RE: iSCSI: Last call issues for the Naming and Discovery document

["KRUEGER,MARJORIE (HP-Roseville,ex1)" <marjorie_krueger@hp.com>]

> This doesn't seem consistent with the iSCSI 
> (draft-ietf-ips-iscsi-16) use of "address". I can't find 
> anywhere where it calls the address the iSCSI name plus a 
> transport address. For example, in 11.8, it defines target address as:
> 
> TargetAddress=domainname[:port][,portal-group-tag]
> 
> This doesn't include iSCSI name and it does include 
> portal-group-tag which isn't mentioned in the Naming and 
> Discovery document. To harmonize the two documents, iSCSI 
> address should not include the iSCSI name and the option for 
> portal-group-tag should be added to Naming and Discovery.

The iSCSI protocol document is not defining the concept of an iSCSI address,
it's specifying the format of a key value pair.  The N&D doc is purely an
informative discussion of concepts.  I don't see any problem with speaking
of an iSCSI address as both the iSCSI name and IP address, since you can't
address a target without knowing it's iSCSI name.
 
> Also, the use of Address in the iSCSI draft is asymmetrical. 
> The iSCSI draft only applies the structure of address, 
> <domain-name>[:<port>], to target addresses. Initiator 
> address is generally not mentioned. While the iSCSI protocol 
> doesn't explicitly use initiator addresses, they may be 
> useful for other things such as management interfaces so a 
> target can report address along with who is connected. The 
> Naming and Discovery document should at least explain the 
> difference between the asymmetry in the use of address in 
> iSCSI and the symmetry in Naming and Discovery. One might 
> also mention that a target will be only able to report the 
> address of a connected initiator in one of the IP formats 
> because it doesn't ever receive a DNS based name.

A target could perform reverse lookup on an address, just as the initiator
would have to do a name lookup to find the target's IP address if it had a
host name.

Marj