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

Re: [MMUSIC] Last Call:draft-ietf-mmusic-sdp-capability-negotiation (SDP CapabilityNegotiation) to Proposed Standard



Hi

Comments inline below 

Regards
Ingemar

> -----Original Message-----
> From: Bob Gilman [mailto:bob_gilman at comcast.net] 
> Sent: den 11 oktober 2008 00:43
> To: Ingemar Johansson S
> Cc: Flemming Andreasen; Jean-Francois Mule; Christer 
> Holmberg; mmusic at ietf.org; Roni Even
> Subject: Re: [MMUSIC] Last 
> Call:draft-ietf-mmusic-sdp-capability-negotiation (SDP 
> CapabilityNegotiation) to Proposed Standard
> 
> Ingemar-
> Thanks for your comments.  I've added a few remarks below.
> -Bob
> 
> Ingemar Johansson S wrote:
> > Hi
> >  
> > To put it short, what we would like to do is to include a 
> mechanism in 
> > the SDP capability negitiation framework (SDPCapNeg) that tells the 
> > endpoint to modify m-line (or c-lines..)  and only indicated the 
> > selected configuration on the a=acfg line. The proposed signal for 
> > this is an extra SDP attribute.
> [rrg]: If we do this through a middlebox that doesn't 
> understand capneg,
>         then wouldn't we wind up fooling the middlebox into 
> thinking the
>         m-line option was chosen? - unless we do something 
> like set the
>         ports to zero as described in RFC3264, section 5.1.  I suppose
>         that could be done by the answerer as part of the rules for
>         a=sdpcapneg-acfg-indication-only processing.  If the answerer
>         doesn't recognize such an attribute, then it doesn't support
>         SDPCapNeg, and it would respond with it's best answer to the
>         m-line(s) and be done with it.  Can we think of a shorter
>         attribute name?
Reading RFC3264 your suggestion to pair this with port=0 seems to make
sense. I guess there are three alternatives here.
1) port number not set to zero in offer or answer: 
  may fool a middlebox as you hint above 
2) port number i set to zero in 1st SDP answer: 
  This may be an indication that the offer is rejected. 
  However the offerer will likely be able to deduce the correct behavior
and issue a second offer/answer.
  Below a list of 1st answer SDP examples with port=0
  a) potential config selected
    ======
    m=audio 0 RTP/AVP 98
    ..
    ..
    a=acfg:1 t=1 ..
    ======
    Offerer draws conclusion that 2nd offer answer is needed as 
      "legacy" SDP body and acfg is uncorrelated    
  b) no potential config selected
    ======
    m=audio 0 RTP/AVP 98
    ..
    ..
    ======
    Normal handling of rejected offers accroding to RFC3264  
3) port number set to zero already in 1st offer:
  RFC3264 explicitly says that an m-line with port=0 MUST NOT be used. 

My conclusion is that alternative 2) is the safest.

Shorter attribute name?, yes please! I am out of ideas for the moment. 


> >  
> > Latent configurations may indeed be useful to in order to 
> solve this 
> > problem but I am not convinced that it solves the problems 
> all the way.
> >  
> > The idea is (to my understanding) that the 1st SDP offer 
> may contain 
> > zero m-lines but a number of latent configurations for 
> audio and video.
> > As the offerer will send an offer with zero m-lines it is easy to 
> > realize that a 2nd offer/answer is needed.
> >  
> > There are a few issues with this however:
> > 1) Assumes that SDPMedCapNeg (SDP media capabilities 
> negotiation) is 
> > supported by the endpoints. An endpoint that does not understand 
> > SDPMedCapNeg is not helped by this trick and will have to 
> resort to a 
> > session setup without SDPCapNeg at all.
> [rrg]: But it's not particularly hurt either - the offerer expected to
>         make two exchanges, so if the answerer doesn't support
>         SDPMediaCapNeg, the offeror will know to construct a 
> conventional
>         offer anyway.
OK, correct

> > We will have this particular situation for the 3GPP MMTel/MTSI 
> > services at least for rel. 7.
> [rrg]: My apologies for the delays in getting the 
> SDPMedCapNeg document
>         finished.  I hope endpoints will all support it when 
> it becomes
>         available - or soon thereafter.  Your situation is a 
> consequence
>         of specifying capability negotiation in pieces.
> >  
> > 2) The examples and the description for lcfg is a bit unclear (no 
> > offence meant, I realize the headache the authors must have 
> had with 
> > this, also the readers, among them "yours truly" should 
> have been more 
> > active with comments) .
> [rrg]: Please let me know what you have in mind here so I can 
> try to fix
>         it for the upcoming meeting.
I believe the issue I have was that it was not quite clear on the
untouched vs. pruned lcfgs below. I will anyway try to get time to do a
thorough read on the draft and see what other parts are unclear. 

