Re: [AVT] New draft of "Rapid Acquisition of Multicast RTPSessions (RAMS)" available

["Roni Even" <[e-mail]>]

Ali,
Your SDP example will not help anyone who will try to develop interop
application, the c-line in the unicast stream and the port should refer to
the same device since there is no other attribute in the example that points
to a different semantics to anyone who will look at it. The semantics that
you are using where the c-line is for the RS and the port is for the RR and
the RS assuming that the RS somehow knows the IP address of the RR based on
the RAMS-R request may work in some cases but should not be the one shown as
the primary example and should not be the recommended way, The recommended
way should be to convey the address of the RR in the RAMS-R and this is the
reason I pointed to RTSP. The semantics of the SDP in the example is not
documented in your text and is not obvious to anyone who will just look at
it.
   
I also strongly object to mandating that the client MUST use a port defined
by the server to receive the unicast stream. The RS can recommend a port but
it is the client (RR) decision to decide which port to use.



Roni Even

-----Original Message-----
From: Ali C. Begen (abegen) [mailto:[e-mail]] 
Sent: Sunday, April 26, 2009 8:02 PM
To: Roni Even; [e-mail]
Cc: Bill Ver Steeg (versteb)
Subject: RE: [AVT] New draft of "Rapid Acquisition of Multicast RTPSessions
(RAMS)" available

Hi Roni,

Thanks for the detailed answer. Admittedly, I don't know much about RTSP
beyond its basics. However, I doubt how much it applies in our case.
"Unicast" in RTSP context mostly refers to video-on-demand kind of
applications where the primary distribution is over unicast. In contrast,
our primary distribution is over multicast, and we use unicast as a repair
stream. Thus, I still believe the SDP we have should work fine. RS gathers
the IP address of the RR when it receives the RAMS request. The samething
happens when RR sends a NACK for requesting retransmissions.

That is not to say that this SDP is the only SDP that will work in this
system. There may be others. But, I still don't agree with you when you say
that this SDP won't work. My expectation is that we will include other SDP
examples as we move forward.

On your last point, NAT is a different issue and there has to be a way to
deal with it. Muxing sounds reasonable, and we may or may not mandate it, I
am not sure at this point. An alternative could be that RR opens ports on
its NAT device and includes them in the request message it sends. These
could be in RAMS-R TLV extensions, or could be a separate RTCP packet solely
for this purpose within the same RTCP compound packet that carries the
RAMS-R. I believe Bill mentioned these briefly in his earlier email. We are
open to suggestions at this point and the NAT section will cover these
issues once we have a WG item.

-acbegen

 

