Andreas Rossberg

@lukewagner, multiple segments would still be needed to support merging of modules.

@askeksa-google: > Some of the most common use cases that I have seen described throughout the discussions are: > > 1. As the representation of integers in a uniform representation...

> > 0. As the representation of small scalars in contexts requiring a reference. > > This looks to me like a general description of the purpose of `i31ref` which...

@RossTate, it's correct that GHC does something special, mostly because of laziness, as described in their ICFP'07 paper. The authors also point out that they are the odd one out:...

It is generally expected that the use of less precise types is less efficient than more precise types. Hence, a module that doesn't need its generality should certainly avoid anyref,...

> I think the comparison to other optimizations like nested structures is reasonable, but the key differences are that we have a fully fleshed out semantics and engine and tools...

@tlively, no, the name section is unrelated. What I was referring to was uses of duplicate symbolic field names I encountered in the J2CL example's .wat file, e.g., like this:...

None right now. It would fall out rather naturally from #274, so I'm happy to defer.

The test suite requirement should be read as tests integrated into the wast-based spec test suite. A JS-specific test suite is of little use to non-JS engines. As @lars-t-hansen suggested,...

Proper wasm threads would be nice to have. :) I see that the first of your links already abstracts out spawn and send functions. Do you think it would be...