From: Axel Kohlmeyer (akohlmey_at_cmm.chem.upenn.edu)
Date: Thu May 07 2009 - 10:45:00 CDT

On Thu, 2009-05-07 at 10:19 -0500, John Stone wrote:

> In general, the best thing for Python users to do is to compile VMD
> from source code, so that they can use whatever Python version they
> prefer, along with the existing Python libraries they may already have
> installed. We've always had difficulty pleasing Python users because
> people prefer whatever version they're already using, and since we can
> only compile/link VMD with one specific version, there's always a chorus
> of people that would prefer we use a different one than we've selected,
> due to incompatibilities from one version to the next.

btw: this is not a problem of VMD by itself, but rather of all large
packages that use python and move at a slower pace than the python
development by itself. i've come across similar problems when trying
to set up the plone CMS (using zope/python 2.4). the common resolution
for all of these seems to be to have a specialized complete python
installation to make sure it is consistent with the requirements.

i think an acceptable compromise solution could be to make certain
that the compilation of .rpm and .deb packages would work, which in
turn would be distribution specific and using the native tcl/tk,
fltk, and - of course - python libraries. i've managed to get reasonably
well working .spec files for building redhat/fedora style .rpms
in the past, and am willing to update them and help others to make
them work on their machines.

> I've been using Python 2.5.1 and NumPy 1.0.3.1 for VMD 1.8.7 test
> builds up until now. Before the final release of VMD I'll be
> reevaluating whether I plan to use those versions for the final
> builds, but if people have a strong opinion about it, I'd be happy
> to hear from them now...

i've been building the current source tree successfully against
native python 2.4, 2.5, and 2.6 installations on various flavors
of fedora and red hat and have seen very, very few problems, and
to the best of my knowledge all resulting patches have been included,
with the exception of the rename of the swig based python interface.

perhaps this discussion is the right place to ask, if there is any
interest to keep that one working and have it included. w/o the
rename only one python interface can be included (well, two if you
also count the higher level API, that is written in python and on
top of the low level native python API). i've seen in other places
python programmers complain about swig interfaces being not very
"pythony", so perhaps that one would be a candidate for removal?

cheers,
    axel.

> Since Python is still a relatively fast moving target, my general
> suggestion is still to build VMD from source code and link against
> your favorite Python version.
>
> Cheers,
> John
>

-- 
=======================================================================
Axel Kohlmeyer   akohlmey_at_cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.