[Python-Dev] Snakebite, buildbot and low hanging fruit -- feedback wanted! (Was Re: SSH access against buildbot boxes)

C. Titus Brown ctb at msu.edu
Mon Nov 8 03:58:56 CET 2010


On Sun, Nov 07, 2010 at 09:32:33PM -0500, Scott Dial wrote:
> On 11/7/2010 7:09 PM, Martin v. L?wis wrote:
> >> Luckily, the problems that we faced 2.5 years ago when I came up with
> >> the idea of Snakebite are still just as ever present today ;-)
> > 
> > Is this bashing of existing infrastructure really necessary?
> > People (like me) might start bashing about vaporware and how
> > a bird in the hand is worth two in the bush. Cooperate, don't
> > confront.
> 
> +1 Respect your (software) elders.
> 
> The Snaketbite rhetoric has always been less than generous with regard
> to Buildbot, but Buildbot has been providing an infinitely more useful
> service to the community for much longer than Snakebite has for those
> 2.5 years.

Yes, yes, I agree that some graciousness is a good idea.

Oh, wait... you're not helping.

Anyway, I think buildbot is a good local optimum for python-dev, largely
because it's maintained by someone who cares enough to do it well.  And, if
Trent had been talking about buildbot only, MvL's comment would be more than
fair.  But Trent, and I, and others, have talked about quite a bit more than
buildbot being "the" problem. Things like enabling *and maintaining* easy EC2
spin-up with buildbot, or providing SSH key access, or making a 'try' server
available and maintaining it, would be clearly beneficial.  And that's some of
what Trent has been talking about providing.  It turns out it's hard to do
without lots and lots of time and money.  If you truly think it's not useful,
I'd be interested in hearing your opinions, because we've spent an ungodly
amount of the above on it.

In the larger context, I worry very much that we're settling for a rather
suboptimal support setup (on svn, and on cont integration, and on some other
aspects of Python infrastructure) because the current maintainers are so
overloaded and few others are stepping up to bear burdens.  This is a big
concern of at least some people in the PSF.  But it's not an easy problem to
solve - quelle surprise.  And I'm not in a personal position to help, so I've
basically tried to shut up about it :).

As for buildbot, I've been pretty hard on buildbot myself, and I'm happy to
justify it to others -- I've done so in public fora so I'm sure you can find
the records, if you care to look.  But it's not really very relevant to this
conversation, especially since Trent has always been interested in building off
the buildbot setup rather than replacing it.

--titus
-- 
C. Titus Brown, ctb at msu.edu


More information about the Python-Dev mailing list