Note about versioning scheme changes about website HOT 3 CLOSED

I found the discussion here. The gist is that pip would need to change its handling of implicit epochs for it to work well, and I think it's solvable if folks want to work on it (probably warrants a PEP, which is tiring to consider for such a relatively narrow use case). I think it's worth writing up how epochs could work that would make this transition more feasible. It doesn't have anything to do with what epochs mean (pep440 covers that), the issue is mainly in the installer's handling of unspecified epochs in version constraints, which I don't think PEP440 dictates. It's also worth mentioning that epochs are a PEP440 thing that pip may understand (if not how we'd like it to), but other packaging systems like conda may not have equivalents and that becomes a second hurdle to adoption in packages that are distributed via multiple channels.

from website.

Thanks for the response @minrk. That pretty much sums up my feelings too. This discussion seems to have come to an end so I'm going to close this issue.

from website.

Hey there! Thanks for opening this.

I similarly thought epochs were the answer to this, but they don't quite work the way users would intuitively expect so may cause more of a headache.

I was discussing this with @minrk and others a few months ago (although I can't find the original conversation now). Check out this repo with some examples of the behaviour and challenges https://github.com/minrk/epicepoch

from website.