I suggested once before a looser approval process for the bindings,
since there were a) fewer committers competent to review, and b) the
bindings aren't part of the core software.
We didn't really get a consensus on that proposal, and we seem to be
doing okay with using the three +1's system for Perl bindings at least
-- there are enough people competent to review them.
But for the javahl bindings, we are having a real problem getting
enough votes. Partially, this might be because Patrick Mayweg doesn't
know he can vote on them (otherwise, I assume he would have voted on
them already.) Confidential to Patrick: they're in your domain, so
you can vote on them, see the top section in STATUS for details :-).
But even if Patrick +1's everything, many changes will still be short
of three votes.
This is getting ridiculous. We should *not* hold up the javahl
bindings just because there are only a few experts. There's a reason
we have partial committers maintaining those areas in the first place:
we recognized that we should open the doors a bit wider for non-core
bindings surfaces that didn't pose much risk to Subversion itself. We
should be applying that same philosophy to getting them into 1.0, else
why did we have the looser commit policy in the first place?
I'd wanted to review/vote on these changes myself. All the ones I've
checked so far look good. Unfortunately, I don't have the environment
or expertise to do a truly thorough review & testing. Now, I could
just vote "+1" in the faith that someone else has done their homework
(and I will, if we can't solve this any other way), but that would be
a travesty of the review process. It'd be much better to simply admit
that we need to treat non-core areas differently.
Here's my specific proposal:
A bindings change can go in with a +1 from either a bindings
maintainer for that area or a full committer, and at least one "+0" or
"+1 (concept)" from any other committer.
The goal of this system is to make sure there's at least two pairs of
eyes on the change, but not to demand that every reviewer have the
same amount of expertise that the bindings maintainer does. One can
review for general sanity, accurate comments, obvious mistakes, etc,
without being forced to assert "Yes, I understand these changes in
every detail and have tested them."
If people prefer, we can apply this system to just the javahl bindings
for now, but I think it would be more comprehensible if applied to all
the bindings. And after all, nothing is stopping someone from
reviewing a change if they want to, no matter how many votes it
already has.
-Karl
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Jan 13 18:24:19 2004