[Biopython-dev] Ideas for Biopython 2.0

[Peter Cock]

On Wed, Jun 21, 2017 at 7:16 PM, Tiago Antão <tiagoantao at gmail.com> wrote:
>
>
> On 21 June 2017 at 11:44, João Rodrigues <j.p.g.l.m.rodrigues at gmail.com>
> wrote:
>>
>> Regarding the pledge, the more I think about it the more conflicted I
>> feel. On one hand, this will push adoption of Py3.x for the end-user and
>> allow us to take advantage of new language features. On the other hand, I
>> still work on clusters with system Py2.6 and with code that chokes with
>> anything over 2.5.
>
> This will mean that we, with our scant resources would be given more support
> than many of the other projects out there. More than CPython, or some
> scientific libraries.
> With regards to existing installations: I think the argument works the other
> way around: The Python community is mostly deprecating Python 2 in 2020 so
> system administrators and individual users have plenty of time to adapt,
> lets hope everybody is competent enough. If not, I do not think it is fair
> for us to be supporting the burden people that were well warned in advance
> of this change and dud not take appropriate action.

Our local cluster is still on CentOS 6 right now, where the default system
Python is only Python 2.6, which Biopython already does not support.
It is comparatively straightforward to install a recent Python, even under
a $HOME directory without admin rights, and deal with isolating package
dependencies via Python's virtual environments (venv) or similar.

(We do this kind of thing right now by installing Python 2.7 in order to run
Galaxy under CentOS 6.)

So while I accept that even after 2020 there will be many academics
stuck with compute resources still using Python 2, this is not a real
barrier to actually using Python 3.

>> For Biopython 2, I think it might be reasonable to say look, we support
>> only 3.x, and maintain versions 1.x supporting Py2.x but dropping support
>> altogether might be a bad move.
>
> This means revisiting a previous discussion and mostly be alone out there
> compared to many Python projects.
> This is less bad than having biopython 2 supporting python 2.
>
> Notice that maintaining Python 2 support after 2020 might be problematic: at
> least in the code parts that depend on libraries that will require Python 3.

Taking NumPy as an example, once they drop Python 2 support anyone
really wanting to use (an old version of) Biopython 1.xx under Python 2
would just have to install an old version of NumPy as well.

Peter