Rack: Rack 3.0 minimum Ruby version

[ioquatix]

By the time we release Rack 3.0, I believe Ruby 2.4 will be EOL.

There are a few functions we would prefer to use, including #tranform_keys.

Do we want to support EOL Rubies? What would be a reasonable policy?

[matthewd]

It's certainly not the deciding factor, but FWIW, the Rails 6.x series requires Ruby 2.5.

If anything, that's mostly relevant in the other direction: we're ~going to~ _expecting_ to support 2.5 for a new release in 2021, so we'd prefer new Rack releases to also maintain support for that version until then (because otherwise we won't be able to use them).

That desire aside, I think it's generally reasonable to drop support for EOLed Ruby versions during major version bumps, even for deep infrastructure level projects like Rack. The disruption of the library upgrade will dwarf any disruption from a corresponding Ruby upgrade; it's not reasonable for people to expect us to help them be able to do the hard thing while skipping the easy one.

[simi]

Just to remind, there are some stats available at http://ecosystem.rubytogether.org/ and 2.3 seems still widely used for some reason.

[natematykiewicz]

As of today, Ruby 2.4 support has ended. https://twitter.com/rubylangorg/status/1249125876492615681?s=21

[jeremyevans]

IMO, central infrastructure such as rack should only drop old ruby versions when doing so becomes a maintenance burden. There is some Ruby <2.4 compatibility code we could remove (dropping 2.3 support), but I don't see any reason to drop Ruby 2.4 support just because it is no longer supported by ruby-core.

[natematykiewicz]

I just mentioned it, since the original post mentioned Ruby 2.4.