RE: iSCSI: Wrapping up SendTargets

[Black_David@emc.com]

> I don't see any reason to disallow what Marjorie is asking
> for.  However, we should also decide whether or not a SendTargets
> response may contain a well-known/default/canonical target to
> which the initiator is expected to recursively issue a
> SendTargets command.
> 
> Supporting that feature is much more likely to be a problem
> if we want implementations to transition to other discovery
> mechanisms in the future, as well as creating requirements to
> avoid loops.
> 
> Any thoughts?

On further thought, I agree.  Forbidding recursive use
of SendTargets prevents the nasty problem scenarios such as
recursion problems/bugs that prevent discovery of things that
should have been discovered and loops in the recursion structure.

This prohibition is easy to specify - Initiators won't do the recursion,
hence SendTargets MUST NOT return information about other
canonical targets.  IMHO, this is sufficient, and the tighter
restriction based on same <IP address, TCP port> is not
needed.  If someone wants to have SendTargets return
information about multiple systems, keeping that information
correct and up to date across the systems is their
problem, not ours (long version of "Marjorie was right").

Thanks,
--David
---------------------------------------------------
David L. Black, Senior Technologist
EMC Corporation, 42 South St., Hopkinton, MA  01748
+1 (508) 435-1000 x75140     FAX: +1 (508) 497-8500
black_david@emc.com       Mobile: +1 (978) 394-7754
---------------------------------------------------