ISCSI: Several topics

[Black_David@emc.com]

(1) Section 1.2.1 of the iSCSI draft contains the following:

   An iSCSI target MUST support at least one TCP connection.  An iSCSI 
   initiator SHOULD support several connections in a session.

Based on the long-standing WG consensus that multiple-connection
sessions are optional, that "SHOULD" needs to be replaced
by a "MAY".  I believe this to be WG consensus, and would ask
people not to reopen this issue.

(2) I want to encourage the analysis that Pierre Labat has
started on errors and recovery.  We need that level of detail
to make sure all the mechanisms are right and work as intended
in all circumstances.  Pierre - if you want co-authors, please
ask for them on the list, and I'd encourage writing this material
up as an Internet-Draft so we don't lose the effort invested so far.

(3) Mark Edwards wrote:

	Can anyone quantify for me the percentage difference that adding the
URG 
	pointer would make to this sort of throughput ?  I'm not sure, but
my 
	suspicion about such cards is that URG data does not flow through
the fast 
	path and would get off-loaded to the software.

I think it would greatly help reaching consensus on URG for the
proponents of that mechanism to respond to Mark in technical
detail, not necessarily focussed on current cards.  An
Internet-Draft may be more appropriate than a long
explanation on the mailing list.

--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
---------------------------------------------------