> -----Original Message-----
> From: Roni Even [mailto:[e-mail]] 
> Sent: Sunday, April 26, 2009 1:14 AM
> To: Ali C. Begen (abegen); [e-mail]
> Cc: Bill Ver Steeg (versteb)
> Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> Multicast RTPSessions (RAMS)" available
> 
> Ali,
> I looked at RTSP
> http://tools.ietf.org/html/draft-ietf-mmusic-rfc2326bis-20#appendix-D 
> 
> 
> 
> D.1.2.  Media Streams
> 
>    The "m=" field is used to enumerate the streams.  It is 
> expected that
>    all the specified streams will be rendered with appropriate
>    synchronization.  If the session is over multicast, the port number
>    indicated SHOULD be used for reception.  The client MAY try to
>    override the destination port, through the Transport header.  The
>    servers MAY allow this, the response will indicate if 
> allowed or not.
>    If the session is unicast, the port numbers are the ones 
> RECOMMENDED
>    by the server to the client, about which receiver ports to use; the
>    client MUST still include its receiver ports in its SETUP request.
>    The client MAY ignore this recommendation.  If the server has no
>    preference, it SHOULD set the port number value to zero.
> 
> 
> Note that the client has the choice on which port to receive 
> the unicast
> stream, this is why there should be such a parameter in RAMS-R
> 
> 
> Also
> 
> " D.1.8.  Connection Information
> 
>    In SDP, the "c=" field contains the destination address 
> for the media
>    stream.  For on-demand unicast streams and some multicast streams,
>    the destination address MAY be specified by the client via 
> the SETUP
>    request, thus overriding any specified address.  To 
> identify streams
>    without a fixed destination address, where the client is 
> required to
>    specify a destination address, the "c=" field SHOULD be 
> set to a null
>    value.  For addresses of type "IP4", this value MUST be "0.0.0.0",
>    and for type "IP6", this value MUST be "0:0:0:0:0:0:0:0" 
> (can also be
>    written as "::"), i.e. the unspecified address according 
> to RFC 4291
>    [RFC4291]."
> 
> 
> As for the c-line in the unicast stream the whole section you quoted
> 
> " If the session is multicast, the connection address will be an IP
>       multicast group address.  If the session is not multicast, then
>       the connection address contains the unicast IP address of the
>       expected data source or data relay or data sink as determined by
>       additional attribute fields.  It is not expected that unicast
>       addresses will be given in a session description that is
>       communicated by a multicast announcement, though this is not
>       prohibited."
> 
> Now the expectation here is for the address of the stream 
> described by the
> port in the m-line, I do not see any other attribute. Also 
> look at the above
> text from RTSP. The RS does not know the IP address on the RR 
> to where to
> send the RTP unicast repair stream. If you assume it is to 
> the IP address
> from where the RTCP RAMS-R came from this is not mandated 
> that the client
> will have such implementation. Furthermore the address that 
> the RS will see
> in the received RTCP RAMS-R packet may have been change by a 
> NAT or SBC.
> 
> 
> Roni
> 
> 
> 
> -----Original Message-----
> From: Ali C. Begen (abegen) [mailto:[e-mail]] 
> Sent: Sunday, April 26, 2009 1:23 AM
> To: Roni Even; [e-mail]
> Cc: Bill Ver Steeg (versteb)
> Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> Multicast RTPSessions
> (RAMS)" available
> 
> Inline.
> 
> >         m=video 41000 RTP/AVPF 98
> >         i=Primary Multicast Stream #2
> >         c=IN IP4 224.1.1.2/255
> >         a=source-filter: incl IN IP4 224.1.1.2 192.0.2.2
> > 
> > Source (192.0.2.2) sends the rtp packets to the multicast 
> > group (224.1.1.2)
> > with a dest port 41000.
> > 
> >         a=rtpmap:98 MP2T/90000
> >         a=rtcp:41001 IN IP4 192.0.2.1
> > 
> > The feedback target (RS) address for this SSM session is 
> > 192.0.2.1 and its
> > port is 41001. This is the address/port where RR sends the 
> RAMS-R, or
> > receiver reports for the SSM session.
> > 
> >         a=rtcp-fb:98 nack
> >         a=rtcp-fb:98 nack ssli
> >         a=ssrc:123321 cname:[e-mail]
> >         a=mid:3
> >         m=video 41002 RTP/AVPF 99
> > 
> > The retransmission packets go to port 41002. This is the port 
> > RRs listen to
> > for retransmission and RAMS.
> > 
> > Roni E: This implies that the multicast source defines the 
> > port that all
> > clients MUST use to receive the unicast information from the 
