Another long-term issue for the platform: make all platform package maintainers cryptographically sign their releases that become part of a platform release. I documented here how one can check that a Coq tarball was actually signed by a Coq release manager, and similar practices should apply to the platform: https://github.com/coq/coq/wiki/VerifyingReleases
unfortunately, a "Coq platform key signing party" to increase trust may be some time off in the future, and not only due to the time it takes to get the platform established...
Why does this page explain how to verify the signature of the commit instead of the signature of the tag itself?
@Théo Zimmermann as far as I can tell, there is no dedicated git command for verifying a tag, git verify-commit
is all there is (or maybe it was added recently)
also, it seems the V8.11.0
, V8.11.1
and V8.11.2
tags were never signed :shrug:
Sigh! This is despite the release documentation giving exactly the commands to use to prepare the tags.
I guess I can update for 8.12 and add a note about instructions (for verifying release locally) not working for 8.11. It's not really feasible to describe both workflows at the same time (verifying a tag vs. verifying a commit)
We could also ask @Pierre-Marie Pédrot (not subscribed to this stream) to recreate the tags for 8.11.
but wouldn't that change the tarballs/checksums and lead to a lot of work in opam, etc.?
I don't think so. The tarballs only depend on the content in there, not the tag meta-data AFAIK.
I think it would make sense to have a meta field in opam to specify how sure one can be about the authorship and check this. Btw.: interestingly opam does not fail when the given hashes are wrong. It spits an an error, but continues happily.
@Théo Zimmermann so should I open a Coq issue about the missing tags? it may become a problem in the future if we integrate authorship verification in some pipeline
Sure!
Last updated: Jun 03 2023 at 05:01 UTC