Thomas Lively

That's super cool, thanks for the update @alexcrichton!

Whether or not we add those restricted union constructors to the MVP (which I would be interested in benchmarking for code size and performance wins), consolidating the 5 casting instructions...

This sounds reasonable to me. @rossberg, @jakobkummerow, wdyt? I believe this is also orthogonal to what we do with RTTs, which would replace the `ht` immediate in the above instructions...

Using new opcodes and doing an opcode cleanup at the end sounds good. The relevant lessons I learned from doing the same (multiple times!) in the SIMD proposal are 1)...

Yes, in this repo we certainly should remove the old versions. V8 and Binaryen will probably continue supporting the old instructions for some time to avoid unnecessarily breaking builds as...

@rossberg, can you be more specific about what you have in mind so we can better compare it to @askeksa-google's proposal?

I am strongly concerned about the prospect of making the type hierarchy a DAG rather than a tree and am also concerned about introducing a new basic heap type solely...

I meant for that sentence to just say that a function's import signature in one module is the same as its export signature in another module. Can you elaborate on...

Sorry about the imprecise language :) > Soundness is a meta-level correctness property of a language semantics (more specifically, its type system). I could certainly have been more precise about...

> The suggestion in this issue seems to be to use dynamic mechanisms to make up for the lack of static isolation. So let's consider what that would look like....