Florian Angeletti

Note that the feature is unlikely to be ever implemented in ocamldoc. Ocamldoc is in maintenance mode and further development efforts are focused on odoc.

I am fine with both merging the `.gitignore` rule in the manual or moving them, prefixed by `manual/` in the root `.gitignore`. Both variants seem quite similar to me.

Currently, that would break libraries that use both equality between phantom types and the relaxed value restriction. For instance, with type-level addition implemented with ```ocaml type (+'a,+'b) nat val zero:...

It might be simpler to start without this specialization refinement? That would require the use `[@alert "-polymorphic_comparison"]` to disable this warning when using a polymorphic-but-specialized comparison. But keeping the `polymorphic_comparison`...

For comparison functions, the solution used by `base` or `containers` is to restrict the type of the binary comparison operators (`=`,``,``) to `int -> int -> bool`. We could add...

I am in favor of switching to autoconf 2.71 now that most desktop linux distribution have switched, that will make the release process slightly simpler.

I agree that the syntax looks fine, we could also be more conservative and allow just the `let.[:alpha:]+` and `and.[:alpha:]+` subcase (which doesn't change the tokenisation of existing code).

My point was indeed that requiring a `.` dot before the alphanumeric part removes the minor conflict with the current syntax. Another potentially interesting restriction would be to constrain the...

I am curious: any reason to cherry-pick all the individual commits by hand rather than the PR merge commit? (Also I personally like to record the original commit with the...

I think it is certainly worth it to make it clearer in the documentation that both modules are experimental. We could also possibly add an alert? Similarly, for the release...