Re: Alpha Releases: Docs? - Mailing list pgsql-hackers

Robert Haas wrote:
> On Wed, Aug 5, 2009 at 6:59 PM, Josh Berkus<josh@agliodbs.com> wrote:
> >> As far as the release notes, I think we would have to have proof that
> >> the alpha-generated release notes are as good or close to the quality of
> >> the release notes using the current process. ?If they are, we can use
> >> them for 8.6, or even for 8.5 if the quality is similar, but we can't
> >> know that without creating identical release notes for 8.5 and comparing
> >> them, to make sure the alpha process has not missed any items, etc.
> >
> > I can't speak for Robert or Peter, but for me this gives me exactly zero
> > incentive to bother. ?If you're just going to do the same amount of work
> > anyway ... and potentially delay the release by just as much ... then
> > there's really no point on me spending my nights and weekends wrestling
> > with SGML formatting. ?I'll leave it to you.
> 
> I think I am in agreement.  Parsing Bruce's words carefully, he seems
> to be saying that the only way to determine whether the release notes
> are of sufficient quality is to repeat the whole process of release
> note generation ab initio to determine whether what has been produced
> is good enough.  Presumably this would be followed by some comparison
> of the two work products (by a panel of impartial judges?).
> 
> I can't believe this is necessary.  It ought to be possible with
> careful bookkeeping to make it easy to verify that every commit has
> been either included or intentionally omitted.  The obvious system
> that occurs to me is to track the git hash of each commit and the
> release note text associated with it, but I am sure there are other
> unique identifiers that could equally well be used.  Once you've
> verified that, then the only remaining issue is the actual quality of
> the work product, and I would think that it could be much faster to
> edit someone else's work than to do the whole thing over.  Peter and
> Josh both have excellent written communication skills, and I like to
> think that I do as well; I would think that the necessary work would
> be more on the order of fine-tuning than a wholesale rewrite.
> 
> That having been said, I am not going to spend a lot of time trying to
> push water up a hill.

I would love to get out of the release-note-writing business, but I
can't imagine how such a document could be written incrementally, so it
is logical that I would want some kind of test to see that the method I
didn't think would work would actually work.

I could state right now that I will not do any 8.5 release notes and
force folks to cobble something together, and hope it works, but that is
hardly repsonsible.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + If your life is a hard drive, Christ can be your backup. +