Re: working draft: comments to 0706

[Matt Wakeley <matt_wakeley@agilent.com>]

Comments to draft 0706:

Section 2.2.2:

The "retry" bit is introduced here, but it's use is not really described
anywhere else in the document.

"NOOP" commands are introduced, but these are typically described as "NOP" or
"NO-OP".

Section 2.2.6:

Don't forget to describe how an IPv6 address would be used.

Section 3.2.6:

You need to change "40-byte" to "48-byte".

Section 3.2.7:

The length of the CDB algorithm should be:

if (I==0) then
  cdb_length = 16+ length
else
  cdb_length = 16

This also means that the new "object oriented" CDBs will never be able to have
immediate write data.

Section 3.3

In the reserved area bytes 40-47, the "SNS_LEN" and "RSP_LEN" fields should be
included, as they are in FCP-2.

Section 3.3.3

I don't understand the text in this section.  Typically, MaxCmdRN will be
greater than the current value of MaxCmdRN, and certainly greater than
ExpCmdRN.  Was this a cut-and-paste error?

Section 3.4

ExpStatRN should be at offset 28 to be consistent with the other message
headers.

Section 3.10

The Data Packet for READ should not have the Transfer Tag field (should be
reserved).

Section 3.17

The description and use of the "map command" needs a lot more work.

Section 6.1

RFC1122 says that TCP checksum checking is *required*.  So the "weakening" or
"nullifying" of the TCP checksum text must be removed.

This is another bummer with TCP - audio or video distribution streams will not
be able to make use of iSCSI with TCP due to the retry mechanisms of TCP.

That's all for now...

-Matt Wakeley
Agilent Technologies