[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [MMUSIC] Comments on draft-rosenberg-mmusic-ice-nonsip



On Sun, 9-Mar-08, at 09:26 , Jonathan Rosenberg wrote:

> I think the real meta-question on this draft, is whether:
>
>   a. it is the same functionality as ICE, but not tied to SIP
>   b. it is (a) plus a subsetting, removing features that we think  
> only SIP needs
>
>
> The list of things in the (b) bucket that I removed were:
>
>   1. multiple streams
>   2. multiple components per stream
>   3. updating ICE parameters mid-negotiation
>   4. restarting ICE
>   5. role conflict detection
>   6. ICE mismatches
>   7. default candidates
>   8. ice-lite
>   9. forking support
>   10. frozen algorithm
>   11. remote candidates mechanism
>
>
> So first we need to decide path a or b, and then we can discuss  
> which of the above get included or not.


Right now, I vote for (a) for the following reasons:
* I think we want to encourage just one "ICE" protocol, not an ICE  
and a NICE protocol. This is because I think we will find things we  
want to change/fix with ICE and having two protocols will make things  
harder.
* I think that selecting the list in the (b) bucket will be hard, at  
least until we get more experience.



>
> few more comments below:
>
> Philip Matthews wrote:
>> Below are some comments on this draft. I have grouped the comments  
>> by subject matter.
>> Comments on the subset of ICE chosen:
>> * I think this draft should allow multiple streams between the two  
>> clients. I feel this is useful in non-SIP contexts. For example,  
>> my draft draft-matthews-p2psip-id-loc requires multiple streams  
>> for non-SIP purposes. I suggest this draft simply point out the  
>> simplifications that occur if there is only one stream.
>> Comments on the MIME object:
>> * I think this draft should not require the MIME object of section  
>> 15, but simply present it as one possible encoding. The draft  
>> should instead list the parameters that ICE needs as input, and  
>> allow protocols to use other encodings instead. Protocols may also  
>> derive some of these ICE-related parameters from other fields in  
>> their protocol messages, rather than pass them explicitly.
>> For example, right now the HIP work is proposing to use an  
>> encoding that is more consistent with HIP. HIP is a binary  
>> protocol, so a text-based encoding seems a bit out-of-place.  
>> Moreover, the HIP work is proposing to not generate an ICE- 
>> specific username fragment and password, but use the Host Identity  
>> Tag (HIT) as the username fragment, and the keying material  
>> derived from the Diffie-Helmman exchange as the password. Thus  
>> these fields of the MIME object are not needed for HIP.
>
> OK. I got a few folks that said they hated this MIME thing. THe  
> intention was that it was optional - there if you want it. I think  
> it would be a fine idea to have a section that also lists the  
> params you'd need to include if you didn't use the MIME object.

Sounds like we agree.

>
>> Comments on Terminology:
>> * I dislike the term "session protocol". In SIP/SDP, it is the  
>> direct connection which is the "session", not the signaling  
>> protocol, so the "session protocol" should be the protocol that  
>> runs over the direct connection. I suggest that you replace the  
>> sentence "NICE works only with protocols that fit the pattern of a  
>> session protocol" with "NICE only works with signaling protocols  
>> that have a rendezvous mechanism".
>
> OK.
>
>> * In section 2, the document talks about establishing sessions  
>> between "hosts". I think it should instead talk about sessions  
>> between clients or protocol instances or similar.
>> * The document introduces the terms "Initiate Message/Accept  
>> Message" and "Initiator/Responder" in place of ICE's "Offer/ 
>> Answer" and "Offerer/Answerer". Though the new terms might be  
>> better than the existing terms, I think it will just be confusing  
>> to have to different sets of terms for the same concept. Anyone  
>> who implements NICE and any protocol designers who use NICE will  
>> have to at least skim draft-ietf-mmusic-ice and become familiar  
>> with the Offer/Answer terminology there, so I don't see the point  
>> of introducing new terminology.
>
> Hmm, well I thought it was confusing to use the same terms when  
> this is not based on O/A anymore.

Personally, I have not found it confusing. For the NAT traversal for  
HIP work, I think in terms of an "ICE Offer" and an "ICE Answer" and  
of two endpoints (the Offerer and Answerer) even though it has  
nothing to do with SDP.

>
> -Jonathan R.
> -- 
> Jonathan D. Rosenberg, Ph.D.                   499 Thornall St.
> Cisco Fellow                                   Edison, NJ 08837
> Cisco, Voice Technology Group
> jdrosen at cisco.com
> http://www.jdrosen.net                         PHONE: (408) 902-3084
> http://www.cisco.com
>

_______________________________________________
mmusic mailing list
mmusic at ietf.org
https://www.ietf.org/mailman/listinfo/mmusic