Ross Tate

Interesting use case. More generally, it seems like JS interop should respect object identity. Thanks!

Um, none of the short-term motivations mention privacy, and I mentioned that this would probably be workable for structural types. Also, I thought the plan everyone just agreed to was...

Thanks for clarifying, @tlively! I see your reasoning now. My concern is that my understanding of yesterday is that we were not going to be choosing a nominal vs. structural...

Ah, good point. Is there a public collection of production-scale binaries to examine? Of course, there are motivations besides smaller size that are worth discussing in parallel. I remember @jakobkummerow...

Wow, thanks @askeksa-google for performing the experiment! That's interesting how compression cancels out the savings. I had hoped the numbers would give a clear win one way or the other,...

Ah, that is compelling. For completeness, I believe there is a way to solve the multiple-bounds problem for the index-only approach. What you have to do is—rather than a bunch...

I had a chance to consider the hierarchy approach in more depth and was able to make it much simpler than what I described above. That said, I can't figure...

I agree that we don't want 3 loads for a field access (especially chained ones!). But I don't believe what I'm suggesting here makes that necessary. Let me break it...

Ah, thanks for the clarifications. But now I have a question about how things are currently implemented. I had thought that WebAssembly modules were designed so that things like globals...

Oh wow, that's grim. Thanks for the improved understanding (and for clarifying load-count vs. chain-count).