> >   As I see it the answerer can either return an lcfg untouched or 
> > pruned to indicate just one possible alternative.
> [rrg]: correct.
> > ...... In the first case the problem
> > is that it will be necesary to send a second offer with several 
> > potential configurations which will leave us in the same 
> situation as 
> > before i.e that the 2nd offer must include a number of potential 
> > configurations which leaves us in the same situation.
> [rrg]: Not necessarily.  If the offerer provides a set of latent
>         configurations, and the answerer responds with some 
> (sub)set of
>         those, then the offerer can select its preferred one and offer
>         only that in the m-line(s) of second exchange.  The second
>         exchange doesn't need (indeed, really shouldn't have) any
>         capability negotiation attributes in either the offer or the
>         answer.
This will work, I believe however that it must be written in the draft
(unless it is already..)

  
> > 3) Other extenstions (such as the extension considered for 
> the c-line) 
> > would need to assume that SDPMedCapNeg is supported by the 
> endpoints 
> > (maybe not a big issue).
> [rrg]: I hope it's not an issue (or at least soon won't be). :-)
> >  
> > 4) Would an SDP offer without any m-lines look odd from an 
> > intermediary point of view ?, according to RFC3264 it 
> should not be an issue.
> [rrg]: Since we're assuming an old intermediary, I wouldn't be too
>         surprised if the intermediary were surprised by a "media-less"
>         session.
> >  
> > My conclusion is that latent configurations does not help us but 
> > please correct me where I make the wrong 
> interpretations/conclusions.
> [rrg]: There seem to be three time periods under consideration:
>         1. Endpoints support SDPCapNeg, but not SDPMedCapNeg, and
>            middleboxes don't support SDPCapNeg.  In this case, lcfg
>            clearly can't help.
>         2. Endpoints support SDPMedCapNeg, but middleboxes don't;
>            lcfg-only works, but requires two exchanges.
>         3. Endpoints and middleboxes support SDPMedCapNeg.  Now
>            lcfg isn't needed, and negotiation can be done in a single
>            exchange.  Note however that the two-stage lcfg technique
>            for period 2 will still work here.
>         It's my hope that period 1 will be short.
> >  
> > /Ingemar
> >  
> >  
> > 
> ----------------------------------------------------------------------
> > --
> > *From:* Flemming Andreasen [mailto:fandreas at cisco.com]
> > *Sent:* den 8 oktober 2008 16:58
> > *To:* Ingemar Johansson S
> > *Cc:* Jean-Francois Mule; Christer Holmberg; mmusic at ietf.org
> > *Subject:* Re: [MMUSIC] Last
> > Call:draft-ietf-mmusic-sdp-capability-negotiation (SDP
> > CapabilityNegotiation) to Proposed Standard
> > 
> >     Thank you for the comments Ingemar. As Jean-Francois 
> noted, we have
> >     been through this issue many times before and the 
> solution specified
> >     in the draft reflects WG consensus, which you also 
> supported in the
> >     past (after admittedly considerable discussion). I 
> think it's worth
> >     taking a step back again, and look at what it is you 
> actually want
> >     to do here, and how we can address it. Quoting from an earlier
> >     exchange on this subject ("SDPCapNeg, modified m-line issue",
> >     5/21/2008 on the MMUSIC list):
> > 
> >     <quote>
> >     Taking a step back, I believe the essence of your 
> comments are, that if
> >     we want to use SDP Capability Negotiation to establish 
> media streams,
> >     then we cannot do that in a single offer/answer 
> exchange. Instead, you
> >     want to have a mechanism whereby we first exchange 
> capabilities between
> >     the offerer and the answerer (without establishing a 
> stream) followed by
> >     another exchange where we actually negotiate the media stream
> >     parameters. In other words, the current 
> single-roundtrip O/A exchange
> >     provided by RFC 3264 and supported by SDPCapNeg is replaced by a
> >     two-roundtrip solution.
> > 
> >     <snip>
> > 
> >     Note however, that the SDPCapNeg framework as currently 
> defined can be 
> >     used more or less the way you seem to prefer by merely 
> including 
