Rust Programming Language Rust and WebAssembly Working Group, 27 Sep 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 27 September 2018 Meeting

Description

The Rust WebAssembly Working Group meeting from 27 September 2018.

A

Hello welcome to another restaurant assembly working group meeting.

A

We don't have any new RCS today, just jumping right into the schedule. We probably should make a little bit of progress on the NPM packages. Rfc I think that's stalled, a bit, that's just kind of a note to self and then I think also. Maybe it makes sense to.

A

Was it closed or like decide not to do the trait inheritance RFC and then write up a new DRF RFC? Instead, I can probably move to do that later. Today, all right and I don't think. There's any new issues tagged for the release candidate.

A

Yep, no, nothing new, all right so without any further ado jump into 2018 goal status updates.

A

So the big news for lesson been Jenny stuff is that we published web sis.

A

So if you didn't see the announcement blog post check out the rest, isin blog.

B

A

And then, yes, there were some a couple things that just the act of publishing ended up, uncovering one how bad publishing, although asthma, by train crates still is.

A

There's a bunch of no interface object, stuff that the format of our warnings was different, and so it didn't get caught by the analysis and audit they did before and kind of the bash scripting. So there's like 70 something interfaces that are marks, no interface object and we don't expose at all and I. Think.

A

The best thing to do here is to just consider them structural, because what no interface object means is that basically, like the constructor and its prototype aren't visible like through the window, the only way to get at them is to like have an actual instance and then ask for the prototype and then like ask for the prototypes constructor.

A

So since we can't really make like ahead of time, bindings I think it just makes sense to just consider them a purely structural. Do you think this.

C

Will be a breaking change or you just think: okay, just throw it in ahead, more appearance, yeah.

A

This is additive because we're just skipping them completely right now, yeah.

D

A

Yes, that's the one thing and then the other was audio context on web assembly or on WebKit has a prefix and we haven't really talked about how to deal with like prefixes at all. Yet, but all.

E

A

Yeah, so on WebKit, it's not called like audio context, it's called WebKit audio context, and so I don't know if the best way to handle that is like to make better in websites or to try and like add custom attributes to. As imagine for us to kind of do it in the bindings that everyone else can just use audio context. Yeah.

D

Just one for difference to make sure that it works rather than working that for the core.

A

I'm sorry I can't understand what you're saying there's a lot of background noise. Yeah. Okay, now is it fine? You know it's a.

F

Little bit better.

E

Well, I think I go through. You was talking about poly, fitting right, we're in college. What if it is I.

A

Mean this isn't quite as drastic as like, including a polyfill. It's the exact same API. It just happens to have a different name and like the way that people do this in JavaScript is they say, like audio context, equals like WebKit, audio context or or audio context and I guess my question is: do we want people to write the equivalent of that in rust, or do we want to have Wesson bind gen automatically? Do that kind of dance in the generative bindings I.

E

Mean the thing is that you you're already using wasn't binding as some sort of in interoperability layer, so I wouldn't expect to write this kind of logic. I would expect the generated code to handle that, for me,.

C

Yeah okay I feel like we should have like some sort of whitelist in the build script. Saying that like extend that ast for as I minded a little bit more to say, like you, can check these names as well and basically just generate exactly that code by saying audio contracts equals to equals, WebKit or new context, or that and we just somehow plumbing the generated bindings.

C

A

You think so I think it makes sense to have like I'm wondering what the the UI would be and I think like. It makes sense to have our own custom extended attribute in the web, IDL, where we can just say like like I, don't know polyfill test or whatever feature detect, equals WebKit audio context and then probably a similar like plasma engine attribute for people writing bindings by hand yeah.

C

Something like that, where we just say like alternative: this is alternatively, name this and then you just blank it use that or their the depending on what the context is and.

D

Then we could, we could optimize the.

C

Binding a little bit too much of the if statements just yeah.

A

Okay, the downside is that now we have to maintain all of these and like find all of these, which kind of happy but I guess if we do a good job, that's nice for user I.

C

Think it's fine in this, since they're, like we I, don't think we should seek them out, but rather people open issues, and we just add another wait list as it comes up. Otherwise, it seems like everything should like in the expanse of time, reached what Firefox web video, which we used is actually saying and so like, eventually it even Safari to be renamed audio context where we won't even need this so I just like in that since I'm, not a too too worried about the cost. Long term of this kind of strategy.

