This should be of interest for INT area folks as well:

Begin forwarded message:

> From: Lars Eggert <[EMAIL PROTECTED]>
> Date: May 19, 2008 12:03:53 GMT+03:00
> To: TSV Area <[EMAIL PROTECTED]>
> Subject: non-SIP usage of ICE
>
> Hi,
>
> the MMUSIC WG is discussing whether they should take on a work item  
> on using ICE in non-SIP/RTP environments. This subset of ICE  
> specified in draft-rosenberg-mmusic-ice-nonsip (called NICE) would  
> be a general-purpose NAT traversal mechanism, and is for example  
> being considered by the HIP WG as a NAT traversal solution for HIP.
>
> I strongly encourage transport-area folks to join this discussion on  
> the MMUSIC mailing list. The proponents of ICE argue to develop it  
> into the IETF-preferred NAT traversal protocol, for which it IMO is  
> currently not sufficiently developed.
>
> There are some challenges here that are of immediate interest to  
> transport folks. One is that ICE-for-SIP had been pacing its  
> connectivity checks such that they use bandwidth in way that is not  
> more aggressive than a following RTP media stream. With NICE, this  
> approach is no longer possible, because NICE has no knowledge of the  
> transmission behavior of arbitrary flows.
>
> Another is that if NICE chooses candidates that it has obtained via  
> STUN and TURN, non-RTP traffic may be tunneled through such relays,  
> which is currently (at least to me) not a well-understood operation.
>
> Because ICE handles gathering of candidate IP addresses and ports,  
> it is undefined how ICE/NICE would intersect with layer-3 mobility  
> and multihoming solutions, or transport protocols like SCTP.
>
> Finally, ICE has so far mostly focused on relaying UDP traffic, with  
> a draft on relaying TCP still under WG discussion (draft-ietf-mmusic- 
> ice-tcp), which so far is still SIP/RTP-specific. There is currently  
> no possibility to relay SCTP or DCCP with ICE.
>
> Lars



_______________________________________________
Int-area mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/int-area

Reply via email to