have been trashed. (967 is just a random number.)
>
> I'm sure that if I let this issue drop, and let the issues of
> cooperation drift, we could all "get along" with no visible disputes.
> But this is because the substantive problems would quietly continue to
> grow. We need to do more than get along; we need to work together.
Correct. Joining the Linux developement team is quite easy: contribute.
But don't expect that this gives you dictator-like rights.
>
> People using almost the same collection of software have defined two
> different community identities--"GNU user" and "Linux user". And the
There is a big important difference: Their goals. "Linux user"s don't
think in terms of HURD. "GNU user"s (at least the developers, I assume
this is what you mean by user *grin* :) ) do.
> people in one community tend not to make an effort to cooperate with
> those in the other. This suggests that the real root of the problem
There is cooperation. But basically you have to admit that HJ Lu
tried to cooperate. But when the GNU maintainers didn't take his patches,
which he deemed vital for Linux developement, he had only one option:
Make his own distribution. This is his right, I assume you know the GPL
by heart.
> is in this social structure, not in the individuals concerned. To
> have more cooperation, we need to change the structure. We need to
> pay more conscious attention to our ties and common interests, and to
> the larger community that includes all of us.
A side question: Which is the last ``released'' glibc version?
1.09 doesn't support linux (and it's not that old!), and the
development ones doesn't count. (As I remember it from the GCC snapshot
problems, the general linux hacker is not considered member of the
GNU developement team, so he may not use developement releases *grin*)

> To effect a change, we have to do something different from what we
> have done in the past.
Yes. The FSF has to take Linux seriously, and it has to play by the
Linux rules.
>
> I've proposed the term "Linux-based GNU system" as a way of doing
> this, because I think it is an accurate description. Other people
Hmmm, let my think:
- GCC
- fileutils,
- textutils,
- binutils,
- some roots of linux libc
(- Emacs) -> I know some developers that don't use Emacs.

This a four components, which partly could be replaced by PD stuff.
The cause why this is not done: Because they are superior in quality
to their potential PD counterparts.

If you want the Linux community to use glibc, than make it better in
quality than Linux libc. But don't expect the Linux community to
``surrender'' their work to you, just because you re the Great RMS from FSF.
(Remember: The ratio between the general population that knows RMS and
that knows Linus Torvalds is quite in favour to Mr. Torvalds!)
> have suggested the term "GNU-based Linux system". That too would
> promote cooperation, if people use it often. Both terms emphasize the
> relationship rather than implying separation and isolation.
Playing political games with labels, and not changing anything in the GNU
developement style won't help you. (The problem with the ``Lost Son''
Linux is, that it didn't want to go away, but GNU made it, and it lives
quite well now without sharing a house with GNU now, if you allow me this
biblic analogy.)
Can you explain why patches that HJ Lu considers vital are rejected?
The GNU maintainer should perhaps adopt Linus's working model:
patches that come from known developers are only checked very briefly, and
go in very quickly.

Enough of politics,

Andreas

--
Andreas Kostyrka
Email: andreas@medman.ag.or.at
Fax: +43/1/7070750 Tel: +43/1/7077571, +43/664/3020166 (cellular)
Copyright 1996 Andreas Kostyrka.  Microsoft Network is prohibited from
redistributing this work in any form, in whole or in part.