Response to SPI-4 proposal 01-251 (continued)

Richard Moore richard.moore at qlogic.com
Fri Aug 31 11:33:35 PDT 2001


* From the T10 Reflector (t10 at t10.org), posted by:
* Richard Moore <richard.moore at qlogic.com>
*
Yesterday I wrote about the negative effect of this proposal. I should also
point out that
this proposal has little if any positive benefit.

First, the proposal does nothing about SPI-3. So the initiator behavior that
is considered
offensive by this proposal will still be present in SPI-3 environments even
if this proposal
is passed.

Second, the proposal only changes the behavior for data group transfers. It
has no effect
on information unit transfers. To my knowledge, nobody is developing SPI-4
hardware without
packetized, so packetized will be the default behavior in pure SPI-4
environments. If someone
elects to mix SPI-3 and SPI-4 devices on a bus, then the possibility of
encountering a
combination of hardware that would cause this problem is still there even if
this proposal is
passed.

Third, the method of counting that I described for DT transfers is actually
quite consistent
with the method that SPI-4 *requires* to be used for counting in paced
transfers (see 10.7.4.4).
The only difference is that paced requires one edge (not two) of the
non-clocking REQ or
ACK signal to every two edges of the clocking signal. With this proposal we
would be
banning a behavior (that was previously allowed) in one mode, while
requiring very similar
behavior in another mode. While technically this is only a minor drawback
(in than that it
creates a separate case in several parts of the logic), it does reflect a
philosophic
inconsistency. It wouldn't surprise me if there are other vendors using an
approach similar
to ours.

The proposal only addresses an issue in DT data group transfers. It will not
have any effect
on packetized DT transfers or on paced transfers, so it will not be of any
value in pure SPI-4
environments. It won't retroactively change SPI-3 devices either. Given the
negative effect of
forcing at least one vendor (I can't speak for others) into another turn of
silicon, is there
really a need for a measure that offers so little real benefit?

Richard Moore
QLogic Corporation

*
* 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