> >     capabilities in the offer without the actual potential 
> configuration 
> >     attributes (which are used to trigger the extended O/A 
> exchange). The 
> >     thing that would be missing is expressing valid combinations of 
> >     capabilites as well as preferences, however this is 
> part of what the 
> >     media capabilities extension document is addressing by 
> introducing the 
> >     "latent configuration" attribute, which you could then use.
> >     </quote>
> > 
> > 
> >     In lieu of all the discussion and WG consensus on the currently
> >     specified mechanism, can you please explain why the 
> above cannot be
> >     used to address your concerns, i.e. either just send 
> capabilities
> >     initially, or define an extension that provides the 
> exact semantics
> >     you are looking for  ?
> > 
> >     Thanks
> > 
> >     -- Flemming
> > 
> > 
> >     Ingemar Johansson S wrote:
> >>     Hi
> >>
> >>     I believe the threads below pretty much summarizes the previous
> >>     discussions around this topic. 
> >>     The difference is that the new effort is to introduce 
> an extra indicator
> >>     that essentially says that the 1st offer/answer is 
> only capability
> >>     exhange in the sense that if 
> "a=spdcapneg-acfg-indication-only" (or
> >>     something with the same meaning) is present at the 
> session level of the
> >>     offer SDP, then the answerer MUST only specify actual 
> configuration on
> >>     the a=acfg line. As mentioned before the proposed 
> change is minimal and
> >>     as far as I can see it does not break the framework.
> >>
> >>     The indicator is not mandatory to use other than that it may be
> >>     specified as mandatory in another standards text that 
> use the SDP
> >>     Capability Negotiation Framework (for instance a 3GPP 
> specification).
> >>
> >>     /Ingemar
> >>
> >>
> >>       
> >>>     -----Original Message-----
> >>>     From: Jean-Francois Mule [mailto:jf.mule at cablelabs.com] 
> >>>     Sent: den 7 oktober 2008 23:08
> >>>     To: Ingemar Johansson S; Christer Holmberg
> >>>     Cc: Flemming Andreasen; mmusic at ietf.org
> >>>     Subject: RE: [MMUSIC] Last 
> >>>     Call:draft-ietf-mmusic-sdp-capability-negotiation (SDP 
> >>>     CapabilityNegotiation) to Proposed Standard
> >>>
> >>>     (Removing ietf at ietf.org to discuss your Last Call comment on 
> >>>     the mmusic
> >>>     list.)
> >>>
> >>>     Ingemar, Christer,
> >>>
> >>>        Thank you for the comment. 
> >>>
> >>>        I would like to note the following history related to your 
> >>>     IETF Last Call comment:
> >>>     - On June 17, 2008, an email thread "Closing on the SDPCapNeg 
> >>>     modified m-line issue" was initiated:
> >>>     
> http://www.ietf.org/mail-archive/web/mmusic/current/msg06939.html
> >>>     It was addressed to Flemming, Ingemar and Christer 
> >>>     explicitly, copied to the list.
> >>>     - Between June 17 and July 7, 3 folks responded to the thread 
> >>>     in favor of moving forward (including Ingemar), and there 
> >>>     were not one email asking to reopen or address the issue:
> >>>     
> http://www.ietf.org/mail-archive/web/mmusic/current/msg06940.html
> >>>     
> http://www.ietf.org/mail-archive/web/mmusic/current/msg06941.html
> >>>     
> http://www.ietf.org/mail-archive/web/mmusic/current/msg07002.html
> >>>     - at IETF#72, on 8/1, Joerg and I put up the WG status slides 
> >>>     (http://www.ietf.org/proceedings/08jul/slides/mmusic-8.pdf, 
> >>>     page 5) showing SDP capneg as intended for publication 
> >>>     request as soon as draft
> >>>     09 was out.  Both of you were at the meeting and commented on 
> >>>     other draft discussions.
> >>>     - On 8/29, the mmusic WG meeting minutes were posted and 
> >>>     announced to the list.  The minutes
> >>>     (http://www.ietf.org/proceedings/08jul/minutes/mmusic.html) 
> >>>     show the ID is awaiting publication request.
> >>>
> >>>     Just to make sure I get the facts right, was there any new 
> >>>     email on this that I missed?  If so, please resend.
> >>>
> >>>     Thank you
> >>>     Jean-Francois.
> >>>     MMUSIC co-chair
> >>>
> >>>         
> >>>>     -----Original Message-----
> >>>>     From: mmusic-bounces at ietf.org 
> [mailto:mmusic-bounces at ietf.org] On 
> >>>>     Behalf Of Ingemar Johansson S
