February 2020 in-person meeting agenda

[dtig]

What should we discuss at the next in-person meeting? Please comment with agenda items, or things we should discuss so these can be added to the agenda.

I've added some items, and estimated times. Suggestions to remove, or edit time estimates also appreciated.

• Garbage Collection (2 Hr)
• Interface Types (1 - 1½ Hr)
• Fixed-Width SIMD (½ Hr)
• Exception Handling (½ Hr)
• Stack switching / Coroutines / Effect handlers (½ - 1 Hr)
• Feature Detection (½ Hr)
• Type Reflection for WebAssembly JavaScript API (½ Hr)
• Multiple Memories (½ - 1 Hr)
• Typed Function References (½ - 1 Hr)
• Bulk Memory Operations (½ - 1 Hr)
  • Poll to Phase 4?
• Reference Types (½ - 1 Hr)
  • Global function references require two-pass analysis reference-types#76
  • Poll to Phase 4?
• Tail calls (¼ - ½ Hr)
  • Poll to phase 4?
• Debugging topics
  • Limitations of Source maps
• WASI topics (1½ Hr)
  • WASI Embedding API, Interface Types and WASI
  • WASI security model, reference types, interface types, and POSIX
• CT-Wasm (¼ - ½ Hr)
• Custom Annotation Syntax in the Text Format (¼ - ½ Hr)

[tlively]

I'm happy to only spend half an hour on feature detection. I think 10 minutes for an update and 20 minutes for discussion seems reasonable.

[lars-t-hansen]

Reference types to phase 4 is on my Christmas wishlist, too. The main blocker seems to be WebAssembly/reference-types#31 (last comment): getting declared segments into the spec & interpreter. And then at least one more implementation needs to implement that + nullref. So it's possible pushing for phase 4 is premature, but I'd like to explore the possibility.

[dtig]

Thanks @tlively and @lars-t-hansen - I've added an item for reference types, and updated time required for feature detection.

[ErikMcClure]

I would be interested in discussing the current state of debugging support and limitations regarding Source Maps I encountered while attempting to implement pass-through C++ debugging support in an AoT compiler.

[dtig]

@rossberg @lukewagner @aheejin @tschneidereit @binji @dschuff @sunfishcode

Does the time for discussions seem reasonable? Are there other things we should discuss?

[aheejin]

Not sure we need 1.5 hrs for EH. 1 hr will be probably sufficient, if not too long. I'm also planning to advance the proposal to phase 2, but haven't decided if we are gonna do it before the in-person CG meeting or not.

[peterhuene]

I would like to propose a discussion on the C embedding API (or as separate extensions that layer on top of the C API) regarding interface types and WASI.

Topics regarding interface types:

• Checking for support of interface types in the engine.
• Getting interface-types-based import and export signatures for a module.
• Instantiating a module with interface-types-based imports where the engine handles the adapter logic.
• How to handle host string encodings.
• Is it desirable to have an API to describe the interface types custom section for custom embeddings?
• Where does this API live (as part of the current C API or as a layered extension)?

Topics regarding WASI (see this issue for more context):

• Checking for support of a WASI implementation in the engine (alternatively: how to gracefully handle unsupported WASI versions vs. confusing import binding errors).
• Configuring the engine's WASI implementation (preopens, rights, etc.).
• Instantiating a module with imports from the engine's WASI implementation.
• Sharing WASI instances between different module instantiations.
• Where should the API live (e.g. a WebAssembly c-wasi-api repo)?

I think 45 minutes will suffice for this. I'll have API proposals ready to present to kickstart design discussion.

[deian]

If there is room, I would love to talk about Constant-Time WebAssembly (CT-Wasm) and more broadly why and how we can extend Wasm to allow developers to safely write cryptographic algorithms in Wasm. Frank Denis already wrote up a nice and concise summary of why we need this: WebAssembly/WASI#65 (comment).

This topic was brought up before by others: WebAssembly/design#1299 WebAssembly/WASI#65 denoland/deno#3097

[rossberg]

Sorry for being slow to reply.

I suspect that we will need more time for a discussion on GC to make sense (2 hours or more). I want to test the room for some back-to-the-drawing-board ideas around the central issue of casts that might be fairly technical and take time to digest.

If there is room, I'd also be happy to present preliminary ideas for stack switching / coroutines / effect handlers, since I have mentioned them and being asked about them on a number of occasions and feel bad about not having provided a more "official" intro and pointer yet. (1/2-1 hour, would fit in well after exceptions)

[rossberg]

Oh, and given that it's heavily used by interface types, we should probably discuss the annotations proposal again. (1/4-1/2 hour)

[aheejin]

Not sure we need 1.5 hrs for EH. 1 hr will be probably sufficient, if not too long. I'm also planning to advance the proposal to phase 2, but haven't decided if we are gonna do it before the in-person CG meeting or not.

I think we are going to vote on the phase advancement in the meeting. And come to think of it, even including the vote, I don't think we need a full hour, unless we revisit previous discussions. I think 0.5 hrs should be sufficient, or people can have a long break :)

[dtig]

Thanks everyone, I've updated the issue and will shortly be formulating this into a more formal agenda. Please continue to use this issue to propose other topics if any. If you have scheduling constraints and would like to be present for a session, submit a PR to CG-02.md so that any constraints can be taken into account while allocating time slots.

[paolosevMSFT]

The link for "Debugging topics" should probably be https://github.com/WebAssembly/debugging/issues.

[binji]

Deepti added the agenda here: https://github.com/WebAssembly/meetings/blob/master/2020/CG-02.md#agenda-items

Please take a look and see if the order works for you. As always, we'll likely be moving things around up to (and including) the days of the meeting. :-)

[sunfishcode]

In the WASI section, I'd like to propose a presentation, covering the WASI security model, the anticipated relationships with reference types and interface types, and the relationship to POSIX.

Seperately, CT-WASM shouldn't be a sub-item of WASI, since it's a core language feature that isn't specific to WASI.

[dtig]

Thanks @sunfishcode, I've moved CT-WASM out from WASI topics. Could you also provide a time estimate for the proposed presentation?

[deian]

For CT-Wasm 1/4-1/2 should work

[sunfishcode]

@dtig For the WASI presentation, I estimate a half hour.

[dtig]

@lars-t-hansen Looks like WebAssembly/reference-types#31 is now closed, is there more there that may need discussion? If not I can take it off the agenda.

[binji]

There's a related topic here that we may want to discuss if it isn't resolved by then: WebAssembly/reference-types#76

[lars-t-hansen]

@lars-t-hansen Looks like WebAssembly/reference-types#31 is now closed, is there more there that may need discussion? If not I can take it off the agenda.

@dtig, you can take it off the agenda.

[chicoxyzzy]

Would it be possible to attend remotely?

[KronicDeth]

I’d be interested in remote too. I had too many conferences this month to also make the meeting.

[dtig]

Attending remotely is possible, email for the calendar link. Just a caveat though, that given the size of the group this time it's possible that all the discussion may not come through over VC, but all the presentations should work well.

[binji]

Thanks for a great meeting everyone!