Dimitri Fontaine <dimitri@2ndQuadrant.fr> writes:
> "David E. Wheeler" <david@kineticode.com> writes:
>> You keep making extension authors have to do more work. I keep trying
>> to make it so they can do less. We want the barrier to be as low as
>> possible, which means a lot of DRY. Make it *possible* to do more
>> complicated things, but don't *require* it.
> So really, using %v to say "put the current version number here" does
> not seem like a problem for me, it allows me not to have to think about
> *any* files naming rules nor version numbering scheme.
Maybe I misread David's meaning, but I thought he was saying that
there's no value in inventing all those control file entries in the
first place. Just hard-wire in ALTER EXTENSION UPGRADE the convention
that the name of an upgrade script to upgrade from prior version VVV is
EXTNAME-upgrade.VVV.sql (or any variant spelling of that you care for).
What is the point of letting/making extension authors invent their own
naming schemes? That has no benefit that I can perceive, and the
disadvantage that lack of uniformity will confuse users.
As for the question of what characters should be expected in version
numbers, +1 for digits and dots only. There's no good reason for
something else. Even the Debian document you quote points out that
hyphens in upstream version numbers give them problems, and Red Hat
style packaging rules flat out disallow hyphens. (hyphen-something is
for the packager to use, not the upstream software.)
regards, tom lane
