Re: Audio Workers - please review

On Wed, Sep 10, 2014 at 9:29 AM, Ehsan Akhgari <ehsan@mozilla.com> wrote:

> Hi Chris, and everyone else on the WG!
>
> I took a look at the merged version of this proposal and have some
> feedback, I would appreciate if you can please let me know what you think.
> As you know, I haven't been closely involved with the spec for a while, so
> if the below reflects my ignorance on some of the previously discussed
> matters, please accept my apologies in advance.  And of course, sorry that
> this is a *long* list.  :-)
>
> 1.  I think the processing model can use some further clarification.  The
> current text of the spec is very vague on a number of different issues, so
> I would appreciate if you could please clarify what you intended to happen
> in the following cases:
>
> a) It is not really clear what should happen after you call terminate() on
> AudioWorkerNode, since that seems to be orthogonal to whether the node
> participates in the graph.
>
> b) The interaction between worker closing/termination and the audio
> processing is not clear at all.  Specifically, dedicated workers typically
> have a long time that is as long as the last pending task (assuming that
> the main thread doesn't hold a reference to them), however, for these
> workers, we need to keep firing audioprocess events.
>
> c) How would different AudioWorkerNodes allocated through different
> AudioContext objects interact with each other?  (Here is an example of
> something that is not clear, should it be allowed to send an AudioParam
> belonging to a node form context 1 to a worker node from context 2?  The
> spec is currently silent about this.)
>

Agree the spec is silent and that it should be clarified, but it seems the
clarification is straightforward: if we took the extensibility perspective
and assumed that the worker type of node is what explains how all types of
nodes operate "under the covers", then we'd need to imagine the same rules
applying as for other node types. Needs spec language for sure, but the
behaviour shouldn't differ.


> d) What is the exact semantics of the audioprocess events dispatched to
> these workers?  Do they block running all audio processing on the audio
> processing thread?  Note that because of the halting problem, we cannot
> even verify that the worker script will ever terminate,
>

It can be pre-empted. We do this for long-running UI thread tasks (often
presented to the user as "this script is taking a long time, do you wish to
stop it?").


> let alone in a reasonable amount of time, so in practice the UAs need to
> preempt the execution of the script.  It would be nice if this was somehow
> mentioned in the spec, at least in terms of what the UA needs to do if it
> decides to abort the execution of one of these scripts.
>

Does pre-emption have an effect on the processing model that's different to
a worker that simply doesn't return data? That's the case in which I'd
imagine you want spec text.


> e) What is the order in which these worker nodes receive the audioprocess
> events (assuming obviously that one such node is not an indirect
> input/output of the other)?  Note that with the currently specified API, I
> think it is possible for one of these workers to send a MessagePort through
> the main thread to another one, and therefore be able to communicate with
> the other AudioWorkerNode workers through MessagePort.postMessage, so the
> order of execution is observable to script.  (I admit that I don't have a
> good solution for this -- but I also don't know what use cases for
> transferring information back and forth between these nodes and the main
> thread we're trying to address here.)
>
> 2. I feel very strongly against exposing this node type as a web worker.
> I think that has a number of undesired side effects.  Here are some
> examples:
>
> a) Even though the HTML web worker termination algorithm says nothing
> about killing an underlying worker thread, I think that is what authors
> would typically expect to happen, but obviously doing that would not be an
> option here.
>

HTML Web Workers have no impact on the threading model outside observable
side-effects; e.g., it's entirely possible to fuse worker execution to the
UI thread and pump from a single message loop. Nothing prevents it (except
that it sucks). Some UAs have even done this. Futher, some UAs decline to
spin up a thread per worker, meaning some workers can starve others. It's
unclear why users should have a specific mental model for this already when
there's so little prior use by users and/or consistency among impls.

It is also not specified what needs to be output from the node after
> terminate() has been called on it (the input unmodified? silence? something
> else?)  Also, given the fact that you can already disconnect the node from
> the graph, why do we need the terminate() method in the first place?
>

The discussion I had with Chris about terminate() revolved around it being
grandfathered in from the Worker spec. Removing it is a pain. Better to
spec it to do something sane -- this group should decide what's most sane.


