[arch-dev-public] [RFC] Python 3.3.0 and PEP 394

[Sébastien Luttringer]

On Fri, Sep 14, 2012 at 3:00 PM, Rashif Ray Rahman <schiv at archlinux.org> wrote:
> Xyne didn't appear to have write access so passing on his message as follows:
>
> ---------- Forwarded message ----------
> From: Xyne <xyne at archlinux.ca>
> Date: 14 September 2012 19:57
> Subject: Re: [arch-dev-public] [RFC] Python 3.3.0 and PEP 394
>
>
> Rashif Ray Rahman wrote:
>
>>Good one, we must provide for that. Nevertheless, what I meant was
>>_not_ to have:
>>
>>python-qt
>>python2-qt
>>python-gtk
>>python2-gtk
>>
>>But to have:
>>
>>pyqt
>>python2-pyqt
>>pygtk
>>python2-pygtk
>>
>>As we have now. Prepending 'python' everywhere is fine, as long as the
>>'pyX' name remains. Now the ideal scenario would be:
>>
>>John: Fetch me 'pyalpm'
>>Pacman: Do you want python(3)-pyalpm or python2-pyalpm?
>>
>>Now that I look at it that way prepending 'python' would be better.
>
>
> I also think that prefixing "python" in all cases makes sense. The prefix
> indicates that it is a library/module in our naming scheme and omitting it even
> in the case of naming redundancy (e.g. python-py*) creates exceptions to an
> otherwise uniform rule.
>
> I also strongly support the idea of using
>
> python3-foo
> python2-foo
>
> even if Python 4 may be a long way off. It is a consistent naming scheme that
> is completely unambiguous and it will avoid considerable hassle when Python 4
> is finally released.

I think it's not a matter of time. If python 4 is released with no
major breaking with python 3, update will be smooth and we will never
want to have a python4-foo and python3-foo like we don't want to have
python3.2-foo and python3.3-foo.

Another case, I talked with some python dev last week which explain me
that python3 can be considered as a new language. They also suggest
that /usr/bin/python should remains python2 and new program should use
python3 in there sheebang (as mark of compatibily)

Cheers,

-- 
Sébastien "Seblu" Luttringer
www.seblu.net