RE(4): [AVT] RTP - RTT

["Vikal Kumar Jain" <vikalkj at slingmedia.com>]

> All computers have a clock. Round trip times (RTT) do not require a
> very accurate clock at all, as you are sending out a packet and
> compare time stamps when you receive a response. That can be done even
> if your clock is not set to the accurate time, has a rate error, etc.
> All you need to do is to have some clock accuracy over a few hundred
> milliseconds to a few seconds.

As, I mentioned there are many embedded system that don't have RTC
(real-time clock) and If NTP is used to adjust the system clock (not a
RTC) that 
has it's own complexities I am not aware of. 
 
>If you have neither round trip nor a time standard, you will not be
>able to do what you want. You are trying to make bricks without either
>straw or clay.

As I mentioned earlier it can still be done and would be more accurate
by extending RTCP headers by adding "delay since last RR" in the SR
report even if server does not get any packets back from the client. 

It is possible to calculate the RTT on the Sender (Server) using the
scheme described in RFC 3550 section "6.4.1 SR: Sender Report RTCP
Packet" even for the scenario I mentioned below in which receiver is not
sending any RTP packet back to server without any extensions.

Regards,
Vikal.
-----Original Message-----
From: Marshall Eubanks [mailto:marshall.eubanks at gmail.com] 
Sent: Friday, September 14, 2007 7:20 PM
To: Vikal Kumar Jain
Cc: avt at ietf.org
Subject: Re: RE RE: [AVT] RTP - RTT

On 9/14/07, Vikal Kumar Jain <vikalkj at slingmedia.com> wrote:
>
> I would like to use round trips but that is not possible in the
scenario
> I mentioned below for the receiver (client).
>
> I may be able to do so by extending the RTCP headers in the scenario I
> mentioned below; I can't do that using the standard headers since
there
> is no RTP data sent back from receiver to sender like in case of a
video
> conferencing system.
>
> I am more interested in having a solution that exist and does not
> require me to extend any of the standard headers.

There is an old saying, you can't make bricks without straw.

All computers have a clock. Round trip times (RTT) do not require a
very accurate clock at all, as you are sending out a packet and
compare time stamps when you receive a response. That can be done even
if your clock is not set to the accurate time, has a rate error, etc.
All you need to do is to have some clock accuracy over a few hundred
milliseconds to a few seconds.

One way time requires that both ends have a clock, and that these
clocks be set to some time standard to better than the desired
accuracy. With NTP, you can set clocks to about a millisecond, so you
can do one way timing if everything is set up to about a millisecond.

If you have neither round trip nor a time standard, you will not be
able to do what you want. You are trying to make bricks without either
straw or clay.

Regards
Marshall


>
> Moreover, there may be no notion of a wall clock on many systems and I
> am not very sure of the complexities involved in using a NTP server to
> get NTP timestamps.
>
> Regards,
> Vikal.
>
>
>
> -----Original Message-----
> From: Marshall Eubanks [mailto:marshall.eubanks at gmail.com]
> Sent: Friday, September 14, 2007 5:05 PM
> To: Vikal Kumar Jain
> Cc: avt at ietf.org
> Subject: Re: [AVT] RTP - RTT
>
> Hello;
>
> On 9/14/07, Vikal Kumar Jain <vikalkj at slingmedia.com> wrote:
> >
> >
> >
> >
> > Hi all,
> >
> >
> >
> > I would like to know the ways to calculate RTT on a RTP receiver
> (client)
> > that does not send back packets to sender (Server) in RTP session.
> Since,
> > receiver does not send back any packets to the sender, the sender
> should not
> > include receiver report blocks in the SR RTCP packets and hence
> receiver
> > does not know the "delay on the sender since it received last RR"
and
> can
> > not calculate RTT using the scheme mentioned in section "6.4.1 SR:
> Sender
> > Report RTCP Packet".
> >
> >
> >
> > How does a receiver that does not send back any packets back to the
> sender
> > calculate RTT? I will be obliged if some one could help me on the
> same.
> >
> >
> >
> >  I would like to add one more point.
> >
> >  RFC 3550 also mentions that NTP can be used to calculate RTT, but I
> think
> > it can be used to calculate only one side delay and multiplying it
by
> 2 to
> > get RTT may not be       a great idea as the delay may be different.
> >
> >
>
> I may be missing something, but it seems that you want to calculate
> the round trip delay without actually having any round trips. Isn't
> doubling the one way delay (which of course may be inaccurate) the
> best you can do in that case ?
>
> Regards
> Marshall
>
> >
> > Regards,
> >
> > Vikal.
> >
> >
> > _______________________________________________
> > Audio/Video Transport Working Group
> > avt at ietf.org
> > https://www1.ietf.org/mailman/listinfo/avt
> >
> >
>

_______________________________________________
Audio/Video Transport Working Group
avt at ietf.org
https://www1.ietf.org/mailman/listinfo/avt