RE: [iSCSI:] Logout request -- reason? Correction

[pat_thaler@agilent.com]

Luben,

I was refering to your statement "I was under the impression that byte 1 in
BHS is a flags byte -- i.e. bit values." The BHS is what is defined in the
template which says it is opcode specific, not a flags byte. If you want to
be able to count on it being a flags byte, than that additional information
would need to be added to the template.

Some PDU types have 2-bit fields in the first byte (e.g. CSG and NSG), some
have 3-bit fields. All these values are described as numeric values (e.g.
CSG/NSG is described as having values 0,1 and 3). I don't see how a 2-bit
value is okay in that location but a 7-bit value is not okay. I also note
that it appears that only 3 bits of the Function field are actually used
currently as the 

As an aside, I don't see that the spec describing a 2 to 8 bit field in
decimal, hex or binary makes any difference to what the hardware/software
does in processing those bits when the field is just a mapping of a type to
a set of bits. It would be different if one was expected to do arithmetic
operations on the field. The existing uses of the first byte are all
mappings of a meaning to a bit pattern regardless of the representation of
that bit pattern in the field. Of course, some day it could be used for a
real number.  

It isn't a sound architectural idea to assume that fields which are OpCode
specific will have a consistency of format across opcodes. If it isn't in
the template then future OpCodes may be created that use that byte in
different ways. 

I understood what you meant and I don't have to have the same reason to
object as other people.

Pat

-----Original Message-----
From: Luben Tuikov [mailto:luben@splentec.com]
Sent: Tuesday, May 28, 2002 11:24 AM
To: pat_thaler@agilent.com
Cc: ips@ece.cmu.edu
Subject: Re: [iSCSI:] Logout request -- reason? Correction


pat_thaler@agilent.com wrote:
> 
> Luben,
> 
> I don't see where you got that impression. The BHS layout shows bytes 1
through
> 3 as opcode specific fields. The kinds of opcode specific fields will be
different
> depending upon opcode.


Pat,

I KNOW about the BHS template.

If you had read my mail carefully you'd notice that
I mention that putting a bit value _and_ a numeric
value in the same octet is not a sound architectural idea.

Then you'd deduce that since the F bit is used in _that_
PDU, it'd be a good idea to leave the rest 7 bits
be bits (thus making one octet) and use the _next_ _whole_
_empty_ _reserved_ octet for the status/response code.

Certainly everyone else understood what I meant.

And the objection was _not_ your argument, but
the fact that we are running out of time.

-- 
Luben