Re: [rtcweb] "20 lines" (Re: RTCWeb default signaling protocol [was RE: About defining a signaling protocol for WebRTC (or not)])

[Iñaki Baz Castillo <[e-mail]>]

Hi Ravindran, comments inline:


2011/9/26 Ravindran Parthasarathi <[e-mail]>:
> The aim of having standard RTCWeb signaling protocol is to make basic dialog between two different browsers (For ex: Mozilla & chrome) user.

And me and others already said that the aim of WebRTC is not
converting web browsers into generic SIP phones capable of
establishing a call between them without visiting the same page
(replace "SIP" with whichever protocol).

WebRTC adds value to the WWW world, so it's assumed that first of all,
browsers visit some website, so they can get something (JavaScript
code).



> Having seeing the list discussion on signaling intricacies in this alias, it is no point in asking each web developer to create their own signaling protocol. It is better for IETF to recommend one protocol which best suit RTCWeb signaling requirement and best simple API for signaling protocol will be defined by W3C.

Ok, but such protocol should be carried within HTTP or WebSocket,
rather than being a native SIP/XMPP/whatever-custom-protocol built-in
the web browsers. Do we agree on this? I think the answer is "not", so
please continue reading.



>>Ravindran, you insist of that even when multiple comments tell you the
>>opposite. If Mozilla, Chrome or IE browsers are visiting the *same*
>>website then they *already* share the *same* JavaScript (loaded from
>>the website) so they can signal the media thought the web server (by
>>using HTTP or WebSocket) as defined in the retrieved JavaScript code.
>>This is easy to understand IMHO and this is how WWW works for years =>
>>interoperability between clients visiting the same website. Period.
>>
> <partha> WWW works with plugin. If you think otherwise, there is no much
> Work for IETF in this WG </partha>

AFAIK one of the aims of WebRTC is allowing multimedia sessions within
browsers *without* the need of external plugins (like Flash or Java
applets).




>>If the browsers speak *native* SIP (as you suggest in many threads)
> <partha> I asked initially because it is defacto recommended IETF
> real-time protocol </partha>

I'm sure that fans of XMPP would not like that (even less taking into
account that XMPP is much more deployed in pure Internet than SIP).



>>then the web server is unaware of what is happening at media level
>>between the visitors (because the web server, where the application
>>logic is, does not see the signaling traffic).

> <partha> IMO, it is your bad assumption. There are "n" number of way to
> Design server even if they speak different protocol which is a common
> Practice in gateway implementation </partha>

So you want a "built-in protocol" not to bore web developers with
custom and complex signaling protocols, but at the same time you want
to force them to build a hyper-complex gateway in server side. ¿?¿?



>>> Here, there is a need of defining the common underlying protocol
>>(which is not RTCWeb developer concern).
>>
>>No, it's not.
> <partha> There is a strong need to define signaling protocol which helps
> Browsers to interop and also reduce the complexity for web developers </partha>

Again: browsers MUST not *directly* interop with other browsers in the
*signaling* plane. If two browsers visit the same page and download
the *same* JavaScript, such JS already can say the browsers how to
signal the media.

Please, don't insist on the same again and again. There is NO need at
all for web browsers to natively and directly speak
SIP/XMPP/other-protocol between them. In fact, that is undesirable.



>  You can repeat 1000 times by ignoring any rationale
>>already replied to you in multiple threads. But that changes nothing.
>>WebRTC does not need native SIP or native XMPP/Jingle spoken by web
>>browsers. In fact, that would stop innovation, would mandate website
>
> <partha> I have clarified multiple time that your innovation is not going
> to stop by defining standard signaling protocol. Let say X protocol become
> standard signaling protocol which does not mean that you should not
> implement your SIP over websocket using Javascript. So, this is false
> propaganda by few folks that it would stop innovation. </partha>

Ok. But in other mails you propose that WebRTC should be an island for
web browsers. Wouldn't that stop innovation?



>>admins to provide a SIP/XMPP server within their infrastructure (how!)
>>and would make WebRTC unfeasible in shared web hostings.

> <partha> In case you talking about manageability of the servers, lot of more
> Stuff involved than simple protocol decision </partha>

No. What I meant is very clear: if the signaling (or the "default
signaling" as you propose) is not carried via HTTP or WebSocket, then
it would require another kind of centralized server (as a SIP proxy in
the provider side). So all those websites hosted in shared hosting
datacenters (i.e. using a shared Apache with different virtual
domains) would have no chance to use WebRTC. Just it.



Regards.


-- 
Iñaki Baz Castillo
<[e-mail]>