A

Okay and then the final issue that came up I believe was a internet explorer doesn't have text to coder at all yeah and even when that's polyfill. That seems some of the Arden me. Some of the examples don't run sorry, not in Explorer MS edge.

E

Actually posted the issue in the agenda and I was actually wondering how much because I wasn't binding rely on that, because we use it for string I guess it needs to be here like yeah.

C

Basically, nothing works because yeah we use it for string both passing the Azam and pulling out for muslim and had we just unconditionally assume that if your new browser it is available or like if you're not unknown, is available. Yeah.

A

And the what's frustrating here is that this is not like, so this is the whole polyfill case right. It's not just like a different name for the same interface. We really don't want to include this code, just like polyfill code for non MSS browsers, so I think we have to rely on Mike, bundler configurations and like whatever, when, when people build this stuff like this is not something I included by default. Yeah.

E

I mean this is not a unique issue and the rest wasn't by the rest. Wasn't team has in every JavaScript project. You have these kind of issues and usually the bundler. We did with that sure you could just omit and import statements or a require statement and just leave it leave it like this.

A

That is what we do with comment. Yes with like browser code, it's on the global scope, so there's no require which makes it a little harder but I think at minimum I guess we need to document like how to do this or like pointers to how to do this. I guess yeah.

D

I remember that I've used the library context encoding, which you can use in a message and it works.

D

C

Is send, can you correct, or can you clarify, if do we need to do anything to hint to webpack that we are using text acuter and if you're targeting edge a polyfill is required or as like, as our current code generation already flying that.

E

The thing is, as Nick mentioned, since you are just coding out global object, the window won't notice any require statement or stuff like that, so you won't know anything about it, but usually what what you do is include a polyfill before that been there actually notes and conventionally so.

C

The idea is that, most of the time, you actually have like an import statement or require state like an import, yes, sick statement, and then that turns into a no op. If you end up you, don't just you don't actually target edge, yeah.

E

I mean I, don't know if you saw the there's something in the library community called bubble. Preset in the approach is, if you have already seen this now.

C

You know okay, so.

E

Basically, in the very configuration you can specify your target, which is either a browser like Safari, 7 or H, it events or faileth, and we actually scan your code searching for incompatibilities in your API that you're using and for example, if you stick to decoder it and it isn't defined on the targets you specify it will put in a polyfill and just polyfill your code for you and if you don't target edge, it would just omit it because it will assume that the target has it already and.

C

It specifically only works there with import like you to the scanning. Your code is like you have to have an import statement for you. I have to have like any a6 import to be a polyfill. No.

E

No, no yeah, you don't think that it will actually look at the usages like if I mean this will probably work. I! Guess because if wasn't binding, it's the JavaScript bender and then you pass it to blabber in a web pack. A bubble will actually scan the generated code for incompatibilities in the API and just polyfill them for you. So I guess, with a little more testing. I think this will work.

C

Yeah, so in that sense, I definitely think we should have caveats and instructions for like if you want support on all browsers, make sure you configure your bundlers to do this or like run this battle pass to do this, but the only other question for me would be we. We do somewhat support, like the new modules case, where you're not using a bundle or is like somewhat largest in the rest community and so I, don't know if we want to go out of our way to just say like this doesn't work an edge.

C

Please use a bundler. If you want that to work or do and I have a polyfill I feel like we should just say it doesn't work, but just have a kind of that saying, like new modules, doesn't work an edge yeah.

E

I mean if you're targeting nodejs and you have an incubator in competitivity in the api, it's kind of awkward to say just burned area code, because there's no need to burn a little, no js'. So I think it's. As you mentioned. It's a big, bigger issue than that.

C

So you think we should like have some sort of a feel for the new modules case.

D

No J's I, don't think you need, and you have to anyhow support it, because you have only one dog stop engine there I mean.

E

In that specific case, yes, but there are a few specific word platform API like window- that fetch that you don't have on the GS, and in that case, if you don't bundle your code, there is no. Where you have, you would have a polyfill in it. If.

D

I understand it correctly: we already doing that now right, so we have nothing specifically for no generation and for every browser to English. So I think we have that now already.

