RE: The third alternative

[Black_David@emc.com]

> The only argument for SCSI wedge drivers where that they EXIST ALREADY.
> Pretty weak argument for those building new equipment and for
> interoperability.

This severely understates the case.  Not only do wedge drivers exist
already,
but they do a number of things that an iSCSI standard will never do.  Wedge
drivers tend to contain logic specific to the device that they're providing
the
wedge for - that's inappropriate to standardize, either in IETF or T10.  In
addition, there are implementation advantages to building wedge drivers
above the SCSI level - one doesn't have to spread SCSI connection state
across a failure boundary.

> I think that if we keep ourselves honest we have to either:
> 
> - provide for multiple connections at the iSCSI level as it is transport
problem
>    that other TCP applications are not compelled to handle (I hear already
BUT SCTP
>   handles it!) and hope that one day the session concept will drift into
pure transport

Provide for is fine, as long as they're not required.  I would expect to see
significant
deployment of systems that use single TCP connection per session and handle
multiple sessions in wedge drivers.  Arguments like the one Bob has made is
one of the reasons for multiple connections/session to remain OPTIONAL.

> - go to T10 and ask the to standardize wedge drivers!

Just say no!  Exactly what would you want T10 to standardize?

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