Re: [Lksctp-developers] Do piggybacked ACKs work?

[Vlad Y. <vla...@hp...>]

Doug Graham wrote:
> Michael Tüxen wrote:
>> Doug,
>>
>> could you provide a small tacefile (.pcap format)
>> which shows the behaviour you are talking about?
> 
> Ok, I'll do that later today.  Is it ok if I send it to the linux-sctp 
> list instead of this one?
> Vlad prefers the former and I'd prefer to stick to one list.
> 
> Anyway, I've got some reading to do first it seems.  I'm sure hoping 
> that somebody has
> misinterpreted the RFC and that piggybacked ACKs are in fact allowed in 
> the scenario
> I mention, but I guess I need to read the "fine print" in the RFCs 
> before being convinced
> one way or the other.  Sending a lot of small messages around sounds 
> like exactly the
> kind of thing that SCTP was designed to do, and if piggybacked ACKs can't
> be used in that scenario, I think that's a very surprising flaw in the 
> protocol!

it's the request-response scenario that suffers.  If you only have 1 packet
in flight at a time, you end up waiting 200ms for a SACK and with Nagle turned
on, the performance suffers.

If a new request is not predicated on a response, there are no problems.

-vlad

> 
> --Doug
> 
> 
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with 
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Lksctp-developers mailing list
> Lks...@li...
> https://lists.sourceforge.net/lists/listinfo/lksctp-developers
>