> > Burst source.  
> 
> Yes, that is what it implies (One nit is that I would call it 
> unicast RTP
> packets rather than information since we may use another port for RTCP
> packets).
>  
> >         i=Unicast Retransmission Stream #2 (Ret. and Rapid 
> > Acq. Support)
> >         c=IN IP4 192.0.2.1
> > 
> > This is where the retransmission packets come from, same as 
> > the feedback
> > target (in this example).
> > 
> > Roni E: If this is the address of the feedback target it 
> > belongs to the
> > above m-line (m=video 41002 RTP/AVPF 99) that according to 
> > you gives the
> > port on the RR and not on the RS. The problem is that 
> 
> This line actually specifies the sender for the unicast retransmission
> session. In our case, it just happens to be the feedback 
> target for the
> primary SSM session, too. I recall some discussion about using a
> retransmission source which is different than the feedback 
> target for the
> SSM. In that case, the address would be different, I guess.
> 
> > according to SDP the
> > combination of the c= and port on the m-line provides the 
> > transport address
> > to where a stream will be sent so according to SDP this 
> 
> The port on the m line is the port where the retranmission 
> packets are sent.
> However, c line denotes the source for these packets. RFC 
> 4566 page 13 says
> that "... If the session is not multicast, then the connection address
> contains the unicast IP address of the expected data source 
> or data relay or
> data sink as determined by additional attribute fields..."
> 
> Am I missing something?
> 
> > m-line specify an
> > address and port on the Feedback Target and not on the RR and 
> > only to this
> > address RTP payload type 99 will be sent. I am positive that 
> > this is not
> > what you want.
> 
> There can be other payload types we might need to send on this
> retransmission session in the future. In that case, we will add them.
>  
> >         a=rtpmap:99 rtx/90000
> >         a=rtcp:41003
> > 
> > This is where the RTCP packets for the retransmission 
> session go. For
> > example, RAMS-I goes to this port on RRs. RAMS-T goes to this 
> > port on RS.
> > 
> > Roni E: So now you assume that the RR must use the same port 
> > as the FT for
> > this communication. I could not find any place that one 
> > m-line gives the
> 
> Not the m line, but the "a=rtcp" line. Is this wrong?
> 
> > address for both sides. This RTCP port is on 192.0.2.1 so it 
> > will be used
> > for RAMS-R but not for RAMS-I.
> 
> RAMS-R should go to the feedback target address and port for 
> the primary SSM
> session since it is a feedback message for the SSM session. 
> But, RAMS-T is a
> feedback message for the retransmission session, thus, it 
> should go to the
> RTCP port of the retransmission session (41003). I hope this 
> also clarifies
> your question in the next email. If you think I got the ports 
> wrong, let me
> know.
> 
> -acbegen
>  
> >         a=fmtp:99 apt=98; rtx-time=5000
> >         a=mid:4
> > 
> > 
> > -acbegen
> >  
> > 
> > > -----Original Message-----
> > > From: Roni Even [mailto:[e-mail]] 
> > > Sent: Saturday, April 25, 2009 1:11 PM
> > > To: Ali C. Begen (abegen); [e-mail]
> > > Cc: Bill Ver Steeg (versteb)
> > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > Multicast RTPSessions (RAMS)" available
> > > 
> > > Ali,
> > > Can you please write three addresses from this strange SDP.
> > > 
> > > 1. The address and port of multicast
> > > 
> > > 2. The address and port of the RS where the RTCP FB should go to
> > > 
> > > 3. The address and port on the RR where the unicast stream 
> > > should be sent to
> > > 
> > > Roni
> > > 
> > > -----Original Message-----
> > > From: Ali C. Begen (abegen) [mailto:[e-mail]] 
> > > Sent: Saturday, April 25, 2009 11:02 PM
> > > To: Roni Even; [e-mail]
> > > Cc: Bill Ver Steeg (versteb)
> > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > Multicast RTPSessions
> > > (RAMS)" available
> > > 
> > > > Ali,
> > > > The example SDP is syntactically correct but it does not 
> > > say that the
> > > > rtp/rtcp mux will be used and it does not give the 
> > > > information to where the
> > > > unicast stream will be sent when the RR sends a RAMS-R. 
> > > 
> > > To my knowledge, the first line in the following SDP tells 
> > > the RRs on which
> > > port they will receive the retransmission/burst packets.
> > > 
> > >         m=video 41002 RTP/AVPF 99
> > >         i=Unicast Retransmission Stream #2 (Ret. and Rapid 
> > > Acq. Support)
> > >         c=IN IP4 192.0.2.1