C

It is the case that we're not attempting to polyfill things for node renewed we're, just assuming you have some sort of base minimum version and like this at XD code or in some personal note, and if you use EPA, if you use whether it be guys like anything from what says. Basically, nothing will work and where we're not gonna, try and get that working on those. uh No js' specific.

C

D

A

For the no modules use case, I think what we can do is we can just give people a little snippet pointing to an existing that uses an existing polyfill and just does like you, put this script right above your main script, and it just says: if, like window text, encoder/decoder does not exist, then it creates its own little script, node that points to the polyfill and it inserts it in the document synchronously right at like during load time before the next script runs and, like you know, it's not great, but like it's good enough, you actually care about that.

A

Then, like you, know, you're well, on the way towards, like being in a custom world anyway, right yeah,.

E

Agree, and but this this should be documented in order to know what to party favorite, yeah and I. Guess if this I know, is users enough to actually do something about this, we could think about it more later. Yeah.

C

Okay, so document that document, if you're using a model or you have to worry about polyfills and the switches, do load transformation and, if you're not using a little over the modules. Talking about saying, like here's, a drop-in replacement like some existing something somewhere. So both of these documentation, so probably just say, like here's existing documentation for doing this, and this is like probably how you plug it in to us right. Exactly.

A

We don't want to be like the people, maintaining the. How do I polyfill text, you know decoder, encoder documentation. We want to be pointing to the people who are already documenting that right. Yes, all.

C

Right I, don't.

E

Know if you wanna note that but I will try to test the bubble thing into our existing website, complete.

C

A

Cool all right, Rasim, tacky, stuff I- can do this update or yeah I guess ketosis gone now, so we released Oh 500 of wesen packs. This is the one that has less impact, build and stuff I, don't know if you all saw that expand and CIN deal but can probably update.

A

The templates and everything it's a lot faster. It has built in headless browser testing. Now things are good, am I missing anything is it can also support targets. The way web engine says target node modules, targeted, commonjs whatever and then based on our discussions with Brian Charlson from tc39 and Microsoft's. We've got a pretty good plan, I think for how to generate like universal modules that work in both commonjs and ES modules and browser, and everything like that and I expect there'll be mmm some sort of straw.

A

Man coming up pretty soon, once Ashley's back from strange loop about that.

A

To talk Butler's and templates.

D

Today about that step in Ystad, yes,.

E

Now, just I think you know so: I've noticed that we have now a passel playing right, but I couldn't look into it. You're, probably the best to talk about you, okay,.

D

Sure my audio is clear right.

D

In plate and I moved in version, I have created two issues for you: one is for creating Billings VI other one is for anything depend about Raja, nothing, but but more than that, it's done so. People can use that thing that works. Fine and I also enabled CBI. For that and the second one is for us, a template. I've been able to see I, but I think something is some rough to later. Dependency is getting messed up with that, and that is causing I.

D

And also done this thing that yeoman generator for a spasm, so you can generate application.

D

Just pushed into a pool so that you can have a look into it and then probably work with that stuff. That.

A

Again in the agenda.

D

A

E

Yes, I mean we.

A

Finished the goal updates.

D

Actually, I will get back to your name. I just had.

D

E

You not talking about verdict by any chance.

D

Not sure I think in role of do you, they still have the PR open. So.

E

Yeah the complicate the thought in word is that they not designed to hand their binary mothers into the graph, so it was pretty yeah. Basically, I. Don't have the time to finish that otherwise, okay.

D

A

If you don't have time to finish, the PR I think it might be helpful to like. Maybe leave a comment in the PR saying that so that they know that they can take it if they need to yeah.

E

I know I, so I was clear with them and they probably know that and but but the thing is that just nobody wants to work on that at the moment, because you basically need to rewrite a lot of pieces mm.

E

Okay and so Cynthia I'm wondering if you know the author of the past Harrigan.

D

No I don't know the lateral parcel clicking. Okay.

E

Because I guess it would be cool if you can join the core at some point. I.

D

Think it makes sense, we can talk.

E

Yeah, it seems that the parcel guys have just no interest of making sure that wherever somebody works in Gaza yeah.

D

They're really interested in that stuff and I think like a lot of workers. Work is going on in that direction. So let's do it, let's talk with them, probably beginning and then so.

E

You are saying that you are already working with them right. I talked.

