customer perspective RE: Mixed bus widths

Lohmeyer, John JLOHMEYE at
Wed Oct 23 15:48:00 PDT 1996

* From the SCSI Reflector (scsi at, posted by:


The discussion thread you forwarded does not say why the CD-ROM did not   
work (perhaps it was unknown).  My personal experience with CD-ROMs is   
that they had a lot of software interoperability problems that have been   
largely addressed in the newer operating systems.  Since most of these   
problems were due to devices not following the SCSI-2 standard, it is   
rather unfair to blame the standard or the standards process.

In defense of "creeping" improvements I'd like to point out that the vast   
majority (if not all) of improvements to SCSI over the last decade have   
not been detrimental to interoperability with well-designed products.   
 That is, if the product does something sensible (such as returning   
MESSAGE REJECT or ILLEGAL REQUEST) when presented with unknown messages   
or commands, then the other device will know to back off and not use the   
new message or command.  Unfortunately, not all products are   
well-designed or well-tested; this means that some products will fail to   
interoperate with other products designed to the same standard.

If the standards committees rejected all proposals that did not improve   
performance by 2 orders of magnitude, then SCSI would be dead and   
replaced by something else.  This would guarantee that the new interface   
would not interoperate with the old product.  Seems like we would move   
|from a situation that usually has interoperability to one that is   
guaranteed to not interoperate.

I do agree that the standards process is not perfect.  It is indeed much   
easier to say "yes" instead of "no".  But those who do not participate   
would be amazed at how often the answer is "no" or "no, do it this other   
way that is already available".  Mature standards have a lot more   
resistance to change than those in their infancy.  Manufacturers learn   
rather quickly that adding an option that one company wants will soon   
become a requirement in their customer's purchase specification.

Incidentally, the original SCSI standard (X3.131-1986) documented   
transfer rates of up to 4 MBytes/sec on a 8-bit bus.  Ten years later,   
the draft SPI-2 standard (recently stabilized) documents 40 Bytes/sec on   
a 8-bit interface.  That is only one order of magnitude per decade.  I   
doubt anyone would even remember SCSI if we had waited until 2006 to   
improve it beyond the original 4 MBytes/sec.


John Lohmeyer                 E-Mail: john.lohmeyer at
Symbios Logic Inc.             Voice: 719-533-7560
4420 ArrowsWest Dr.              Fax: 719-533-7036
Colo Spgs, CO 80907-3444    SCSI BBS: 719-533-7950 300--14400 baud

From:  scsi-owner[SMTP:scsi-owner at Symbios.COM]
Sent:  Wednesday, October 23, 1996 2:31 PM
To:  scsi
Subject:  customer perspective RE: Mixed bus widths

* From the SCSI Reflector (scsi at, posted by:
* palasek at (Robert Palasek)
User/customer Perspective:
  A friend forwarded to me part of a discussion thread, copied below,   
was somewhat related to a recent problem.

  I have enjoyed SCSI and its benefits of being STANDARD for
a number of years.  Two weeks ago, when we bought a new desktop
computer, we weren't able to configure it because even though
our external cdrom devices were scsi, and even though the
computer had a scsi connector, they did not interoperate.

Marshal Rose explained one time that the reason the OSI standards, even   
they had the ability to solve problems that the internet is coming
up against just now,the reason they never proved useful to more than a   
was because the standards committee never had the resolve to
say NO, that when there occurred a substantive difference, instead
of making a hard choice, they said, "we'll have it both ways."
How diplomatic!

I would worry that the same problem is going to kill scsi.
This is a plea for you to work hard to reduce the entropy.

In a more narrow view, my personal feeling is
that if generation N+1 does not deliver two orders
of magnitude improvement over generation N, you
should just wait and bide your time until it does.
Don't change it incrementally, the creep costs
more than its real value.

Ususal disclaimer: I am not speaking for
my employer nor its sponsors.
Bob Palasek
P.O. Box 808,  M.S. L-303
Livermore CA  94551.

>>From: "Lohmeyer, John" <JLOHMEYE at>
>>To: "'SCSI Reflector'" <scsi at Symbios.COM>
>>Subject: RE: Mixed bus widths
>>Date: Tue, 22 Oct 96 10:20:00 MDT
>>Encoding: 31 TEXT
>>Sender: owner-scsi at Symbios.COM
>>Precedence: bulk
>>* From the SCSI Reflector (scsi at, posted by:
>>Greg Alvey wrote:
>> ----------
>>>I seem to remember that there used to be something in the FAST-20
>>>that said something like: "the bus width shall be constant" (at least
>>>single ended).... but ...
>>>After reviewing Rev 6 of FAST-20, it appears to have "gone away".
>>>I guess I missed that one!  Maybe someone can inform us as to "when &
>>>I wonder if this statement was moved to another document (SPI or SPI-2
>>Your memory serves you well.  The statement was removed near the end of
>>the Fast-20 development when the working group went into "permissive
>>mode" (this is when the marketing people start reviewing the draft
>>document ;-).  No one knew for sure that mixed widths would cause
>>problems, so the restriction was removed.  About the same time we
>>softened much of the wording surrounding cable length restrictions.
>> --
>>John Lohmeyer                 E-Mail: john.lohmeyer at
>>Symbios Logic Inc.             Voice: 719-533-7560
>>4420 ArrowsWest Dr.              Fax: 719-533-7036
>>Colo Spgs, CO 80907-3444    SCSI BBS: 719-533-7950 300--14400 baud
>>* For SCSI Reflector information, send a message with
>>* 'info scsi' (no quotes) in the message body to majordomo at

* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at

* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at

More information about the T10 mailing list