[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [MMUSIC] RTSP and NATs
Hi,
I would like to try to summarize and give my comments to Tom's and
Jonathan's discussion. These are issues to continue to discuss here on
the list but which will also be brought up on tomorrow's RTSP
teleconference.
1. We might break out the NAT traversal part into a separate
specification. However we must look at how it relates to mechanisms that
will be in the base spec. I think it is important that as soon as
possible we outline these solutions.
2. The use of "client_rtcp_port" versus changing "client_port". Your are
correct that it might not be necessary to call it RTCP port. What we
will need for STUN support is to possibility to specify two explicit
port numbers without relation to each other. The general case would then
to allow to specify any set of port numbers in both directions. This
will cover any future transport protocol. However I went on to a new
parameter to avoid changing the already present parameters syntax. If
people find this acceptable then we can basically do anything with
"client_port".
3. Symmetric RTP.
- STUN over cone nats work without changing the servers, where symmetric
RTP requires server changes.
- Symmetric has hi-jacking potentials that will impossible to protect
against without RTP/RTCP based ID/authentication scheme.
- Needs binding mechanism for RTP/RTCP packets to the correct RTSP Session.
- Potential RTSP middle box (proxies, caches) problems. Should any way
be fixed for future RECORD servers.
- Otherwise of course a nice solution.
4. ALG's
- ALG: Deployment issues, especially for extensions
- ALG: Sub-standard implementations
- Prevent end-to-end security.
- Interactions between ALG's STUN is a potential problem.
5. UPnP or MIDCOM.
- Time to deploy these solutions.
- no real conflict, if UPnP/midcom is available use it.
6. Jonathan's other proposal
Please provide it so that we can discuss it.
7. Use of "Destination" parameter for NAT/ALG's
- Destination might be needed if NAT maps using several IP addresses to
same client.
- Needs client based authentication of that address/port in RTSP message
to work in secure ways.
Best Regards
Magnus Westerlund
Multimedia Technologies, Ericsson Research ERA/TVA/A
----------------------------------------------------------------------
Ericsson AB | Phone +46 8 4048287
Torshamsgatan 23 | Fax +46 8 7575550
S-164 80 Stockholm, Sweden | mailto: magnus.westerlund@era.ericsson.se
_______________________________________________
mmusic mailing list
mmusic@ietf.org
https://www1.ietf.org/mailman/listinfo/mmusic