Re: iSCSI Naming and Discovery

["Jim Hafner/Almaden/IBM" <hafner@almaden.ibm.com>]

David,

<snip>

>I assert that within the iSCSI transport the service level naming
>never appears (I am ignoring 3rd party transfers for now). Thus
?using some sort of URL like name as part of a connection login
>is redundant information, the target doesn't need to be told who
>it is.

I agree.

>So I see the WG needed to resolve two independant issues, the first
>is how to name a sub-unit at login, the second is how to define
>the service level name that resolves into an IP/port and what the
>resolver is.  I think a URL is overkill for the former but might
>be appropiate for the latter.

I don't see the reason at all for the first.  Sub-unit are SCSI LUs and
discovery and naming conventions for those are already established by T10,
SAM, SPC-x, etc.  The only issue that I believe this WG needs to resolve is
the mechanisms for the establishment of a connection between initiator and
target (and whatever authentication/security features are required on top
of that).

*Please lets try to stay away from LUNs, LUs, etc. in this context.  They
aren't relevant (IMO).*

I've tried to say in other posts (CONNECT thread) that I think the
naming/discovery issues are in two places.  One is a specification of the
datum needed for an initiator to establish a (TCP/)IP connection to the
target.   The second, independent issue is where that datum comes from
(i.e., the management/discovery/namerserver function).

If every iSCSI target has an ipaddress:port valid *wrt a given initiator*,
then that suffices as the "datum".  If not, then two levels of addressing
are needed (as I suggest in the CONNECT thread).

I'm trying to stay out of the second issue but.. There are many approaches
to this and they all rely on some established infrastructure (an
implementation).  One such solution is that each initiator stores the
target datum refered to above locally and it gets put there manually by
humans typing at the keyboard.  *I don't advocate this as a good thing* (so
don't start flaming), but it is ONE possible implementation.  Others
solutions use LDAP or similar directory services (like changes to DNS or an
additional service like ScsiNS).   Others on this list can address those
choices.

In short, can we split this into two independent questions:

1) what datum does an initiator need to establish the IP connection to the
target?

2) where can an initiator get that datum?

Jim Hafner