python setuptools?

[Kevin Turner]

On Tue, 2007-03-20 at 00:36 +0000, Simon Willison wrote:
> Joe Gregorio's rant is relevant here:
> 
> http://bitworking.org/news/Please_stop_using_setuptools__at_least_exclusively__for_now____

I know setuptools has had a chance to improve things a bit in the year
since that post.  Can anyone tell me if the pkg_resources.require thing
still applies?  I may be persuaded that using a third-party package for
installation is an acceptable cost, but adopting a whole new set of
import semantics or facing reports of "it doesn't work when I used this
option on the installer" is not somewhere I want to go.

The extensive use of .pth files is something that makes me a little
uncomfortable as it is, and the fact that
http://peak.telecommunity.com/DevCenter/EasyInstall#custom-installation-locations has four distinct options (most of which involve editing a new configuration file) isn't making me any less nervous about the whole thing.


And Ben Finney wrote:

> My main motivation for wanting Setuptools support is the automation
> of 
> dependency satisfaction. [...]
> 
> For my case, I could then write a package that simply depends on 
> 'python-openid >= 1.2.0', and the user doesn't need to go through 
> separate steps to get 'python-openid', 'python-urljr', 'python-yadis' 
> all at the right version, and install them. This is a big advantage
> to 
> me as a developer -- it becomes much simpler for potential users to 
> try my application.

It seems to me (from the rapid bug reports I get when I break the
dependency metadata on cheeseshop ;) that Zope (Plone?) is already doing
this with python-openid.  What would either ez_setup usage or .egg
distribution buy us in this regard?

Thanks,

 - Kevin


p.s. Ben, I did request this list be carried by Gmane shortly after it
was created, and Norman Rasmussen resubmitted a request last November,
but I don't see it there now.  Not sure what's going on there.  Maybe
something about our tweaked mailman confuses its subscription attempts?