Re: [Gnu-arch-users] GNU Arch status update

[Tom Lord]

   From: Ben Finney <address@hidden>

   The roadmap says:

       "GNU Arch 1.3.1 15 Mar 2005 -- This will be primarily a
       maintenance/rejuvination release of GNU Arch 1.x. Experience has
       shown that permitting the baz line to serve as the de facto mainline
       for GNU Arch 1.x will be highly problematic."

   For those who came in late, can someone point to these experiences
   (blog entries?  mailing list thread URL?) showing that following
   baz will be "highly problematic" for GNU Arch?

>From my perspective, 1.4pre1 has some serious regressions.  For example,
it does not build at all on one of my boxes (on which earlier releases
had no problem) and, once tweaked to build, fails `make test'.
The box in question is an exceptionally clean, robust, and rich Posix
compatible environment hence my labling this regression "serious".

In the pipeline (I am told) are baz changes that will make that
even worse (e.g., a deliberate hard dependency on GNU libc).

And from my perspective some of the new features and core design
changes in baz, while certainly interesting and worth trying out,
require more work than it appears they will get if the GNU Arch
project continues to coast.  (I guess you can see a *little* evidence
of that in recent gnu-arch-dev threads such as the one about standards
documents that seems to have petered out and a few about the baz
`merge' command and new archive format.  There's baz guys on the one
hand explaining just fine the why's and wherefore's of how they got
where they are and where they are going; there's both audible and
silent folk on the GNU side noting that their concerns are coming out
as low priorities in this mix.  Probably there is some brilliant
economic analysis one could do that would explain why this tension
exists in this particular case but it hardly matters why --- there it
is, and there's no easy way to make it disappear.)

Basically, in spite of good efforts all around, on both branches, 1.4pre1
falls way short of my quality expectations and goals and the expense of
directly trying to fix it from its current state looks quite daunting
to me.   Don't be confused about that though: my judgement re 1.4pre1
doesn't necessarily imply anything at all about whether or not baz itself
is doing a good job of achieving the goals for which that branch was 
created.

Finally, I should point out that the 1.4 series isn't being killed or
even direction-shifted.  It's still there.  Should my judgement prove
wrong, and coasting on Baz's wake again appear the right thing for GNU
Arch to do, the option is still open.  Meanwhile, the 1.3.X series is
a context in which to take a more proactive approach to GNU Arch
(including, presumably, merging at least /some/ changes from baz).

Does that clarify things a bit?  People often complain that I seem to
always express myself with a "negative tone".  That's certainly not my
intent here.  1.4pre1 was a valuable experiment, in my book, and the
final chapter on merging from baz to GNU Arch is far from written.

-t