[webkit-dev] New feature announcement – Video conferencing and peer-to-peer communication
abarth at webkit.org
Mon May 16 13:30:27 PDT 2011
On Mon, May 16, 2011 at 6:54 AM, Tommy Widenflycht <tommyw at google.com> wrote:
> The plan is to have the P2P VC WebKit code to do as little as possible
> and leave the implementation to the browser port, just like the
> MediaStream patches. And the code touches very little other code;
> Event, EventTarget, Navigator etc but obviously heavily depends on the
> MediaStream functionality.
There's a trade-off here between adding complexity to WebCore and
sharing the functionality between different consumers of WebKit.
> Is this a problem? Happy to modularize the functionality if someone
> can give me some pointers, show our work in progress etc.
I started a separate thread about modularity. That's somewhat of a
larger discussion, and certainly doesn't need to block your work.
It's unclear whether that's the direction the project will go in any
case. In the near-term, I'd recommend putting this feature in a new
directory and being careful about dependencies to and from your code.
> To get the standard (on whatwg) finalized an implementation is needed
> so that some real world usage can be had, and previously we have
> received some favorable comments re go-ahead on implementation.
> Personally I think this is a great feature and are committed to
> implement it properly.
The high-level question is whether this feature is beneficial to the
WebKit project. I've seen folks raise the following concerns:
1) Eric Seidel raised the question of whether this feature is
appropriate for WebCore/WebKit, presumably because P2P video
conferencing seems like something you'd build on top of the web
platform (like an email service) rather than something that's directly
related to rendering web pages.
2) Eric Carlson (in Bug 56459) says that the feature is important for
WebKit but is concerned that the spec is a moving target and that
implementing it at this stage might be costly for the project in the
log run. After some discussion, he seems to agree that we should
proceed with an off-by-default, vendor-prefixed implementation so we
can provide feedback about our implementation experience to the
My perspective is that I would like to see WebKit continue to be a
leading implementation of the web platform. Leading is more expensive
than following, of course, because followers can adopt a wait-and-see
approach and only implement features that are already successful
rather than placing bets that sometimes pay off and sometimes don't.
>From that perspective, the question boils down to whether this feature
is a good bet for the platform.
> On Sat, May 14, 2011 at 01:50, Adam Barth <abarth at webkit.org> wrote:
>> One approach we talked about at the contributor's meeting to dealing
>> with the ever-growing feature list of WebCore is modularity. Looking
>> at the specification for this feature, there doesn't seem to be much
>> interdependence between these APIs and the rest of WebCore. For
>> example, there's nothing that couples with dom, css, or rendering.
>> Would it make sense to try building this feature as a self-contained
>> module? For example, you could imagine the following dependency
>> WebKit -----> WebP2P
>> | |
>> | \/
>> +------------> WebCore
>> Ideally, WebCore would have some moderately well-defined interface for
>> these modules, much like the WebKit <--> WebCore interaction is
>> mediated by clients. Ideally, these modules would just describe
>> themselves using WebIDL and we could just generated bindings for them.
>> As another example, webaudio might also profit from such as design.
>> Of course, splitting this feature off into its own module doesn't
>> remove all the costs of the new feature, but it would make it easier
>> to work on WebCore proper without having to deal with all the
>> complexities of every feature.
>> On Fri, May 13, 2011 at 11:10 AM, Eric Seidel <eric at webkit.org> wrote:
>>> That sounds like about 8 features.
>>> Seems we should think about this in smaller chunks...
>>> On the surface "peer to peer video conferencing" does not seem like
>>> something appropriate to add to WebCore/WebKit. Just like "an API for
>>> reading my email" is out of scope for the project. (But certainly lots of
>>> people build such things on top of WebKit.) :)
>>> On Fri, May 13, 2011 at 6:19 AM, Tony Gentilcore <tonyg at chromium.org> wrote:
>>>> For those interested, the cover bug for this work is here:
>>>> On Fri, May 13, 2011 at 8:27 AM, Adam Bergkvist
>>>> <adam.bergkvist at ericsson.com> wrote:
>>>>> I want to inform people on this list that we have been doing some early
>>>>> implementation work of the video conferencing and peer-to-peer communication
>>>>> chapter in the HTML spec
>>>>> This section of the spec is in an early stage and should be considered
>>>>> experimental, but as it becomes more understood, we intend to contribute to
>>>>> the WebKit implementation. Others are also (based on bugs reported) working
>>>>> on this.
More information about the webkit-dev