Redhat and Python 2.x.x

[Steve Holden]

"Trond Eivind Glomsrød" <teg at redhat.com> wrote in message
news:xuyofjm6tf6.fsf at halden.devel.redhat.com...
> Gordon McMillan <gmcm at hypernet.com> writes:
>
> > Steve Holden wrote:
> >
> > > It would be useful to define some sort of "shared" area into which
> > > modules/packages acceptable to several different distributions could
be
> > > installed once. However, extending this to mutliple versions would get
> > > confusing: a directory for 1.5.2 and 2.0 packages, a directory for
2.0,
> > > 2.1 and 2.2 packages, etc.
> >
> > While the source might be the same, the .pycs aren't, so somebody
> > would suffer greatly.
>
> Not only that, most such libraries would have C components as well
> (e.g. newt).
>
True. So that would exclude them from the set of "modules/packages
acceptable to several different distributions". Having said this, perhaps I
can add that my original remarks were motivated by the tiresome necessities
of maintaining multiple Python platforms on a single machine.

My Windows 98 laptop currently has ActivePython 2.0, Cygwin Python 2.1.1 and
PythonWare Python 2.2b1. It's not easy to maintain a compatible set of
modules over just this set of distros, let alone introducing my SPARC RedHat
6.2 and Intel RedHat 7.2 into the picture, along with a Windows 95 Python
2.0 final and some flavor of 2.1 on a couple of NT systems. Some extensions
simply aren't available to all distros/platforms, MySQLdb/Cygwin on Win98
being a current thorn in my flesh (the MySQL libraries I have on Win98 are
compiled with mingw32).

Obviously it would be desirable to keep these machines all in step as to
Python levels, but I've spent some time recently trying to ensure that a
bunch of code was reasonably cross-platform. Hence my wish to find some way
that simple (i.e. pure Python) modules might be used from a common
directory.

Given the byte-code differences between major versions, as Gordon pointed
out it would not be practical with current Pythons, since they all try to
create the .pyc file in the same directory they find the source. Obviously,
compiled C, C++ et al give even greater problems. It would be nice to think
that we could make code management problems easier in the future, but I'm
not suggesting I yet have any practical ideas. Sorry.

Maybe this will be discussed at Developers' Day.

regards
 Steve

--
Consulting, training, speaking: http://www.holdenweb.com/
Python Web Programming: http://pydish.holdenweb.com/pwp/