Deprecate/Drop Support for Python 3.3?

Currently Python 3.3 support is not a major headache for support (unlike 3.2) because we’re largely being limited by Python 2.6 and 2.7 in terms of what features we can support. However, I think it’s important to periodically look at the usage and make sure that we’re not supporting Python versions that are not really being used as no matter what, each version of Python we support incurs a cost in terms of overhead for support (more build matrix items, minor incompatibilities, etc).

With that in mind, I took a look at what % of the pip initiated traffic that PyPI received in the last month to see what our usage numbers look like.

Only pip 8 initiated traffic:

2.7 83.8%
3.5  6.3%
3.4  5.5%
2.6  3.8%
3.3  0.3%
3.6  0.1%
3.2  0.03%

All pip initiated traffic:

2.7 86.8%
3.4  5.2%
3.5  3.9%
2.6  3.4%
3.3  0.4%
3.2  0.07%
3.6  0.04%

Given that 3.3 support is well under 1%, do we want to deprecate support for Python 3.3 with intent to drop support for it in either pip 9 or pip 10? For myself, I say yes-- either as a pip 9 or a pip 10 deprecation.

@pypa/pip-committers ?

[dstufft]

I think I’m going to go ahead and drop support for 3.3 for the next major version of pip (10.0). We’re only a couple of months away from 3.3 being EOL with upstream and they haven’t released a new version since 2014-10-12 anyways so while it is supported, I would argue that it is more like “supported”. In addition PyPy3 is soon to have a 3.5 compatible release, so that is no longer a reason to keep this around.

Ultimately though, I think the usage just doesn’t warrant spending much effort in keeping it here.

[dstufft]

Sure that works for me. The main thing to gain is one less Travis build 😃