<html>
<body>
<font size=3>At 10:48 AM 11/10/2005, Caitlin Bestler wrote:<br>
<blockquote type=cite class=cite cite=""> <br><br>
<br>
Mike Krause wrote in response to Greg Lindahl:<br>
<x-tab> </x-tab><br>
<x-tab> </x-tab><br>
><x-tab> </x-tab>If it is to
be reasonably robust, then RDS should be required to<br>
support<br>
> the resync between the two sides of the communication. This
aligns<br>
with the<br>
> stated objective of implementing reliability in one location in<br>
software and<br>
> one location in hardware. Without such resync being required
in the<br>
ULP,<br>
> then one ends up with a ULP that falls shorts of its stated
objectives<br>
and<br>
> pushes complexity back up to the application which is where the<br>
advocates<br>
> have stated it is too complex or expensive to get it correct.<br>
<x-tab> </x-tab><br><br>
I haven't reread all of RDS fine print to double-check this, but my<br>
impression is that RDS semantics exactly match the subset of MPI<br>
point-to-point communications where the receiving rank is required<br>
to have pre-posted buffers before the send is allowed.<br>
</blockquote><br>
My concern is the requirement that RDS resync the structures in the face
of failure and know whether to re-transmit or will deal with
duplicates. Having pre-posted buffers will help enable the resync
to be accomplished but should not be equated to pre-post equals one can
deal with duplicates or will verify to prevent duplicates from
occurring.<br><br>
Mike</font></body>
</html>