Re: [Simple] Questions regarding MSRP drafts

[Ben Campbell <ben at nostrum.com>]

On Aug 11, 2005, at 8:30 AM, Remi Denis-Courmont wrote:

>     Hello,
>
> Le mer 10/08/2005 à 23:33, ext Paul Kyzivat a écrit :
>
> > The cases where reordering can occur are actually pretty obscure I
> > think. (Somebody correct me if I am wrong.) The one I can think of  
> > is:
> >
> > - Two connections between a pair of relays. Two messages from same
> > connection get sent, one over each connection. The 1st message is  
> > very
> > long, while the 2nd is very short. The last byte of the 2nd  
> > message may
> > be received before the last byte of the first message. Then it gets
> > forwarded on.
>
> That only happens if relays are to maintain multiple connections to  
> the
> same peer. Might be wrong, but I doubt ensuring they only have one
> connection at a time (of course renewing should it fail) would pose  
> any
> problem nor much additionnal load.

Multiple simultaneous connections to the same peer are not required.  
One corner case that was discussed is that a relay sends a message to  
a downstream relay, terminates the connection for whatever reason,  
then reconnects to send a new message--but this connection is in fact  
to a different downstream relay for load balancing reasons. The two  
messages now follow different paths through the network, and cannot  
be guaranteed to arrive in order.


> On the other hand, supporting re-ordering, without any sane limit  
> on the
> re-ordering window size sounds very dangerous and prone to abuse to  
> me.
> At least, there ought to be an error code for client not able or not
> willing to accept a chunk because of memory constraints.

Reordering should be rare enough that I am not convinced we need to  
solve this problem. But I will as usual yield to the work group  
consensus on that matter. What do others think?

> --
> Remi Denis-Courmont <remi.denis-courmont at nokia.com>
> Nokia-NRC/Helsinki
>
> _______________________________________________
> Simple mailing list
> Simple at ietf.org
> https://www1.ietf.org/mailman/listinfo/simple


_______________________________________________
Simple mailing list
Simple at ietf.org
https://www1.ietf.org/mailman/listinfo/simple