Incompatibility between socket and httplib in python CVS build.

[François Pinard]

"Fredrik Lundh" <effbot at telia.com> écrit:

> [...] anyone insisting on using the bleeding edge development CVS version
> (instead of a stable version) can save a lot of time by sub-scribing to
> the python-cvs mailing list.  in that way, you'll *know* what's changed
> in the bleeding edge version.

Just to put things straight, I'm not "insisting" to use it.  Whatever
in production for me still uses 1.5.2.  But there is no chance that I
contribute to the development effort (by reporting bugs, essentially)
if I do not try using 1.5.2+ in some limited environments, where I know
it is bearable that things can break.

But your suggestion is a good idea.  I'm not much of a CVS user, that's
why I did not think of subscribing to anything having `-cvs' in it :-).
But if modifications are described in there, it could be useful indeed.

Thanks for your feedback.

Oh, by the way, is there a specific mailing list to use (and to be) for
discussing findings if any, or is `python-list' just acceptable?

I took a slightly better look at Unicode implementation in 1.5.2+, and feel
quite satisfied by it.  Some of the things are left to users to handle
(like dynamic combining, as per one of my previous messages), but Python
is such that this can be worked out without too much pain.  I now have
the feeling it would be easy to bring my Unicode applications to Python.

Looking at Lib/encodings/, I see that a few are missing from what the
official `recode' has, and would feel like providing now, or soon, directly
in Python, all those encodings which interface naturally with Unicode
(that is most of them, by the way).  In fact, if I go further and consider
the `recode' in pretest, all Asian encodings are missing, which are many.
But since `recode' relies on `libiconv' for those, I believe it would be
much more natural to directly interface Python and `libiconv'.

I'm also seriously pondering the idea of developing a Python implementation
of `recode' that I could use for faster prototyping, and which I would
even include within the `recode' distribution, next to all the C code.
Prototyping, while using C uniquely, requires a lot of energy, and there
are many new ideas to explore.

-- 
François Pinard   http://www.iro.umontreal.ca/~pinard