> >>>>     Sent: Tuesday, October 07, 2008 6:38 AM
> >>>>     To: ietf at ietf.org
> >>>>     Cc: Flemming Andreasen; mmusic at ietf.org; Christer Holmberg
> >>>>     Subject: Re: [MMUSIC] Last 
> Call:draft-ietf-mmusic-sdp-capability-
> >>>>     negotiation (SDP CapabilityNegotiation) to Proposed Standard
> >>>>
> >>>>     Hi
> >>>>
> >>>>     In draft-ietf-mmusic-sdp-capability-negotiation it 
> is proposed that 
> >>>>     the transport in the m-line is modified in the SDP answer. 
> >>>>           
> >>>     While this
> >>>         
> >>>>     is a good idea in order to reduce the number of
> >>>>           
> >>>     offer/answer exchanges
> >>>         
> >>>>     it can in fact cause problems with intermediaries 
> that for instance 
> >>>>     compare offer and answer SDP's. The problem may be
> >>>>           
> >>>     exacerbated further
> >>>         
> >>>>     with the addition of extensions to the framework.
> >>>>
> >>>>     Section 3.12 in the draft says:
> >>>>     "The solution to this problem is to upgrade the 
> intermediary to 
> >>>>     support the SDP Capability Negotiation framework".
> >>>>     We find the conclusion unsatisfactory as the problem 
> is that such 
> >>>>     upgrades are not done overnight and there will 
> therefore, for a 
> >>>>     foreseeable future, exist middleboxes in the network 
> that does not 
> >>>>     understand the SDP Capability Negotiation Framework.
> >>>>
> >>>>     Our proposal to solve this issue is to introduce an
> >>>>           
> >>>     indicator in the
> >>>         
> >>>>     SDP that tells that the actual configuration MUST only be
> >>>>           
> >>>     indicated on
> >>>         
> >>>>     the a=acfg line. A proposed SDP attribute for this may be 
> >>>>     "a=spdcapneg-acfg-indication-only".
> >>>>
> >>>>     In essence it means that (if the indicator is included in
> >>>>           
> >>>     the SDP) the
> >>>         
> >>>>     first offer/answer exchange will only be done to get 
> the actual 
> >>>>     configuration (a=acfg...). This would affect section 
> 3.6.2 in the 
> >>>>     draft.
> >>>>     If the indicator is included in the offer SDP it makes a 2nd 
> >>>>     offer/answer exchange a MUST in order to complete 
> the offer/answer 
> >>>>     exchange. Section 3.6.3 specifies a "SHOULD" regarding the
> >>>>           
> >>>     behavior if
> >>>         
> >>>>     the actual cofiguration and the SDP does not match. 
> It is possible 
> >>>>     that "SHOULD" must be changed to "MUST" here.
> >>>>
> >>>>     A UA that for some reason knows that intermediaries don't
> >>>>           
> >>>     understand
> >>>         
> >>>>     the new framework it will add the said SDP attribute 
> at the session 
> >>>>     level in the offer-SDP.
> >>>>     Indications that intermediaries don't understand the new
> >>>>           
> >>>     framework may
> >>>         
> >>>>     for instance in the 3GPP IMS case be that for a specific
> >>>>           
> >>>     3GPP release
> >>>         
> >>>>     the said attribute is mandated, a possible location for
> >>>>           
> >>>     such a text in
> >>>         
> >>>>     this particular case is 3GPP TS26.114.
> >>>>
> >>>>     We believe that the addition of the new functionality to
> >>>>           
> >>>     the framework
> >>>         
> >>>>     will increase acceptance for the SDP Capability Negotiation
> >>>>           
> >>>     framework
> >>>         
> >>>>     greatly and this without breaking the framework.
> >>>>
> >>>>     PS. This issue has been raised previously in the MMUSIC WG,
> >>>>           
> >>>     still we
> >>>         
> >>>>     want to bring this up again for your consideration.
> >>>>
> >>>>     Regards
> >>>>
> >>>>     Ingemar Johansson
> >>>>     Christer Holmberg
> >>>>
> >>>>
> >>>>     _______________________________________________
> >>>>     mmusic mailing list
> >>>>     mmusic at ietf.org
> >>>>     https://www.ietf.org/mailman/listinfo/mmusic
> >>>>           
> >>
> >>       
> 
> 
> -------------------------------------------------------------
> Bob Gilman        bob_gilman at comcast.net      +1 303 898 9780
> 
> 
> 
_______________________________________________
mmusic mailing list
mmusic at ietf.org
https://www.ietf.org/mailman/listinfo/mmusic