D

With one person who I think his name is SK, diem name is Catholic MA and that guy was creating a plugin for so just offer them I.

F

Sigma actually part.

A

Of the parcel team, I kind of had the impression that Kat Sigma just was an individual contributor yeah.

D

I think he's an individual contributor parameter, I.

E

Mean acidified my positing.

D

We'll talk with them, we can I.

E

Still have my open issue about where were somebody's been in work and interest research, and they don't seem to response to that.

D

A

All right so moving on to documentation, I, don't think. There's too much here, other than I updated the tutorial for the new as impact release. So basically it's just saying use Westpac, build and said it wasn't packing it now and there's still to do is to make a testing chapter now that we have smooth user experience for testing.

A

See website to bugging everything else, I, don't think, there's really any updates, as someone else has updates, but.

D

Nick I guess for the website I think there were some messages going on for Stage two. Was that already done or do we need something to do for that.

A

I I think, like content, wise, we're pretty much there and basically we need to talk with Aaron, I think, probably about final bits of Polish, but I. Don't think, there's anything really pressing that we need to do right now. Okay,.

D

A

Sounds good to me all right: shall we talk about web worker support? Yes,.

D

E

So I first saw an issue: I don't remember where exactly, but someone was saying that wasn't been Jane wasn't able to do beautiful service working. Oh, he wasn't able to load the wasn't module into it, service, worker and so web worker. Yes, I'm.

E

Sorry it simply is a web part plugin or it's a web back loader actually, but it's kinda in 9u wasn't module into the bender, which is kinda, which is pretty bad actually, so people cannot use this, but they can use something else and the whole point of that is that I posted an example into the nodes where someone made a clear configuration for is some code about how to load a web module into the serviceworker and I think it will make sense to either make a new example into the wasn't engine repo about how to build a serviceworker, how to pass messages in this kind of stuff?

E

And maybe why not create a new template or add it into an existing template to show how to load them was a module into a serviceworker. So.

A

What is the reason why our default kind of setup won't run in a web service worker like what is the missing API? Did we rely on.

E

D

E

D

I think the problem is basically on the back plug in their back module or sorry, the back loader are they using and that kind of inline binary content of the box and file into the JavaScript file itself. So we need to do some work around if you are using a pack and I.

E

Found the initial issue again and yeah yeah. Actually it's more like web touch, it sure actually.

A

A

We should fix that then yeah cuz I'm, like one of the the kind of properties that we've had so far. Is that generally, like other than these things, like the text, encoder decoder bit is like, if you don't use, say a Dom API in some respects on project, then you don't rely on the Dom and your code should be able to run in environments that don't have access to the Dom like a worker or service worker, or something like that so it'd, be nice to you know continue to make that a reality. Yep.

D

I think one other thing that began to depart and played for a fool is that we can say people like ok for the software. You have this problem, which you can do it. You can have this workaround to make it work and when you in put the example over there, where a sample over there and then we point that thing to people can use that workaround fill that butter is getting fixed in very fact, site. Yeah.

E

D

E

As far as understand the issue here is that there's there isn't actually an issue. Is it kinda works?

E

If you know how to do it, it's pretty hard to know this stuff, so I think we should just making sure that the documentation is good and that we have examples, but then so I'm just wondering if you move, if you would make sense to create a new worker or service worker or web worker target into wasn't binding, because you could have changes in the API like if you want to include a new like basically the important you require statements into a worker, just coding a function called import scripts or something like this.

E

So basically the API is different, so it kind of makes sense to create when you target to that.

A

Because is import script synchronous because that's like safe in a worker right, yes,.

A

Okay, cool yeah I do agree. That would be nice to have an example of just doing worker stuff and like post messaging back and forth in the West budging cut I.

A

Think we're pretty close to running out of time on the zoom meeting. Correct, oh okay, that's so weird that they don't give the person running the meeting the timer whatever, but it seems like we've covered everything here.

A

Does anyone have any last-minute things to talk about.

B

C

After a and touch.

B

Base for you sensitive, that's about it, there's also like a Watson package.

A

I'm I honestly didn't really catch any of that other than follow-up and something about allies impact issue. But that sounds good to me. We can definitely communicate online.

E

A

Yeah, okay, so thanks everyone for coming to the meeting and participating and I wish you all a wonderful week and see you next time. Yep see you next time.