serious conflicts between python pks installed via yum vs pip

[Neal Becker]

Matej Cepl wrote:

> On 10.2.2012 18:09, 80 wrote:
>> Python guidelines recommends that packagers installs python eggs using
>> distutils (python setup.py install as recommended in guidelines) while
>> pip use the same install method as easy_install (provided by
>> setuptools/distribute). The former one install egg metadata as a file,
>> the latter as a directory, that's not a packaging/rpm issue.
> 
> a) I don't think the answer “Then don’t do it” is a good one. Some other
> Fedora-packaged languages (Perl comes to mind) allow three levels (or
> maybe even four) of installation of packages (in CPAN meaning of the
> word), system-wide-RPM-packaged, system-wide-unpackaged (to
> /usr/local/*), and per-user-in-$HOME. Not sure how it is with Ruby and
> PHP, but I believe this should be a standard in all major
> Fedora-packaged languages.
> b) distutils v. setuptools conflict is just an unfortunate testimony of
> immature bad state of the Python upstream packaging, but it seems to me
> that generally Python world is moving towards setuptools. Shouldn't we
> follow the suite and move towards setuptools as well?
> c) If we want to have as many Python packages packaged in RPMs (the
> terminology is going to kill me soon) do we have some pip2spec (in the
> same manner as there is cpan2spec)?
> 
> Best,
> 
> Matěj
> 

We will never have all python packages that users may want to try out packaged, 
nor should we want to.  I will only package items for Fedora that I have used 
enough to convince myself that they are:

1. useful enough in general
2. likely to continue to be developed/maintained

I always try out packages first via easy_install or pip (after checking they are 
not already available via yum).

Yes, I could try things via virtualenv, but I'm just not in that habit - and I 
suspect many others are in the same boat.