What do you think?

The PEP speaks aboutpreserving order, this should also mean, that the order wont changebetween calls. I cannot reproduce the examplegit.remote.pull(), given in the original commit. But I can reproduce therandom order of kwargs between two executions of Python 3.5 with this code:

defpkwargs(**kwargs):print(kwargs)pkwargs(**{'one':1,'two':2,'three':3,'four':4})

This gives me one of these possible outputs:

{'three': 3, 'four': 4, 'one': 1, 'two': 2}{'two': 2, 'one': 1, 'three': 3, 'four': 4}

For Python 3.6 it is stable and preserves the order:

{'one': 1, 'two': 2, 'three': 3, 'four': 4}

The PEP link is also in the commit message:https://www.python.org/dev/peps/pep-0468/

Thanks a lot! Now this PR would be ready for merge even though I feel a little uncomfortable to remove support half a year after py3.5 reached its end of life given that it took 2 years to remove 3.4 support to be able to support inline types which has incredibly high value.

My suggestion is to wait with the merge till Sept. 5, 2020 giving folks another half a year to stop using python 3.5 or face the potential for breakage.

If you would like a merge right now one could add a python version check to keep the sort code for py3.5 maybe along a depreciation warning to let people know the date of removal.

Thanks for sharing your thoughts.

I don't think, shifting the date further back will have any effect. Most of us (the users) will drop py3.5, ony if we notice, that its not supported by the package we are using. But in the end, this is your call.

I also do not have a problem, to make this conditionally on the python version in use, so stay tuned for an update, without the py3.5 drop.

Thanks a lot for your help!

I will set a reminder for myself to remove python 3.5 support on September 5th. It would be super helpful to have a PR waiting which does that - the code was present here but seems to have been overwritten.