Xref: utzoo comp.periphs.scsi:571 comp.sys.dec:3321
Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!cs.utexas.edu!sun-barr!newstop!sun!wonky.Eng.Sun.COM!mjacob
From: mjacob@wonky.Eng.Sun.COM (Matt Jacob)
Newsgroups: comp.periphs.scsi,comp.sys.dec
Subject: Re: Is DEC RZ23 real SCSI?
Keywords: Does DEC use some odd proprietary controller query scheme?
Message-ID: <136590@sun.Eng.Sun.COM>
Date: 4 Jun 90 16:11:10 GMT
References: <1990Jun2.213812.13909@spock.UUCP> <5454@crltrx.crl.dec.com> <136580@sun.Eng.Sun.COM> <5461@crltrx.crl.dec.com>
Sender: news@sun.Eng.Sun.COM
Organization: Sun Microsystems, Mt. View, Ca.
Lines: 56

In article <5461@crltrx.crl.dec.com> jg@zorch.crl.dec.com (Jim Gettys) writes:
>>>Experience has shown, however, that SCSI is far from the
>>>tightest specification in the world.  When we started hooking up
>>>other SCSI devices to DECstations, there was about a 50% chance
>>>they would work (to my knowledge, it our case, it was always microcode
>>>problems with the SCSI peripherals; in your case, it could be either
>>>the drive or your controller); this percentage has
>>>been increasing with time, as people get their implementations
>>>conformant to the specification, and the specification gets better
>>>understood by the implementors  (moral is:  KISS...  SCSI violates
>>>this rule...).
>
>>Weak, buddy. Real weak. This would have been a truer statement 3-4
>>years ago, but not for the last couple of years.
>
>>
>>-matt jacob
>> Sun Microsystems
>
>Boy, it looks like you don't like to read....

First of all- apologies for being so acerbic. And yes- I didn't read
it as carefully as I might- I was dialed in from home (23 line tty
pages don't latch up in my mind like they should...*sigh*)

>
>If you read my statement above, you'll see that I was talking about several
>years ago, and noted that the percentage has been increasing with time.
>And it was a very small sample. (statistics of small numbers....)
>
>Most of the problems we saw were of the form:
>
>Disk firmware has timing window that only a fast machine can hit...
>Drive might generate error or hang after pounding on it for some hours.
>
>Getting solid implementations of complex specifications is always
>difficult.

True.

>
>Are you claiming that Sun makes slow machines and slow SCSI implementations
>that don't see that kind of problem?  :-).  Sun certainly didn't
>build such fast machines 18 months ago... :-).
>				- Jim


Speed of the CPU or the Host Adapter is largely irrelevant. A good and
careful s/w implementation of the transport layer (Host adapter driver)
can undertake various strategies to make up for deficient hardware.

-matt

p.s.: Let me present my credentials, which you can then use specifically 
defame me with: I am 99% responsible for the SCSI S/W on the SparcStation-1
(and its clones, e.g., SLC, etc.).