Avoiding deadlock in iSCSI

["John Hufferd/San Jose/IBM" <hufferd@us.ibm.com>]

Folks,
We seem to be going around and around on this.

Julian, David Robinson, and myself (I hope among others) seem to think that
with Asymmetric, and a simple ordering rule, everything works with TCP/IP
(because of what TCP/IP itself does).  There maybe some buffer management
in the SCSI "application" (this is the normal stuff that SCSI does), but it
is not an iSCSI/TCP issue.

Can we agree on that for the Asymmetric case?

For the Symmetric case, Julian suggested a few more "minor" rules, and he
claims that then it can work as deadlock free as the above Asymmetric case.

Can we agree on that for the Symmetric case?

.
.
.
John L. Hufferd