On Mon, 2008-06-16 at 18:11 +0300, Sasha Khapyorsky wrote:
> On 08:07 Mon 16 Jun     , Hal Rosenstock wrote:
> > > 
> > > If length is encoded in header you can check whole packet size.
> > 
> > It's only a partial measure. It doesn't account for dropped or
> > duplicated packets. It's UDP, right ?
> 
> Yes in inet mode and unix datagram when locally used.

So then even a length in the header is a coarse comparator for whether
the receive is correct or not.

If there's a reasonable probability that no errors occur (within the
current retry strategy) then this might be OK but there is also the
issue of resync'ing when something is dropped to make sure one is really
starting on a header boundary.

-- Hal

> Sasha
> _______________________________________________
> general mailing list
> [email protected]
> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general
> 
> To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

_______________________________________________
general mailing list
[email protected]
http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general

To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general

Reply via email to