On Sun, Jun 29, 2008 at 06:06:14PM +0900, Yukihiro Matsumoto wrote:
> From 1.8.6, we have reformed our release management.  Ruby 1.8 release
> manager is Akinori MUSHA <knu at iDaemons.org>.  He is responsible for
> the 1.8 head release (i.e. 1.8.7 right now).  Besides that, we
> maintain two prior versions for production maintenance (1.8.5 and
> 1.8.6 now).  Shouhei URABE <shyouhei at ruby-lang.org> is responsible
> for those versions.  We don't think we need to fix this process.

Well ... My experience with the Ruby development/release process since 1.8.5 has
been the following:
* 1.8.6p0 thread support was completely broken. The answer I had at this time
  was that the next release was in three or four month and that a fixed version
  will wait until then. I worked on a patch with the help of Nobu, so the fix
  was available. In other words: Ruby 1.8.6 had not usable thread support for
  four months even though a fixed version existed. Is it not supposed to be
  part of the interpreter core ?
* Ruby/DL had a bad interaction with GC. I spent days tracking the bug and
  fixing it. In the end, I provided a patch taking into account some comments
  from Shyouhei regarding how the finalization process is supposed to behave.
  This patch has been left bitrotting for more than one year. It has been
  integrated in 1.8.7. I don't even know if the current 1.8.6 patchlevel has the
  fix in it. Bottom line: Ruby/DL, which is part of the standard library, could
  crash the whole interpreter. No workaround. Nothing done for more than one year.
* since I had the need for it, I worked on a small patch on gc.c to provide fast
  object allocation statistics that could be used for object allocation
  profiling (and integrated that in ruby-prof). Patch of a few lines. It has
  been rejected on the 1.8.7 on the basis that it should be validated by you
  (matz). In the meantime, 1.8.7 broke backward compatibility. There is
  something wrong here IMO ...

Maybe a better definition of what 1.8 patchlevels should be and when they should
be released could "fix the process". In my opinion, the fact that most of the
development of Ruby is done in japanese on ruby-dev makes it completely opaque
from the non english-speaking people point of view. But I guess that would not
be changed (for good reasons).

I hope that in some ways these experiences won't be repeated for anyone around
there. And I do think that the whole process needs some fixes here and there
(maybe only patches ;-)) to make that happen. Having more than a few Ruby
developpers following bugreports and patches could be a great starting point.

Sylvain