Path: utzoo!attcan!uunet!seismo!sundc!pitstop!sun!amdcad!ames!ucsd!ucsdhub!hp-sdd!ncr-sd!ncrlnk!ncrwic!encad!entec!jlohmeye
From: jlohmeye@entec.Wichita.NCR.COM (John Lohmeyer)
Newsgroups: comp.periphs
Subject: Re: Re: Questions on SCSI device interfacing...
Keywords: SCSI Termination
Message-ID: <295@entec.Wichita.NCR.COM>
Date: 11 Sep 88 16:07:47 GMT
References: <8551@cup.portal.com> <4740003@hprndlb.HP.COM>
Reply-To: jlohmeye@entec.UUCP (John Lohmeyer)
Organization: NCR Corporation, Wichita, Kansas
Lines: 37

In article <4740003@hprndlb.HP.COM> kc@hprndlb.HP.COM (Kurt Chan) writes:
>> The best thing to do is have all initiators (host adapters -- there may be
>> more than one) provide TERMPWR through a diode and a fuse.  This is what
>> SCSI-2 requires...
>         ^^^^^^^^
>
>Do you recall the reason for this requirement?  Bob and I were discussing a
>solution to one of our system problems that involves making sure one and only
>one source of TERMPWR exists on the bus.  
>
>By taking TERMPWR away from other initiators (removing the fuse) we have
>more predictability in terms of which devices experience RST on a cable
>disconnect, since the termination network will ground all the lines, including
>RST, when TERMPWR is lost.  
>
>By keeping other initiators (yes I know, even other hosts) from providing
>TERMPWR, our on-line-replacement procedure is a bit easier.
>
Yes, I remember the reason for the requirement -- it was to insure that a
multi-initiator system could continue to run even though one or more devices
lost power. In such a system, the initiators are all equal peers. It sounds
like you have a system where one of the initiators is more equal (a master).
If it is off, then the SCSI bus is not usable by the other initiators.

If the standard were revised to consider your application, then the requirement
would have to be that initiators provide a facility to _optionally_ provide
terminator power. This means that either a strap be added or the fuse must be
in a holder (as opposed to soldered in). I'm sure there would be some 
disagreement to this idea, but it might be accepted.  A very similar idea was
shot down last month to require targets to provide optional terminator power.

Of course, you can always just continue to ignore this part of the standard.
But if you want the standard revised, then you need to go through the hassle
of bringing a proposal to  the committee.  Call me at 316-636-8703 and I can
give further advice on how to do this.

John Lohmeyer    jlohmeyer@Wichita.NCR.COM    SCSI BBS: 316-636-8700