> b) The API gives you the illusion that multiple AudioWorkerNode's will run
> on different DedicatedWorkerScopes.  That, in Web Workers world, would mean
> different threads, but that will not be the case here.  I think that
> discrepancy is problematic.
>

This mostly means you're bringing a specific impl strategy to a world that
doesn't admit only your worldview. This isn't the fault of the specs.


> c) Using DedicatedWorkerGlobalScope is a bit weird in terms of APIs that
> are exposed to workers.  For example, should these workers support
> onlanguagechange?  What about IndexedDB on workers?  What about nested Web
> Workers?
>

We'll need reasons to NOT expose them, not open-ended questions about if
one should.


> d) At least on Gecko, Web Workers have specific implementation concerns in
> terms of their message queue, event processing model and so on.  It might
> be a lot of effort for us to properly implement the observable behavior of
> these workers running inside our audio processing thread code (which has a
> completely different event processing model, etc.)
>

I expect that every impl will have some work to do. Is the suggestion that
the size of the work is enough to cause you to raise a formal objection? If
not, does the spec look like it can't be implemented?


> e) The topic of whether or not synchronous APIs must be allowed on workers
> is being debated on public-script-coord, and it seems like there is no
> consensus on that yet.  But I find the possibility of running synchronous
> XHR on the audio processing thread unacceptable for example, given its
> realtime requirements.
>

You should remove XHR and, if anything, only surface fetch. Actually, it
was my recollection taht XHR isn't in the base interface for workers. Is
that wrong?:
http://www.whatwg.org/specs/web-apps/current-work/multipage/workers.html#apis-available-to-workers

The only synchronous API I see is importScripts().


> I think tying this node to Web Workers opens a huge can of worms.  It will
> also keep biting us in the future as more and more APIs are exposed to
> workers.  I am wondering if we can get away with a completely different API
> that only tries to facilitate the things that authors would typically need
> to do during audio processing without attempting to make that a Web Worker?
>
> I think at the very least, if we really want to keep tying these concepts
> to Web Workers, it might be worthwhile to bring that up on
> public-script-coord, since it is at least bending the original use cases
> that Web Workers were designed for.  :-)
>
> 3. What is the purpose of addParameter and removeParameter?  It seems to
> me that if we defined a structured clone algorithm for AudioParam (which is
> another detail missing from the spec that needs to be clarified anyway) and
> make it transferable, then the author would be able to postMessage() an
> AudioParam just as easily.  Is there a good reason to have a specialized
> method for what is effectively posting an AudioParam to the worker?  (Note
> that we'd probably need to add methods to AudioParam for extracting a-rate
> and k-rate values at any given time in that case, so that the worker script
> can get the right values when it needs them.)
>
> 4. More on the idea of transferring an AudioParam to a worker, that will
> probably involve some kind of neutering the object.  It might make sense to
> introduce a clone() method on AudioParam as a way for authors to be able to
> keep a copy of the object around on the main thread.  That could of course
> be a future enhancement idea.
>
> 5. Still more on the idea of transferring an AudioParam, another thing
> that we need to worry about is what happens if you try to transfer an
> AudioParam that is currently being used somehow (either through being a
> property on another AudioNode, or being used as an input to one.)
>
> 6. I think addressing #3 above will allow us to completely eliminate
> AudioProcessEvent, and just use AudioProcessingEvent.
>
> I hope the above is useful.  I think there is much more to think about
> especially given what we end up deciding on the first two items above, so I
> hope to provide further feedback as we make progress here.
>
> Please let me know what you think!
>
> Cheers,
> Ehsan
>
>
> On Mon, Aug 25, 2014 at 11:29 AM, Chris Wilson <cwilso@google.com> wrote:
>
>> I've done some tweaking to the Audio Worker (issue #113
>> <https://github.com/WebAudio/web-audio-api/issues/113>) proposal, and
>> most significantly added the ability to create AudioParams on Audio Workers
>> (issue #134 <https://github.com/WebAudio/web-audio-api/issues/134>).
>>
>> The fork is hosted on my fork (http://cwilso.github.io/web-audio-api/).
>>  Start here
>> <http://cwilso.github.io/web-audio-api/#widl-AudioContext-createAudioWorker-AudioWorkerNode-DOMString-scriptURL-unsigned-long-numberOfInputChannels-unsigned-long-numberOfOutputChannels>
>> to review the creation method, and the bulk of the text begins at
>> http://cwilso.github.io/web-audio-api/#the-audio-worker.
>>
>