In discussions with package maintainers I found that it might make sense to have a higher pace in Coq Platform than in Coq. I wonder if we should have a Coq Platform 2021.05 release.
I plan to have only one branch for all Coq platform releases with multiple package files but all scripts shared, so that this would be rather light weight. The scripts would ask which version of the platform to install.
For the sake of releasing the next version quickly, I would favor going for a minor release now.
Also for the sake of not being totally inconsistent with what we announced at the time of the 2021.02.0 release.
And each new major release will require a track creation for the Snap package, and this is not totally lightweight since you have to ask humans to do it and it may take some time.
@Théo Zimmermann : the idea was to have both, a 2021.02.1 bug-fix / patch release and a 2021.05 release.
Then, I think the focus should be on 2021.02.1 and only when it is out of the way should we consider whether a new major release would be useful.
Well this question has a slight influence on the decision which new packages take into 2021.02.1.
The "deal" was to add none, then hott came up, so we had 1 exception. Which other packages do you have in mind?
@Enrico Tassi : well afair the rules said that to a patch release new packages can be added - it is just not allowed to make possibly breaking changes to existing packages.
The package I am thinking about is CoRN.
yes, I meant "the plan". Anyway, I'm not a big fan of the release speed of Coq, I don't see what it brings us (we could release every year IMO). So for the platform, I don't know. I think 6 month is pretty quick already. Can you remind me the main arguments of the package maintainers?
I mean, you can always opam install coq-x.new-version atop the platform, and/or build new instalelr by just forking the repo (CI builds them for your fork). So I don't see myself what it would bring to us and to them.
Indeed. I mean it is still all in a flow and we can experiment what makes sense. If a release it light weight, I don't see a good reason for not doing it. If it is more heavy (Snap) it is a different story.
One package which has a major update is VST, and since it takes a while to compile the pre built installers are especially useful.
The main reason for having a faster pace would be to make decisions what is included when easier.
Are you saying: If we release every 3 months, then minor releases have reasons to add packages
Conversely, if we release every year, we are forced to accept new packages in at minor release time
This is independent of the minor releases which are just there for bug fixes. The "intermediate" releases are more useful for publishing possibly breaking feature updates to existing packages than to add new packages, say VST 2.8 or Flocq 4 or so.
Anyway, my point is that 2021.02.1 should come up ASAP since it will provide an official macOS platform installer.
So IMHO, HoTT should be the only exception for this release.
2021.02.2 could follow shortly after with even more packages.
Iris seems to me like an important candidate (maybe even more than CoRN).
Thinking about it we can now anyway have several 2021.02.X releases with additional packages. A 2021.05.0 release would only be required for considerable changes to existing packages. So the 2021.05 yes/no question does not really influence the question what to add to 2021.02.1.
And yes, Iris is on my short list ...
Last updated: Jan 30 2023 at 11:03 UTC