SBP-2: Isoc min_transfer_length

Stephen_Finch at notes.ssi1.com Stephen_Finch at notes.ssi1.com
Mon Jul 7 13:33:21 PDT 1997


* From the T10 (formerly SCSI) Reflector (t10 at symbios.com), posted by:
* Stephen_Finch at notes.ssi1.com
*


One of the return values in the create stream response packet
is the min_transfer_length field (see section 5.1.4.3).

This is interesting in intent, but very misleading as to its
usefulness.  Is this an absolute number?  Is this an average
number over some large/small set of ORBs?  What happens when
playing a file which contains multiple channels and you ask for
only one to be played?  One could meet this number and still
not be able to maintain the stream, because the amount of data
being read to gain access to the desired data could be large.
The same is true of Null blocks.

Also, this number could change radically based on the number
of logged in initiators, the number of streams that have been
created.

It appears that this field is advisory.  If the purpose is for
the device to guarantee that, if this value is met, that it will
never mess up the stream, then the device should put a maximum
value in this field.  If the device is going to be rated by vendors
based upon this field, then device should put the smallest number
possible assuming that nothing else is happening in the device or
on the bus.  Sounds like we have a lie in this field no matter
what we put in there.

Hmm...

No real answer to what it must be.  No formula to go by.  Basically,
this looks like a marketing thing.  Give the best number in the
direction that the client would want without giving him a number
that he would reject as being out of hand.

Unless I hear from a user out there that has a real need for this
field and can explain and document the formula, I'd vote to remove
this field.

My flack jacket is on.  Fire at will.


steve finch



*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at symbios.com




More information about the T10 mailing list