PEP440 vs SemVer

[bjlittle]

I recently stumbled over the whole PEP440 verses semver issue when cutting a recent iris-sample-data v2.1.0-dev ...

We may have discussed the whole versioning standard previously, but if we're considering cutting iris v2.0, it might be worth while opening up a discussion on whether we still want to follow symver or jump ship to PEP440.

As far as I can see we only really care about

• a final release version
• a release candidate (rc) version
• a development (dev) version, but these never get tagged in a release

I guess this discussion might be also steered by the packaging and distribution tools that we use and how they interpret package versions i.e. conda, conda-forge, setuptools, versioneer etc

I would be great if we could have an explicit group agreement on our choice of version standard moving forward, and comply with that across all future scitools package releases.

[Carwyn Pelley]

> I recently stumbled over the whole PEP440...

Ooh not seen this before.

Can I ask what the motivation is for this re-consideration of a version standard?

semver is a rather prescriptive approach which I doubt iris endusers really understand or care about.  This would be a good motivation from my perspective.
PEP440 is by its own admission, less prescriptive than semver, so perhaps it is a step in the right direction.

I'll take a closer look at PEP440 and get back to you.  Could also be of use for ANTS!

Thanks Bill.