[RTW] [dispatch] Does RTC-WEB need to pick a signaling protocol?
Harald Alvestrand
harald at alvestrand.no
Mon Jan 31 15:28:13 CET 2011
On 01/30/11 22:50, Bernard Aboba wrote:
>
> One approach is to tunnel the signaling protocol over HTTP. This
> requires support for bi-directional communications, such as can be
> provided by BOSH, or Websockets. In these architectures the web
> client communicates via HTTP with a "Connection Manager" which
> encapsulates/decapsulates the signaling messages and routes them
> appropriately on the backend.
>
I think that no matter what we define (or not) as the signalling
protocol, it will be carried in a fashion that looks like HTTP or HTTPS
(most probably HTTPS) to intermediaries and firewalls.
Not using WebSockets (if it's done yet) seems like a strange choice; if
WebSockets are usable for anything, they should be usable for this
application.
>
> To provide interoperability, specifications for encapsulation of
> signaling messages within HTTP are required. Examples include:
>
> XEP-0124 <http://xmpp.org/extensions/xep-0124.html>:
> Bidirectional-streams Over Synchronous HTTP (BOSH)
>
> XEP-0206 <http://xmpp.org/extensions/xep-0206.html>: XMPP over BOSH
>
> An XMPP Sub-protocol for Websocket
> <http://tools.ietf.org/html/draft-moffitt-xmpp-over-websocket>
>
> Information on BOSH implementations can be found here:
>
> http://xmpp.org/about-xmpp/technology-overview/bosh/#impl
>
> *From:*rtc-web-bounces at alvestrand.no
> [mailto:rtc-web-bounces at alvestrand.no] *On Behalf Of
> *Markus.Isomaki at nokia.com
> *Sent:* Sunday, January 30, 2011 2:45 PM
> *To:* bernard_aboba at hotmail.com; richard at shockey.us;
> erik at hookflash.com; jonathan.rosenberg at skype.net
> *Cc:* rtc-web at alvestrand.no; dispatch at ietf.org
> *Subject:* Re: [RTW] [dispatch] Does RTC-WEB need to pick a signaling
> protocol?
>
> Hi,
>
> I pretty much agree with what Bernard and Johathan Rosenberg have said.
>
> There are just a couple of issues I would like to understand better:
>
> -I suppose these Javascript libraries can be cached on the browser and
> need not be downloaded every time the application/site is accessed?
> (These are probably a common practices in Javascript and Browsers but
> are good to clarify in this discussion.)
>
> -It may be easy for a web developer to use the client side Javascript
> library, but I believe the challenge may be bigger on the server side,
> where scalability etc. become issues. How about the server side, is
> there something ready-made for that? I suppose the Javascript on the
> browser can't necessarily connect to vanilla SIP or XMPP servers but
> some HTTP/WebSocket/TLS tunneling and connection management magic is
> needed, especially when dealing with HTTP intermediaries. (But
> presumably the same issues would be faced if we "picked" SIP or XMPP.)
>
> Markus
>
> *From:*rtc-web-bounces at alvestrand.no
> [mailto:rtc-web-bounces at alvestrand.no] *On Behalf Of *ext Bernard Aboba
> *Sent:* 30 January, 2011 07:35
> *To:* Richard Shockey; erik at hookflash.com; jonathan.rosenberg at skype.net
> *Cc:* rtc-web at alvestrand.no; dispatch at ietf.org
> *Subject:* Re: [RTW] [dispatch] Does RTC-WEB need to pick a signaling
> protocol?
>
> > Duh ... what are we abandoning 10 years of work?
>
> The web is a "generative" platform that supports not just protocols,
> but also execution of code. This enables the platform to be extended
> in a virtually infinite number of ways, including the development of
> Javascript signaling APIs, without needing to add yet more core code
> to the browser. This is the preferred approach unless it can be shown
> that something *absolutely* must be natively supported (as was
> discussed at the workshop, STUN/TURN authorization for peer-to-peer
> media is probably an example of something that *does* need to be native).
>
> As an example of what is possible, there is an excellent Javascript
> library for XMPP (strophe, see http://code.stanziq.com/strophe/).
> Poking around, it would appear that there are a number of Javascript
> libraries that claim to provide support for SIP (such as
> http://phono.com/), although I have no idea of their usefulness.
>
> Given the generality and power of the web platform and the ease with
> which sophisticated signaling libraries can be implemented today, the
> bar for getting additional code into any browser is going to be quite
> high. If you doubt this, ask the build-master of your favorite
> browser for the requirements for checkin of a complete SIP stack :)
>
>
> _______________________________________________
> RTC-Web mailing list
> RTC-Web at alvestrand.no
> http://www.alvestrand.no/mailman/listinfo/rtc-web
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.alvestrand.no/pipermail/rtc-web/attachments/20110131/acc82d6d/attachment.html>
More information about the RTC-Web
mailing list