See also: IRC log
<scribe> scribenick: yingying_
<kaz> Web Assembly CG
<jhund> Ben's slides on Webassembly
[Ben from Web Assembly CG is sharing the screen]
Ben: last June we went toward
release.
... WebAssembly in a nutshell.
... data types, functions are flat, state, stack
... data operation: you can call functions directly or
indirectly.
[a picture of trusted and untrusted state]
[compiling c/c++ to WebAssembly]
Ben: can run with similar performance for C/c++
[showing a picture of the module structure]
[Types, Imports, Function Declarations, Function Tables, ...]
Ben: I'll show what the bytecode
looks like.
... the goal are compact, easy to verify and compile and
extensible.
... design stack machine with local variables.
... this is an example of bytecode.
... maybe you are interested in where we are.
... people can still give comments but if not it will go as it
is.
... this is what our roadmap looks like.
... 3 browsers full implementation are expected.
... that's for my presentation. Hope it gave you guys an overview.
Are there any questions from you?
Daniel: features like several return values, JS APIs. Will you bring these features to JS?
Ben: yes. In new version, the features will be added to JS by WebAssembly.
Daniel: you will compile to mini file JS.
Ben: 10-20 percent compaction.
Johannes: people originally write applications in c/c++ not targeted at JS. All the code is run in the sandbox of browser?
Ben: yes.
<Zakim> kaz, you wanted to ask Ben if it's possible for him to send the slides to the public list
Kaz: could you provide your slides to the public mailing list?
Ben: yes.
Kaz: Do you have any idea on capability for state transition?
Ben: I have no contact for
that.
... it does not really benefit the JS.
Zkis: if it's possible that JS can be translated into WebAssembly and used for constraint devices?
Ben: WebAssembly is very low level.
Johannes: if there are high
performance requirements of your code, you could use
WebAssembly.
... it's different from running the JS on constraint devices.
... thank you very much Ben.
... Daniel could you update us the EXI?
Daniel: yes. EXI represents XML
efficiently
... also have XML to support JSON, etc.
... you can process it very fast. Very quick for loading at very
beginning.
... use EXI processor to encode and decode.
... next step is how you could use EXI in JS in the decoder side.
It's still at the very beginning.
... any questions?
Johannes: thanks Daniel
Daniel: we don't change the syntax of
JS.
... it's just the sequence of events that you can jump over.
... you have some pre-knowledge.
... it uses all the built-in types.
Johannes: EXI is built for XML. In
JS, make the parser faster.
... We have been doing some experiments for constraint embedded
devices. I could not make Samsung people join today's
meeting.
... I see both ways (WebAssembly and EXI) seem promising. Maybe
it's good to see every now and then what happened in these two
areas.
... targeting for constraint devices.
Johannes: moved the proposals to https://github.com/w3c/wot/tree/master/scripting/proposals.
... this is where we are so far.
... Zoltan, could you update the UCs we discussed in last call?
zoltan: I haven't yet made it. Will
try to make it for next call.
... I will make a pull request.
Johannes: we are aiming to do the
application layer abstraction. The WG charter is under AC
review.
... the application could run in the runtime from vendor A and then
ports to vendor B, maybe in future on TV. This is the goal of this
TF.
... building blocks: TD, scripting APIs, protocol bindings.
... any other business?
... thank you everybody. See you in next week.
[adjourned]