W3C home > Mailing lists > Public > public-webrtc@w3.org > July 2013

Re: VS: Teleco Integrators vs Web Developers vs Browser Implementers

From: Eric Rescorla <ekr@rtfm.com>
Date: Fri, 5 Jul 2013 14:53:35 -0700
Message-ID: <CABcZeBP5Myum3h+FHR4H2-TT6p3yBxLpcJjTO16VnDzn7fQEsw@mail.gmail.com>
To: "piranna@gmail.com" <piranna@gmail.com>
Cc: Martin Steinmann <martin@ezuce.com>, Robin Raymond <robin@hookflash.com>, cowwoc <cowwoc@bbs.darktech.org>, "public-webrtc_w3.org" <public-webrtc@w3.org>, Roman Shpount <roman@telurix.com>, Christer Holmberg <christer.holmberg@ericsson.com>, Adam Bergkvist <adam.bergkvist@ericsson.com>, IƱaki Baz Castillo <ibc@aliax.net>, Ted Hardie <ted.ietf@gmail.com>, Parthasarathi R <partha@parthasarathi.co.in>, Martin Thomson <martin.thomson@gmail.com>
On Fri, Jul 5, 2013 at 2:10 PM, piranna@gmail.com <piranna@gmail.com> wrote:

> I admit I'm not a networks guy, but:
>
>
> > 0. A full list of ICE candidates:
>
> Also for DataChannels? I don't need to negotiate anything, is just a
> request about reliable or not...


Huh?

DataChannels and SRTP go over the same type of ICE-negotiated
UDP transport, and that means you need ICE.


> >   - Not just the STUN addresses, since you might have two machines  on
> the
> > same network.
>
> Ok, this is good. Also having the full routers-chain would allow to
> connect them directly without going out to Internet, just connect
> directly over the local LAN... :-)


I have no idea what a "full routers-chain" is, but for ICE, what you
need is your host candidates, server reflexive candidates, and
relayed candidates. If a LAN-only connection is possible, ICE is
designed to find it.


> >   - The priorities for the ICE candidates
> > 1. The ICE credentials (ufrag and password)
> > 2. The DTLS fingerprints
>
> Really is needed all this thing just for DataChannels?


Yes. ICE needs the ufrag and password and the data is carried over DTLS.


> Could it be
> hidden someway?


Hidden where? It's hidden in the SDP now. You're asking to break stuff out
of the SDP, and that means it needs to be carried somewhere else.


> Regardless of the exact design of the protocol (SDP or whatever), it seems
> > like it's
> > pretty desirable if implementors need to know these facts in order to do
> > anything
> > useful.
> >
> I'm talking about a high level developer point of view,


So am I.



> I'm only
> interested on be able to make the connection and transfer data, no
> more,


Yes, and this is the information which needs to be carried between the
endpoints to do that. Either it can be carried in one big chunk or it can
be broken up into little pieces.



and I believe that for doing this a simpler API would be capable
> of doing it...


in this case, what's making the API complicated is that you're trying
to break all this information out of the SDP. If all you want to do is
DataChannels, the API is quite high-level and straightforward, and
trying to do a lower-level API forces the user to know quite a bit
more about networking, as this example shows nicely.

-Ekr
Received on Friday, 5 July 2013 21:54:42 UTC

This archive was generated by hypermail 2.3.1 : Monday, 23 October 2017 15:19:34 UTC