[Python-3000] Splitting out packages [was: Is pickle's persistent_id worth keeping?]

[Stephen J. Turnbull]

Fred Drake writes:
 > On Jan 7, 2008, at 10:29 AM, Guido van Rossum wrote:
 > > Actually I suspect that the people pleading in favor of this are also
 > > doing it for political reasons.
 > 
 > Perhaps.  The motivating factor is really to be able to decouple  
 > release cycles, and I think that's a valuable (technical) reason to  
 > split out packages that aren't essential to make Python a good  
 > language to work with.

XEmacs has tried splitting out its packages, and we are very satisfied
with the results.

However, splitting out the stdlib has had two problems that haven't
been mentioned here.  It causes internal politics (different
developers have different opinions on which batteries are really
necessary, and on which internal interfaces should *stay* internal).
And maintenance of many packages does tend to deteriorate, because
*nobody* pays *any* attention to their bugs.  IMO this has not been a
major problem for us only because when we need a quick patch, we can
usually import one from GNU Emacs.  Conditions and developer culture
are different for Python; these may not be real problems.

If interested, I could probably get Steve Baur (who shepherded the
transition for us) to discuss his experience.