Comments on FCP Rev8b
Lansing J Sloan
ljsloan at ocfmail.ocf.llnl.gov
Fri Jul 15 12:57:51 PDT 1994
July 15, 1994
SCSI and Fibre Channel,
A few comments on FCP Rev8b follow. I sent additional
comments, mostly editorial, directly to Bob Snively.
The comment numbers are unchanged from the
-- Lansing Sloan
ljsloan at llnl.gov
#037 (T) Comment on FCP 18.104.22.168
The paragraph about "Ordered_Q" says a request for
Sequential Delivery is necessary to ensure correct ordering.
Since frames can be re-transmitted after being busied or
rejected, a certain amount of out-of-order delivery can
effectively occur even if Sequential Delivery is requested.
Therefore it is far from clear why requesting Sequential Delivery
suffices to ensure correct ordering. If it does suffice, a NOTE
would be valuable if it explains why resends after busies or
rejects cannot cause harm. If it does not suffice, I'm not sure
what should be said.
#070 (T) Comment on FCP Rev8b 5.4
FCP (and other mappings of SCSI to specific media) should
specify the format to be used for Source SCSI Identifiers and
Destination SCSI Identifiers in such places as the segment
descriptors in the parameter block for the COPY command. I have
not found such a specification in FCP.
If there is another standards document that contains the
specification and makes it normative for FCP, that's acceptable.
#071 (T) Comment on FCP Rev8b 5.4
For FCP, the validity and value for Process Associator
values should be included in SCSI Identifiers that occur in such
places as the COPY command's segment descriptors. Alternatively,
FCP should specify that SCSI systems that require Process
Associators may be unusable as SCSI Sources and Destinations in
commands such as COPY.
More information about the T10