New and updated proposals for discussion (vote?) in CO.
Jim Hafner
hafner at almaden.ibm.com
Wed Jun 20 09:01:58 PDT 2001
* From the T10 Reflector (t10 at t10.org), posted by:
* "Jim Hafner" <hafner at almaden.ibm.com>
*
Folks,
I've just uploaded three proposals to the web site. Two are new and one is
a revision. I am hopeful that all of these proposals can get approved at
the next T10 meeting in CO, so if any of these are of interest, I would
politely ask that you read them before the meeting (and make comments and
suggestions before then, if you want).
Here's the summary:
T10/00-425r4: "Long Identifiers in SPC-3, SAM-2, SBC-2 and other XOR
issues". The proposal attempts to solve the problem of long names and
identifiers. It covers a number of minor things particularly with respect
to the XOR commands. The primary focus of this is the new CHANGE ALIASES
and REPORT ALIASES commands. This proposal has been revised based on a
number of comments from the last meeting (that went into rev3) and on a
number of additional comments received since then. There are no
fundamental changes to the model since rev2, with the exception of the
behavior of a CHANGE ALIASES command in the presence of other commands that
are reading the alias list.
T10/01-181r0: "Access Controls TransportIDs for SBP, SRP and iSCSI" (joint
with Rob Elliott). This proposal fills in the missing gaps in the Access
Controls (approved) proposal for these additional protocols. There
shouldn't be anything controversial here. SBP uses the EUI-64 of the port,
SRP uses the 16byte Initiator port identifier and iSCSI uses the iSCSI
Name. The only important things to note are:
(a) We swapped the "protocol identifier" value originally assigned to SPI
and FCP in their formats to match the identifiers assigned already in Table
165 of SPC-2 rev 19. Hopefully this doesn't cause any implementors any
problem.
(b) Because iSCSI's names are potentially long, we had to remove the
restriction that TransportIDs fit in 24 bytes. This caused some extra
bookkeeping in the "logs" stuff, but nothing dramatic or that would break
existing implementations that don't deal with iSCSI.
T10/01-182r0: "SAM-2 device and port names": This proposal adds the notion
of "name" to devices and ports. There is a lengthy discussion (soapbox)
concerning approaches to these concepts in addition to the specific
proposal. The proposal itself should not be very controversial (I've tried
to take the path of least resistence). Essentially, names are optional
both for devices and for ports. Protocols can define requirements for
names. Devices can have more than one name if they live in different
protocol domains (because each protocol may have different requirements and
formats). Ports can have at most one name (because they are bound to a
protocol).
Jim Hafner
*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org
More information about the T10
mailing list