Re: Release 2.12

[Daniel Jacobowitz <drow at mvista dot com>]

On Thu, Oct 25, 2001 at 12:13:33PM +0930, Alan Modra wrote:
> On Wed, Oct 24, 2001 at 09:50:03PM -0400, Daniel Jacobowitz wrote:
> > On Thu, Oct 25, 2001 at 11:08:41AM +0930, Alan Modra wrote:
> > > On Wed, Oct 24, 2001 at 09:26:54PM -0400, Hans-Peter Nilsson wrote:
> > > > I'd like to propose H.J. as release manager for a 2.12 branch.
> > > > Assuming that you (H.J.) are interested, and that you would
> > > > consider doing the lot, not just GNU/Linux.
> > > 
> > > I like the idea.  I was going to suggest that the 2.12 fork should
> > > coincide with one of HJ's releases in order to benefit from the testing
> > > that H.J. puts into his releases.  As HP commented to me in private
> > > mail, H.J. is already acting as release manager.  The number of people
> > > using HJ's releases should speak loudly; if he was doing something wrong
> > > there would be a large scale migration of linux users back to FSF
> > > releases.
> > 
> > While I support the idea, I'd like to strongly disagree with the last
> > part of your reasoning.
> 
> Which part?  "if he was doing something wrong there would be a large scale
> migration of linux users back to FSF releases."?  Regardless of your
> opinion regarding the level of testing, I'd say it was fairly obvious
> that most linux people see HJ's releases as "better" than the FSF ones.
> Mind you, I run CVS head binutils all the time.  ;-)

Yes, that part.  And it's certainly not "most".  Consider:
 - Debian uses them.  That's Chris C.'s decision, and seems to be a
reasonable one, overall.
  - Hard Hat uses them, but we're considering switching back to the
trunk, especially if a 2.12 is forthcoming.  There are obvious
advantages to working from a real release branch.  Three of us made
that evaluation.

These are probably representative: there's no room for a large
migration; most users don't just choose their own.  But that misses the
real point, which is below.

> > Both Debian and Hard Hat use H. J.'s binutils for roughly the same
> > reasons - they have marginally more testing than "pick a day, take a
> > snapshot".  Marginally.  If there was some better cross between every
> > eight months and weekly - for instance, if the release branch received
> > active bug fixes, when critical bugs were found - then things might be
> > quite different.
> 
> The real question here is "why don't debian and hard hat use FSF
> binutils releases in preference to HJ's releases?"

Because critical bugs are found in the released versions and not fixed. 
GNU Libc, for instance, has an unpleasant habit of depending on
binutils not-yet-released.  The patches don't tend to be easily
back-portable for those without a history of binutils experience.  If
we want to make new programs work, we need to move forward fairly
frequently, and HJ offers the only way to do that.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer