[Doc-SIG] soundex module status?
Skip Montanaro
skip@mojam.com (Skip Montanaro)
Mon, 28 Jun 1999 16:20:25 -0400 (EDT)
>>>>> "Guido" == Guido van Rossum <guido@CNRI.Reston.VA.US> writes:
Guido> Sigh. I wish that module went away -- it offends my sense of
Guido> what should be a standard library facility (like several other
Guido> very old modules, including some I wrote, mind you -- it's just
Guido> that this one's in C, which makes it more of a liability). If
Guido> anything, it should be rewritten in Python, I really don't see
Guido> why it has to be in C. It also makes sense to dig out the
Guido> official algorithm from somewhere.
I'd be happy to shepherd it contrib and into Python as well. I'm sure it
can safely be recoded in Python. I doubt my needs for it would suffer if
that happened.
It sounds to me like Python really needs a way to allow people to easily
incorporate "contrib" stuff into their installations. This is way off-topic
for this list, so I'll only toss out one dumb comment about how easily I'd
like things to be, then shut up.
I like the side-effect documentation of using configure to install Python
and associated bits. I can glance at the top of config.status and see how I
configured Python six months ago (or egcs or binutils or anything else).
I'd love to see something like
./configure --with-thread --with-contrib="soundex msql fred"
in config.status that tells me not only how I configured Python, but what
non-core modules I downloaded and installed. CPAN works similarly, though
doesn't necessarily provide the concise documentation of the installation.
The side-effect of --with-contrib would be to download the named modules
from PPAN (or wherever) and install them along with the rest of Python.
If downloading, configuring and installing non-core modules was this easy, I
wouldn't care much if you decided to cut the cgi module out of the core
distribution.
(That said, I know Greg Ward has been working on installation issues.)
Once something fairly transparent is available, I think Guido should
partition the core distribution into disjoint "offensive" and "defensive"
sets, with the "offensive" stuff winding up in contrib and the "defensive"
stuff remaining in the core.
Skip