[Biopython-dev] What would we gain by dropping python 2.3?

[Peter]

>> I was wondering what benefits we would see by dropping support for
>> Python 2.3 after the next release (or next couple of releases?).
>
> Support for Numpy 1.2 ...

We've tested Biopython CVS works on python 2.3, 2.4, 2.5, and are
almost ready for 2.6.  We've also tested Biopython CVS works on Numpy
1.0, 1.1 and 1.2.

The fact that Numpy 1.2 requires Python 2.4+ isn't really linked to
weather or not Biopython continues to work on Python 2.3

> I have not looked at the major Linux distros like Fedora and Ubuntu to know
> when these dropped Python 2.3 for the standard Python install.

According to http://packages.ubuntu.com/intrepid/python and linked pages,
Ubuntu hardy comes with Python 2.3 (very old)
Ubuntu dapper comes with Python 2.4 (pretty old)
Ubuntu gutsy, feisty and intrepid come with Python 2.5

> But I also must add that there is no numpy Windows binary installation for
> Python 2.6 and does not seem likely to be an official one in the near future
> (technical issues with regards to the official Windows binary for Python
> 2.6).

I've been keeping an eye on the numpy list and that is rather
disappointing news - hopefully they can resolve this shortly and maybe
there will be a numpy 1.2.x release for Windows.

>> Note that Mac OS X 10.4 Tiger uses Python 2.3.5, so there could still
>> be a fair number of people out there still interested in using
>> Biopython on Python 2.3 (in addition to my own current Windows
>> development machine).  Before making any plans to drop Python 2.3
>> support we should canvas the main mailing list.
>
> Also some of the older Red Hat / Centos systems still run it - joys of these
> long-term releases.

Yes - this is why I am loath to just drop python 2.3 support without
some benefits.  Some of the linux machine I have access to at work
still run python 2.3 for example.

> How many bug reports are with Python 2.3 from people with an interest in
> Python 2.3 not just testing it?

Our Bugzilla doesn't track the python version, so we can't easily work that out.

> To me the issue is about supporting different versions in the medium term (5
> years) given that NumPy and Biopython will have been rewritten for Python
> 3.0 and most people will be using Python 3.0. I think that if the burden is
> too great to support a Python version it should be officially dropped. Of
> course any criteria bug or feature can be backported to earlier versions if
> requested.
>
> I would recommend that this starts a new minor version i.e 1.5 so it is
> clear that Biopython 1.5+ is Python 2.4+ only.

Biopython doesn't currently have minor version numbers.  On a related
note, perhaps doing the first numpy supporting release as Biopython
1.50 rather than 1.49 would be more memorable / eye pleasing.

> (I also note the recent changes in the cvs that would justify this anyhow.)

Did you mean justify a version number bump, or justify dropping python 2.3?

With hind sight, trying to support both Python 2.3 and Python 2.6 was
more work than I expected - but I think its done now (apart from
Bio.PDB.NACCESS).  If Python 2.7 makes a similar volume of
deprecations needing similar workarounds for Python 2.3, then we may
have more of an incentive to drop Python 2.3.

We've seen some of the drawbacks to continuing to support old python
2.3 while avoiding deprecation warnings in Python 2.6, but what I
wanted to hear was ideas on how any of the newer language features
added in python 2.4 could be useful (in the short to medium term).

Peter