Re: [MMUSIC] Comments on ICE-TCP specification RFC 6544

[Nigel Pattinson <[e-mail]>]

jakubadamek wrote
> could you please tell me roughly how much time it took to implement the
> RFC? 

That's difficult to say - we started from scratch, so had to implement STUN,
TURN (client and server), as well as Jingle. Probably the ICE portion was 2
- 3 man months.


jakubadamek wrote
> And how well does it work for you?

We haven't shipped yet, but it seems like it will work fine. However in
practice we suspect the outcome will be that host candidates will be used
within a LAN, and a relayed connection in almost all other cases. We have
all the usual ICE code for server and peer reflexive candidates, and for
simultaneous open TCP candidates, but we expect that the cases where they
can be successfully used to establish a connection will be rare.

Also as my previous post suggested, your TCP code may need to be more
different from UDP than the specification might lead you to believe.

In the end, if a relayed connection is an acceptable outcome and
peer-to-peer is a bonus, then ICE-TCP seems worth investigating. But if
peer-to-peer is the priority UDP might be the way to go.

Nigel



--
View this message in context: http://ietf.10.n7.nabble.com/Comments-on-ICE-TCP-specification-RFC-6544-tp191890p374094.html
Sent from the IETF - mmusic mailing list archive at Nabble.com.