> > >         a=rtpmap:99 rtx/90000
> > >         a=rtcp:41003
> > >         a=fmtp:99 apt=98; rtx-time=5000
> > >         a=mid:4
> > > 
> > > There is a typo, you are right. That "a=recvonly" line should 
> > > only exist in
> > > the SDP sent to RRs. In the SDP sent to RS, we should rather have
> > > "a=sendonly". I will make this correction in the next version.
> > > 
> > > The feedback target for the SSM session and the RTCP port for the
> > > retransmission session are also defined in the SDP.
> > > 
> > > Hope this clarifies.
> > > 
> > > BR,
> > > -acbegen
> > >  
> > > > 
> > > > I am not sure why the unicast stream m-line has a port 
> > > number with an
> > > > attribute of recvonly. What is the use case for that. The 
> > > > only information
> > > > that the RR will need is the RTCP port on the RS to where to 
> > > > send the RAMS-R
> > > > message.
> > > > Roni
> > > > -----Original Message-----
> > > > From: Ali C. Begen (abegen) [mailto:[e-mail]] 
> > > > Sent: Friday, April 24, 2009 10:37 PM
> > > > To: Roni Even; [e-mail]
> > > > Cc: Bill Ver Steeg (versteb)
> > > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > > Multicast RTPSessions
> > > > (RAMS)" available
> > > > 
> > > > This is a part of an example SDP sent to RS and RR in a SAP 
> > > > announcement,
> > > > for example. So, the SDP describes what both parties should 
> > > > do (RR cannot
> > > > say that he wants to receive this multicast on its favorite 
> > > port). The
> > > > individual SDPs sent to RR or RS may include other portions 
> > > > of descriptions
> > > > that will contain specific information. 
> > > > 
> > > > -acbegen
> > > > 
> > > > > -----Original Message-----
> > > > > From: Roni Even [mailto:[e-mail]] 
> > > > > Sent: Friday, April 24, 2009 12:23 PM
> > > > > To: Ali C. Begen (abegen); [e-mail]
> > > > > Cc: Bill Ver Steeg (versteb)
> > > > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > > > Multicast RTPSessions (RAMS)" available
> > > > > 
> > > > > Ali,
> > > > > I think you get it wrong, this SDP is from the RS and not the 
> > > > > RR so the RS
> > > > > cannot specify to which address it will send it can only 
> > > > > specify to which
> > > > > address it can receive RTP stream. In this SDP the relevant 
> > > > > information is
> > > > > that the request for retransmission will be sent by the RR to 
> > > > > port 41003
> > > > > Roni
> > > > > 
> > > > > -----Original Message-----
> > > > > From: Ali C. Begen (abegen) [mailto:[e-mail]] 
> > > > > Sent: Friday, April 24, 2009 10:01 PM
> > > > > To: Roni Even; [e-mail]
> > > > > Cc: Bill Ver Steeg (versteb)
> > > > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > > > Multicast RTPSessions
> > > > > (RAMS)" available
> > > > > 
> > > > > Oh, I see. The burst goes to the port that we would 
> > > > normally send the
> > > > > retransmissions. For example, consider the SDP from the draft:
> > > > > 
> > > > >         m=video 41000 RTP/AVPF 98
> > > > >         i=Primary Multicast Stream #2
> > > > >         c=IN IP4 224.1.1.2/255
> > > > >         a=source-filter: incl IN IP4 224.1.1.2 192.0.2.2
> > > > >         a=rtpmap:98 MP2T/90000
> > > > >         a=rtcp:41001 IN IP4 192.0.2.1
> > > > >         a=rtcp-fb:98 nack
> > > > >         a=rtcp-fb:98 nack ssli
> > > > >         a=ssrc:123321 cname:[e-mail]
> > > > >         a=mid:3
> > > > >         m=video 41002 RTP/AVPF 99
> > > > >         i=Unicast Retransmission Stream #2 (Ret. and Rapid 
> > > > > Acq. Support)
> > > > >         c=IN IP4 192.0.2.1
> > > > >         a=recvonly
> > > > >         a=rtpmap:99 rtx/90000
> > > > >         a=rtcp:41003
> > > > >         a=fmtp:99 apt=98; rtx-time=5000
> > > > >         a=mid:4
> > > > > 
> > > > > In this case, the burst will go to port 41002 on the RR. 
> > > The RAMS-I
> > > > > message(s), which is an RTCP feedback message, will go to 
> > > > port 41003.
> > > > > 
> > > > > HTH,
> > > > > -acbegen
> > > > > 
> > > > > > -----Original Message-----
> > > > > > From: Roni Even [mailto:[e-mail]] 
> > > > > > Sent: Friday, April 24, 2009 11:44 AM
> > > > > > To: Ali C. Begen (abegen); [e-mail]
> > > > > > Cc: Bill Ver Steeg (versteb)
> > > > > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > > > > Multicast RTPSessions (RAMS)" available
> > > > > > 
> > > > > > Ali,
> > > > > > I will try to explain in simple way
> > > > > > 
> > > > > > When the RS receives the RAMS-R it need to start sending an 
> > > > > > RTP stream to
> > > > > > the RR. 
> > > > > > In order to send a unicast packet, the RS needs to know a 
> > > > > > transport address
