[SciPy-Dev] Required Python Version

David Cournapeau cournape@gmail....
Sat Jul 17 16:33:14 CDT 2010


On Sat, Jul 17, 2010 at 2:25 PM, Ralf Gommers
<ralf.gommers@googlemail.com> wrote:
>
>
> On Sat, Jul 17, 2010 at 8:06 PM, David Cournapeau <cournape@gmail.com>
> wrote:
>>
>> On Sat, Jul 17, 2010 at 4:37 AM, Charles R Harris
>> <charlesr.harris@gmail.com> wrote:
>> > Hi All,
>> >
>> > Is there a good reason we shouldn't raise the required python version up
>> > to
>> > 2.5?
>>
>> Yes, a lot of "enterprise ready"  distributions still use python 2.4
>> (RHEL, Centos).
>>
> That's not a convincing argument for an infinite amount of time. People who
> value "enterprise ready" meaning they run ancient stuff should be perfectly
> fine with numpy 1.4 + scipy 0.8 for a long time from now.

That's not true in my experience. 2 years ago I used those "ancient" stuff.

> And otherwise they
> can upgrade python itself quite easily.

that's not true at all. True, you can update python itself easily, but
updating things like pygtk or pyqt are near impossible without the
support of IT staff (because you generally lacks X11 headers, or gcc
is too old).

>  2.4 doesn't even get security
> updates anymore.

That's not exactly true - python developers will not give security
updates, but I think RH may still do so.

> We now support python 2.4-2.6, for the next versions we'll add 2.7, 3.1 and
> 3.2. Supporting more versions has a cost. And it's clear that the amount of
> people running 2.4 from svn is at or close to zero, because recent syntax
> errors for 2.4 have gone unnoticed for a long period.

People running svn are a small proportion of our userbase. And the 2.4
support cost is fairly minimal IMO. Certainly, 2.4 support is more
important than 3.x at this point.

David


More information about the SciPy-Dev mailing list