Re: question about historical GNU System releases

[olafBuddenhagen]

Hi,

On Wed, Sep 12, 2007 at 03:45:43AM +0200, Arguri wrote:

> Well, I don't know, but I assume that ams stopped working and Marcus
> appeared and did the things, which turned out to become Debian/Hurd.

AFAIK Alfred wasn't around yet at that time. He certainly wasn't
responsible for the 0.2 release.

> Actually there would be (at least thats what ams says) not that much
> difference from a GNU 0.2 release (1997) or a GNU 0.3 release (2007).
> It would include nearly the same amount of packages and the
> installation would be the same. Although many packages would be up to
> date there would be no real difference in system behaviour.

No, that's wrong. What Alfred is claiming is that there would be not
much difference between the latest *snapshot* of GNU -- which he
released at the beginning of 2006 IIRC -- and a new snapshot made now.
(IMHO that's not true BTW; there were some pretty relevant improvements.
But well...)

0.2 is so obsolete it's not even funny. Nobody would claim there would
be no difference compared to *that*.

> Another Point was mentioned by Olaf: GNU Mach 2 (which would have
> included oskit) was abandoned because it caused enough work to let the
> developers say, that GNU Mach 1.x could be fixed and could use newer
> and even better drivers than GNU Mach 2 and still would not require as
> many work as GNU Mach 2 would need. It turned out that the work one
> needs to spend on GNU Mach 1.x is in my eyes nearly the same as for
> GNU Mach 2 (or even less, because oskit is now unmaintained and as
> such the GNU Project would need more or less its own fork).

Seems I was unclear in my explanation. For one, I was talking about the
original situation, when oskit was still maintained upstream. The
situation of course totally changed the moment it became unmaintained;
there was just not much point in oskit-mach anymore after that happened.

What I was trying to say is that originally -- while oskit was still
maintained -- the *total* amount of work it would take for oskit-mach to
fix the few outstanding issues would be smaller than the sum of all the
minor fixes necessary for gnumach1 to reach the same quality. However,
gnumach1 was already working in most situations, and the fixes necessary
to make it work in others were mostly quite simple. So people preferred
adding numerous small fixes for gnumach1 over the larger one-time effort
to finish oskit-mach, although the total amount of work for the latter
would be less, as once finished, it would have avoided most of the
numerous quirks with gnumach1.

What I wanted to show is that if there is a medicore and problematic
solution, but people are using it already and it works most of the time,
people will prefer contributing to this known medicore solution, than to
switch to something better, even if they know it would safe a lot of
effort in the end.

Another good example for this is grub: grub legacy is a horrible
unmaintainable mess; but it is widely used and already works in most
cases, so people dedicate much more effort to shoving even more messy
patches into it, than to finishing grub2, which once finished would make
adding new features and fixing problems so much easier...

> GNU Mach 2 for instance has serious problems (as 1.x has), but the
> problems are even more than for 1.x because 1.x does run on real
> Hardware and 2 won't even compile with recent Distribution. 

Well, no wonder considering that it is unmaintained for several years
now. Nobody seriously consideres switching to oskit-mach anymore.

> Such considerations are surely one, but not the most important, cause
> why there have not been any "offical" Releases for 10 years.

No, not really. The reason is more that the people in charge didn't
care.

Also, some now believe that not having released for such a long time,
expectations have risen, and thus no further releases should be made
unless it's close to perfect... Silly argument IMHO; but well, I'm not
the one to decide.

-antrik-