> > > > > > on the RR to where the RTP stream will be sent. The current 
> > > > > > draft does not
> > > > > > say how the RS knows this address. There is no SDP 
> from RR to 
> > > > > > RS like you
> > > > > > mention in your response.
> > > > > > This is why I say that the current draft does not specify a 
> > > > > > solution that
> > > > > > can be implemented as a working solution
> > > > > > Roni
> > > > > > 
> > > > > > -----Original Message-----
> > > > > > From: Ali C. Begen (abegen) [mailto:[e-mail]] 
> > > > > > Sent: Friday, April 24, 2009 7:38 PM
> > > > > > To: Roni Even; [e-mail]
> > > > > > Cc: Bill Ver Steeg (versteb)
> > > > > > Subject: RE: [AVT] New draft of "Rapid Acquisition of 
> > > > > > Multicast RTPSessions
> > > > > > (RAMS)" available
> > > > > > 
> > > > > > Hi Roni, 
> > > > > > 
> > > > > > > -----Original Message-----
> > > > > > > From: [e-mail] 
> [mailto:[e-mail]] On 
> > > > > > > Behalf Of Roni Even
> > > > > > > Sent: Friday, April 24, 2009 4:24 AM
> > > > > > > To: [e-mail]
> > > > > > > Cc: Bill Ver Steeg (versteb)
> > > > > > > Subject: Re: [AVT] New draft of "Rapid Acquisition of 
> > > > > > > Multicast RTPSessions (RAMS)" available
> > > > > > > 
> > > > > > > Hi,
> > > > > > > 
> > > > > > > I think that the current draft does not give a 
> > description of 
> > > > > > > a system that works since there is no text 
> > explaining how the 
> > > > > > 
> > > > > > What do you mean by "a system that works"? 
> > > > > > 
> > > > > > > RS knows the unicast transport address on the RR to 
> > where to 
> > > > > > > send the stream. 
> > > > > > 
> > > > > > Once RS receives the request packet from an RR, RS 
> knows its 
> > > > > > address. Ports
> > > > > > are defined in the SDP. If you are asking about 
> "NAT" issues, 
> > > > > > we have a
> > > > > > section for it, and we plan to complete it as we move 
> > > > > > forward. It is not as
> > > > > > critical as the other parts for now.
> > > > > >  
> > > > > > > If you mandate the use of RTP/RTCP mux it should say so 
> > > > > > > otherwise the RAMS-R should have an optional 
> parameter that 
> > > > > > > supplies this information and a flag for using 
> RTP/RTCP mux.
> > > > > > 
> > > > > > IMO, we cannot mandate muxing as it is not required to make 
> > > > > > RAMS work. If
> > > > > > muxing is supported, the SDP should reflect it.
> > > > > > 
> > > > > > BR,
> > > > > > -acbegen
> > > > > >  
> > > > > > > Thanks
> > > > > > > 
> > > > > > > Roni Even
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > From: [e-mail] 
> [mailto:[e-mail]] On 
> > > > > > > Behalf Of Bill Ver Steeg (versteb)
> > > > > > > Sent: Thursday, April 16, 2009 7:39 PM
> > > > > > > To: [e-mail]
> > > > > > > Subject: [AVT] New draft of "Rapid Acquisition of 
> Multicast 
> > > > > > > RTP Sessions (RAMS)" available
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > There is a new draft of the "Rapid Acquisition of 
> Multicast 
> > > > > > > RTP Sessions (RAMS)" draft available at 
> > > > > > > 
> > http://www.ietf.org/internet-drafts/draft-versteeg-avt-rapid-s
> > > > > > ynchronization-for-rtp-03.txt 
> <http://www.ietf.org/internet->
> > > > > > 
> > drafts/draft-versteeg-avt-rapid-synchronization-for-rtp-03.txt> 
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > We have incorporated the changes from the technical 
> > breakout 
> > > > > > > session in San Francisco. The major changes in 
> this version 
> > > > > > > of the draft include 
> > > > > > > 
> > > > > > > 1- Changing the document title to avoid confusion 
> > with other 
> > > > > > > ongoing "synchronization" drafts
> > > > > > > 
> > > > > > > 2- changing the message names to reflect the title change
> > > > > > > 
> > > > > > > 3- clarification of the RTCP message semantics, including 
> > > > > > > changes to the "Request" and "Inform" messages
> > > > > > > 
> > > > > > > 4- additional description/motivation for the 
> > various message 
> > > > > > > flows has been added
> > > > > > > 
> > > > > > > 5- RTP/RTCP muxing has been added
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > We hope to make this a Working Group item, and 
> will change 
> > > > > > > the name of the draft to avoid conflicts with other 
> > > > > > > "synchronization" drafts at that time.
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > Bill VerSteeg
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > >  
> > > > > > > 
> > > > > > > 
> > > > > > 
> > > > > > 
> > > > > 
> > > > > 
> > > > 
> > > > 
> > > 
> > > 
> > 
> > 
> 
>