NEP29 blog post

[orbeckst]

As stated in MDAnalysis/mdanalysis#3560 we need to announce that we are following NEP29 (mostly).

@IAlibay knows best what we decided to do.

[orbeckst]

Given that we use Google Colab for workshops and that Colab only supports Python 3.7, are we hurting ourselves by not providing MDAnalysis for 3.7?

Or are we just waiting for Google to finally upgrade?

[IAlibay]

Yeah that's a real pain, the ML workshop had to be delivered using 2.1.0.

There's not much we can do here unfortunately, colab is extremely slow moving, I think it's still using Ubuntu 18.04 as an image. I wouldn't recommend waiting on them, once sklearn, torch et al. have enough non py37 releases they'll be forced to do the update.

[IAlibay]

Just to add here, 18.04 drops support by next April, so they'll have to do something by then. Although NEP29 drops py38 at the same time - it's why we can't guarantee NEP29, just that it's going to be our upper limit.

[orbeckst]

Ok, we'll just go along with the ecosystem. And tell users to complain to Google ;-).

[orbeckst]

Perhaps it might be worth mentioning in the blog post the implications of NEP29 for using Google Colab.

[IAlibay]

Sounds good, I think we can probably keep 3.8 around a bit longer, in reality we usually don't drop until ~ 6+ months after the NEP deadline anyways, purely because the latest version of Python always takes forever to roll out upstream. I think we're be a bit more eager to drop numpy versions than working Python ones

[orbeckst]

@IAlibay is this blog post still something we need to do? Or should we just roll this into the pre-amble of #198 ?