Re: [P2PSIP] Peer protocol, DHT and running code

[Enrico Marocco <[e-mail]>]

Philip Matthews wrote:
>> Actually, I don't think we really need to tunnel SIP over the peer
>> protocol.  In the PCAN draft
>> (http://tools.ietf.org/id/draft-marocco-p2psip-xpp-pcan-00.txt) we
>> address the issue of NAT traversal the following way:
>>
>> 1) clients that need to do so would keep sip-outbound connections with
>>    peers.
>> 2) peers that handle sip-outbound flows for some clients would store
>>    routing records containing the full path to the client.
>>
>> Here's an example: imagine client C registers with peer B and
>> establishes an outbound flow with it. C's routing record (inserted in
>> the overlay by B) will contain two hops: B's contact and C's
>> contact/gruu.  In case B is also suffering restricted connectivity, this
>> would mean that it too would have an outbound flow with another peer,
>> say A. In this case (which is pretty unlikely, because if B is
>> restricted it probably won't be able to receive REGISTER requests form
>> C), C's routing record would contain three hops.
>>
>> Having said that, it's probably worth noting that we have not yet
>> decided whether this approach is better compared to routing SIP messages
>> inside the overlay. Both are easily implementable with XPP and XPP-PCAN
>> so I guess we'd need to discuss this further before we really made up
>> our minds.
> 
> I assume that your peers have public IP addresses.

No, not all peers need to have public addresses.  Restricted peers
receive SIP requests on sip-outbound connections.

> If so,  this is "superpeer" approach that is discussed in the "NATs and
> Overlays"  draft
> and in my NAT traversal presentation in Prague. Dean has also suggested
> a similar approach.
> 
> We have had some discussions of the pros and cons of this approach vs
> the approach
> that Eric and I have been advocating of routing messages through a
> partial mesh of connections.
> 
> The whole goal of draft-matthews-p2psip-nats-and-overlays-01 is to
> discuss these two approaches
> and their pros and cons.  Feedback on that draft would be much
> appreciated, and would help
> the  WG decide on a direction.

Actually our proposal uses a partial mesh too.  The difference is that
SIP messages do not need to follow peer protocol paths.

-- 
Ciao,
Enrico

_______________________________________________
P2PSIP mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/p2psip