Re: Proposed Modules, My Take

[Murray Cumming <murrayc murrayc com>]

On Wed, 2005-01-19 at 16:35 -0500, Sean Middleditch wrote:
> On Wed, 2005-01-19 at 21:45 +0100, Murray Cumming wrote:
> > On Wed, 2005-01-19 at 15:35 -0500, Sean Middleditch wrote:
> 
> > > API basically equals ABI in Python terms.  Python *does* break
> > 
> > For certain values of "break". The application continues to work if the
> > application specifies a certain version of python, as it should. I wish
> 
> Right.
> 
> > that python did not allow applications not to specify a version, but
> > it's not the first development environment to make it easy for people to
> > make mistakes. I think that Java needs the same attention.
> 
> The problem is that Distro A with GNOME 2.10 might have Python 2.3 while
> Distro B with GNOME 2.10 might have Python 2.4.

And I'm saying that a sane distro must have both.

>   So now if the app does
> specify version - like it should - it'll only work on one distro and not
> the other.  The funny part is that *both* distros provide the same
> platform in terms of what GNOME specifies.  In other words, GNOME is not
> specifying a platform at all.

So how can GNOME force distros to provide a certain version of python?
Just by saying that it's what we want?

The application has to specify the python version anyway, or the
application will break when a new python version is installed. That
explicit dependency is far more likely to force distros to support that
version of python, because they'll discover the problem as soon as the
application starts, instead of some time later.

So I think that, if application developers try to specify a specific
ABI, then distros will have to provide that ABI.

> If you say that GNOME 2.10 uses Python 2.3, the above situation can of
> course still happen.  In that case, though, one of the distributions is
> shipping a broken GNOME platform.  If nothing else, you're at least
> shifting the blame, right?  ;-)

-- 
Murray Cumming
murrayc murrayc com
www.murrayc.com
www.openismus.com