Describe the possible extension with function subtyping and exact types

[rossberg]

As discussed at some point in the past.

[tlively]

Thanks for this writeup, @rossberg! If I understand correctly, exact types are introduced to solve one of the problems that the Dispatch Tags pre-proposal solves as well. Would you be ok with adding a link to that pre-proposal in this new section so that readers are aware that there are multiple solutions in flight?

[lukewagner]

lgtm

[lukewagner]

A good follow-up to this would be a sketch of how one could define functions which can't be cast at all or can only be cast using generative/nominal type tags.

[rossberg]

@tlively, AFAICS, there is no particular overlap with dispatch tags (which are more closely related to RTTs from the GC proposal). They would extend indirect calls with new functionality (a new call instruction), whereas this one is just about preserving the existing semantics of call_indirect in the presence of function subtyping. You would need to preserve that regardless of whether dispatch tags are added.

[rossberg]

@lukewagner:

A good follow-up to this would be a sketch of how one could define functions which can't be cast at all or can only be cast using generative/nominal type tags.

Right. In conjunction with the GC proposal, that should be possible by introducing a new form of function definition that has an explicit RTT value. I created issue WebAssembly/gc#99.