[Svnmerge] why python 2.0 compatibility

[Matthew Hannigan]

On Fri, Jul 13, 2007 at 09:31:33AM +0200, Giovanni Bajo wrote:
> On 13/07/2007 3.59, Matthew Hannigan wrote:
> 
> >>Notice that I sticked to Python 2.0 functionalities, since svnmerge.py 
> >>has to be Python 2.0 compatible.  ...
> >
> >I've been meaning to ask, why does it need to be 2.0 compatible?
> >
> >More curiousity at the moment, but I'm looking building some
> >extra functionality using svnmerge.py as a library, and I might
> >need to keep it compatible for the same reason.
> >
> >I had a quick look at the wiki, but I couldn't find the reason.
> >
> >Most linuxes have 2.4 or 2.5 these days; and with the source readily
> >available I thought we could aim for 2.4.
> 
> There isn't only Linux though. BSDs tend to be slower at updates. When I 
> started svnmerge.py, OpenBSD shipped with Python 2.1 (I think 2.4.2 was 
> already out by that day, to put things in perspective).

I just checked the 'legacy' versions of openbsd (3.9) and freebsd (5.5)
and they both have 2.4.x available.  Cygwin have just moved to 2.5.

> 1.5.x is too different to be worth the cost of maintenance, but 2.0 looked 
> like a good compromise. People can still run newer 2.x versions and get 
> more speed (try running the testsuite with different Python versions...). I 
> don't think we're missing *that* much of newer versions anyway.

One thing I missed was the logging module.
Since 2.0 there's more use of iterators and generators and list comprehensions.

> So, I guess the compatibility bar could be raised if there was a *real* 
> need. But I can't see one at the moment.

Well. There's no such thing as a _need_. But I hardly
think you'd be inconveniencing anyone by going to say, 2.2 or maybe
even 2.3.   2.3 is on redhat4.
Jython is 2.2; I'm not sure whether anyone cares about that though.

Would anyone object to an assumption of 2.2? Speak up!


Matt