[Python-Dev] tracker contribution

[Nick Coghlan]

On Mon, Jul 19, 2010 at 9:06 AM, Mark Lawrence <breamoreboy at yahoo.co.uk> wrote:
> Hell, I just wish I was fully healthy and had my MBCS/CEng status back, then
> I'd really feel capable of letting fly.  Having worked on massive UK MOD
> projects (can't say much, Official Secrets Acts and all that stuff) and
> knowing a hell of a lot about change control, configuration management, call
> it what you like, quite frankly Python is years behind.  But there I go
> again, can't rock the boat because someone might get upset, to hell with the
> poor sods putting in their patches years ago and being completely ignored.

I have worked (and do work) on similar projects and in many ways
Python is well ahead of what some large corporations do. In terms of
test automation, baseline control, public auditing of the source
repository, public communications, the very nature of open source
development avoids a lot of the pitfalls private enterprise can fall
into. We *have* to document and automate stuff, because you can't just
yell over at Bob in the next cubicle to ask "hey, what do I need to
install to get [fnord] to build?". We *know* that everything we commit
is going to land in the email inbox of a large number of people, so we
better give it a reasonable checkin message and include a meaningful
comment explaining any apparently-stupid-but-actually-correct code.

The fundamental constraint, though, is that the core developers aren't
paid specifically to hack on Python. We may use it in our day jobs,
and I think some of the folks may get a bit of time to spend on it
during their work week, but making the new versions better isn't the
primary task for any of us. Large corporations, on the other hand,
have a lot more money to throw at things and can pay people to work on
the boring stuff rather than relying purely on volunteers.

If you get to a feature request, find that it doesn't apply cleanly
anymore (or even come close), post a comment to say that. If nobody
steps up to modernise the patch, but it isn't a fundamentally bad
idea, then it's OK for the tracker issue to remain open indefinitely
(e.g. one of mine you commented on recently I had deliberately left
open for a long time because I hadn't made up my mind whether I agreed
with it or not. There were some comments from earlier this year that I
missed at the time, but reading them all now means that I'm inclined
to accept the change for 3.2).

For bug reports, the basic thing is to see if the issue can be
reproduced on currently maintained versions. If it can, update the
version applicability accordingly, if it can't, suggest closure as out
of date.

This isn't a company where the metrics mavens will see a growing count
of low priority feature requests and issue reports and demand that
they either be accepted or rejected and people will be taken from
other tasks and given the responsibility to work through the list. Is
it *good* that things are this way? No, not at all. But it isn't
likely to change anytime soon, either.

Cheers,
Nick.

P.S. Regarding the version bumps with no other comment: I believe
there are some scripts kicking around to bump feature requests up to
the new development version after a new release goes into
maintenance-only mode. It may be good if any such scripts are updated
to also add a comment to that effect, but I don't believe those
scripts are centrally controlled anywhere.

-- 
Nick Coghlan   |   ncoghlan at gmail.com   |   Brisbane, Australia