RE: iSCSI: Support Alias in the protocol

[Black_David@emc.com]

A couple of clarifications ...

> I feel that the alias does not belong in the iSCSI architecture.
> Aliasing is already routinely provided at two other levels in
> the protocol stack and need not be redundantly inserted in iSCSI.
> 
> a)  In the SCSI command set
> 
> There is already available a logical unit level aliasing process
> using the SET/REPORT DEVICE IDENTIFIER service actions defined
> in the SCSI command set standard SPC-2.  
> 
> b)  In the operating system stack
> 
> Most implementations use aliases created at the system level, not
> the protocol level, that are far more useful and can be tuned
> to the particular environment.  A particularly naive example of
> this is the mapping of identified devices to the A:, B:, or 
> C: disk drives in the similarly naive MS-DOS operating system.
> More sophisticated systems and management structures use similar
> aliasing at the system level.

As John has pointed out, both of these are LU or LUN aliases rather
that Initiator or Target aliases.  Further, the second one can be
somewhat fragile.  It's entirely too easy to inadvertently change
device mappings (e.g., pull disk E: out of an MS-DOS system, reboot,
and under the right conditions disk F: becomes disk E:).  I'm sure
someone from Veritas can explain the lengths to which the Veritas
Volume Manager (VxVM) goes to avoid relying on any configuration
information of this sort -- VxVM writes its own header on every
volume so that it knows *exactly* what each volume is no matter
how/where it shows up.  

Let me also acknowledge as valid Marj's opinion that anything of
this sort belongs in a management tool rather than the protocol.

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