Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 7 Dec 2006 20:51:54 +0800
From:      "maillist ifiaas" <maillist.ifiaas@gmail.com>
To:        "Randall Stewart" <rrs@cisco.com>
Cc:        freebsd-net@freebsd.org
Subject:   Re: Urel, a TCP option for Unreliable Streaming. Need your help.
Message-ID:  <161d69110612070451o6945c7c0y9cf7a7e8b6b7225d@mail.gmail.com>
In-Reply-To: <4578057B.9080002@cisco.com>
References:  <161d69110612060709n3bf99bd4y47d94b021b8f1d02@mail.gmail.com> <m21wncv66z.wl%gnn@neville-neil.com> <4577D858.4010300@freebsd.org> <2786D1DE-E8FA-490C-AFF7-E458E262AEF7@lurchi.franken.de> <161d69110612070242p612e2318ya8d285a15a54e9fc@mail.gmail.com> <4578057B.9080002@cisco.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Thanks Randall.

Am I right to say that, in SCTP, the loss information is reported to
the sender, instead of the receiver?

Btw, TCP Urel is a option. To use it, you have to add something like,
int rc = setsockopt( inSettings->mSock, IPPROTO_TCP,TCP_URE, (char*)
&val, len );
to enable Urel option in TCP.

I still think that partial reliability should be performed in the
application layer. Although transport layer knows more about the
channel condition, but they can either be reported to the application
(like we did on segment loss in TCP Urel),  or be infered by
application (e.g. estimating the current bitrate by looking at the
buffersize). As QoS is only meaningful to application, allowing
flexibilty of implementing QoS (such as partial relaibility) mechenism
in application layer rather than transport layer, looks much natrual to me.

-gavin

On 12/7/06, Randall Stewart <rrs@cisco.com> wrote:
> maillist ifiaas wrote:
> > Michael,
> >
> > In PR-SCTP where retranmission is set off, does it allows the
> > applications to know which part of data is lost in the stream?
> > Thanks
>
> Yep.. you subscribe for a notification event and SCTP will
> return you the data that was not sent.
>
> So not only does it let you know you can actually let
> SCTP hold and return the data that did not get
> acknowledged.. The API also has a context so you
> can attach a user defined 32 bit value to the
> data.. to say bind a pointer to an object to
> the actual data... for lookup or other stuff :-)
>
> R
> >
>
>
>
> --
> Randall Stewart
> NSSTG - Cisco Systems Inc.
> 803-345-0369 <or> 803-317-4952 (cell)
>



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?161d69110612070451o6945c7c0y9cf7a7e8b6b7225d>