Not enough Python library development [was PEP scepticism]

[Ralf Muschall]

Grant Griffin <not.this at seebelow.org> writes:

> called "Python Program Archive Network", or PyPAN, for short. That might

I'd say naming and content are different beasts.  People know ctan,
cpan and maybe more, so something syntactically similar would have
maximal chances of being found.  py[cp]an seem to be unused.

Unfortunately, www.cyan.org (IMHO the coolest possible name) is
already used unrelatedly (but only by a pair of index.html and
index2.html which refresh to each other in an endless loop), so only
cpyan (which seems to have been registered by some Pehr Anderson to do
the right thing) remains.

> concept seemed to encourage large interdependencies between modules. 

Unfortunately, the logical alternative to interdependencies is
duplication of code (or, at least, of functionality).

> "application-targeted" distributions.  For example, someone could make a
> "Numeric" bundle of Python, including NumPy and related modules;
> likewise, there could be a "webmaster" bundle.  Of course, this takes a

This does not remove the need to keep them mutually compatible.

E.g. I did a client-server-pair some years ago, where the server
analyzed result files (a huuuuuge number of them) produced by a
proprietary numerics package, and the clients (numpy or STk depending
on what I needed) plotted and evaluated them in hairy ways (the server
machine had only perl, so all I did there was <>, m//, cacheing and
UDP).  Having e.g. sockets and numerics dislike each other (or
requiring to download the distro twice in two versions and sorting it
out) would make that much harder.

Ralf