[Python-Dev] Adding library modules to the core

[Guido van Rossum]

[ESR]
> 1. Has anybody got a vote on the menubrowser framwork facility I described?

Eric, as far as I can tell you haven't shown the code or given a
pointer to it.  I explained to you that your description left me in
the dark as to what it does.  Or did I miss a pointer?  It seems your
module doesn't even have a name!  This is a bad way to start a
discussion about the admission procedure.  Nothing has ever been
accepted into Python before the code was written and shown.

> 1. Do we have a procedure for vetting modules for inclusion in the stock
> distribution?  If not, should be institute one?

Typically, modules get accepted after extensive lobbying and agreement
from multiple developers.  The definition of "developer" is vague, and
I can't give a good rule -- not everybody who has been admitted to the
python-dev list has enough standing to make his opinion count!

Basically, I rely a lot on what various people say, but I have my own
bias about who I trust in what area.  I don't think I'll have to
publish a list of this bias, but one thing is clear: I;m not counting
votes!  Proposals and ideas get approved based on merit, not on how
many people argue for (or against) it.  I want Python to keep its
typical Guido-flavored style, and (apart from the occasional succesful
channeling by TP) there's only one way to do that: let me be the final
arbiter.  I'm willing to be the bottleneck, it gives Python the
typical slow-flowing evolution that has served it well over the past
ten years.  (At the same time, I can't read all messages in every
thread on python-dev any more -- that's why substantial ideas need a
PEP to summarize the discussion.)

> 2. I am willing to do a pass through the Vaults of Parnassus and other
> sources for modules that seem both sufficiently useful and sufficiently
> mature to be added.  I have in mind things like mimectl, PIL, and Vladimir's
> shared-memory module.  

I don't know mimectl or Vladimir's module (how does it compare to
mmap?).  Regarding PIL, I believe the problem there is that it is a
large body of code maintained by a third party.  It should become part
of a SUMO release and of binary releases, but I see no advantage in
carrying it along in the core source distribution.

> Now, assuming I do 3, would I need to go through the vote process
> on each of these, or can I get a ukase from the BDFL authorizing me to
> fold in stuff?

Sorry, I don't write blank checks.

> I realize I'm raising questions for which there are no easy answers.
> But Python is growing.  The Python social machine needs to adapt to
> make such decisions in a more timely and less ad-hoc fashion.  I'm
> not attached to being the point person in this process, but
> somebody's gotta be.

Watch out though: if we open the floodgates now we may seriously
deteriorate the quality of the standard library, without doing much
good.

I'd much rather see an improved Vaults of Parnassus (where every
module uses distutils and installation becomes trivial) than a
fast-track process for including new code in the core.

That said, I think writing a bunch of thoughts up as a PEP makes a lot
of sense!

--Guido van Rossum (home page: http://www.pythonlabs.com/~guido/)