PnP r0.95 comments
Jim McGrath
/G=Jim/S=McGrath/O=QMAILGW/PRMD=QUANTUM/ADMD=MCI/C=US/ at qntm.com
Tue Jan 11 17:10:02 PST 1994
Subject: Time: 1:15 PM
OFFICE MEMO PnP r0.95 comments Date: 1/4/94
Unfortunately I will not be able to attend the meeting in Colorado Springs this
week. However, I will be available next week at the SCSI meeting. I suggest we
might want to schedule a meeting during that time to resolve further outstanding
issues. I can also be reached at my office, and would be happy to discuss these
comments with anyone.
1. Foreword
Please add Quantum to the list of companies that initially worked on the Plug
and Play document (since we started the whole SCAM activity last spring).
2. Hardware (section 5)
It may be a goodness to require SCSI-3 compliance, but I do not see the need in
a standard that is primarily concerned with auto addressing and auto
termination. The SCSI committee is still trying to sort out what SPI compliance
means. If it is unnecessary, then I say drop it, or at least postpone its
implementation until the industry is better equipped (say late '94).
3. Connectors (section 5.1)
Do you really want to have no connector requirements for SCSI devices (i.e. bare
drives)? I suggest we focus on the 50 pin, 100 mil center devices since they
are by far the most popular. I hear people wanting to make provision for the 68
pin high density connector, but I strongly resist making that the only option
(we just started shipping those in 3.5" drives!). And specifying two connectors
WILL lead to more cabling confusion.
4. System configurations (section 6)
We are still requiring the user to buy and install the external (non host
adapter) terminator for the system. This is done at the time that the first
device is installed. Why do that and not simply use the terminator on that
device instead? The device can be shipped with the terminator removed, but
still available in the package. That device should be put at then end of the
cable for future expansion. This is no worse that putting an external
terminator at the end of the cable.
Also, switchable terminators are allowed on the host adapter, but why not the
devices themselves? Devices should be allowed to have switchable terminators.
If you want, they can be specified as turned off by default, to be activated by
a software protocol (to be defined at a later time). In the meantime we can use
the external terminator devices as currently specified. But a path should be
cleared for the day of switchable terminators totally under software control
that should ultimately prove easier and cheaper to use.
In general, I think the terminators are still a real issue (unlike auto
addressing). The only solution I really like is switchable terminators with a
software protocol.
More information about the T10
mailing list