[stdlib-sig] Breaking out the stdlib

[Antoine Pitrou]

Le mardi 15 septembre 2009 à 21:51 +0200, Georg Brandl a écrit :
> 
> Of course it does them reasonably well.  But for most of these "common"
> use cases you talk about above, a replacement library will be able to
> offer the same or a very similar API, so the breakage is not much worse
> than with md5/hashlib.

Well I do think that md5 shouldn't have been deprecated in 2.x. AFAIU,
it was easy to continue supporting it (by creating a simple stub);
annoying users wasn't worth what we gained in maintenance, IMO.

But regardless, this is not the same situation. Migrating from md5 to
hashlib is simply changing an import line and a module name. Migrating
from optparse to argparse involves an API change.

> I wouldn't say it has few bugs filed against it; sadly I have no hard data,
> but since I read each new tracker item (or at least the title), I would say
> that there are many modules that have fewer bugs filed than optparse.
> 
> Of course, this also is correlated with the module's size (it's currently
> the 12th largest single module in trunk/Lib).

It is also correlated with the module's usefulness. Parsing command line
options is certainly much more common than, say, normalizing unicode
strings or adding fractions together.

Regards

Antoine.