Dropping support for 2.7 in 2020

Josh Warner silvertrumpet999 at gmail.com
Mon Jul 4 20:12:16 EDT 2016


I think we should support 2.7 through end-of-life. 2020 sounds right.

There are a tremendous amount of systems out there still running 2.7. I 
know more than one lab which standardized on 2.7 years ago, and while we're 
evaluating the move to 3.x it won't be without major pains. Small, niche 
packages which work perfectly under 2.7 break in 3.x, and they may or may 
not be maintained. It only takes a couple of these to render lab equipment 
with price tags in the 6 figures buggy, problematic, or useless in 3.x. 
This is problematic when you want a consistent Python experience across a 
lab or team, but certain pieces just won't work in 3.x. 

Sounding the trumpets is appropriate, as 2.7 won't be around forever. 
Quarterly or bi-annually it would be reasonable to send out a reminder. 
Three years sounds like a long time, but that's about right for this kind 
of transition.

On Monday, July 4, 2016 at 4:55:34 AM UTC-5, Thomas Kluyver wrote:
>
> On 4 July 2016 at 10:47, François Boulogne <fboulogne at sciunto.org> wrote:
>
>> Does it mean that all new version of those libraries will support
>> python2.7 until 2020 or just that you can release a bugfix to already
>> released version supporting 2.7?
>>
>> If it's the first option, it means we can't use python3 features before
>> 2020 and we will be "in late" with these features.
>>
>
> The statement doesn't commit projects to support 2.7 until any specific 
> date - scikit-bio already went Python 3 only with its release last month, 
> for instance. We're agreeing to end Python 2 support by 2020, but projects 
> are welcome to do it earlier than that.
>
> For IPython, we're planning for the 6.0 feature release (some time in 
> 2017) to be Python 3 only, while we'll make bugfix releases of the 5.x 
> series for longer than normal to support users on Python 2.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/scikit-image/attachments/20160704/5c21924f/attachment.html>


More information about the scikit-image mailing list