🚀 go-pugleaf

>>The particular spec involved the time from reset to ready.  The
>>Commodore will not autoboot if that time is greater than (as I recall)
>>four seconds.  The Miniscribe appears (from my informal measurements)
>>to take about 5 seconds if warm booting with the head on track 2,
>>about 8 seconds if warm booting with the head near the middle track,
>>and about 14 seconds from power up.

> The problem is not the time from RESET to ready, but the time from RESET
> to response to commands.  Seagate drives have this same problem.  When the
> drive is spinning up from cold start, it does not respond to selection at
> all.  This looks to the driver software as if no drive is hooked up.  The
> solution is to try for selection, if it fails wait a second and try again
> up to a timeout limit.

That is fine, except that the COMPETITOR's computers will successfully boot
from these same disk drives.  The customer does not say "what a pity my
computer works with this drive when the drive is out of spec".  The
customer (me) says "I am not happy that the Amiga fails to boot with this
drive.  The other computer can, why not the Amiga?"

> However, SCSI drivers have to check 7 distinct
> SCSI bus addresses to see if drives are hooked up (there is a way under
> the new RDB scheme to shortcut this check, but that`s not important here).
> If the driver were to assume that all drives took at least 5 seconds to
> spin up, that would be 7*5 = 35 seconds of dead time on a cold boot.

Not true.  The Amiga only attempts to autoboot from the first SCSI.
So if the delay is extended to 10 seconds, the max wait would be 10
seconds.  Notice that a customer would rarely see this delay.  The
algorithm looks something like:
  if there is a floppy
	boot from it			no delay
  else if there is a ST506
	boot from it			short delay?
  else
	loop once a second
	  did the SCSI respond?
	    yes: boot from it		delay of 1 to 10 seconds depending
					on how fast the drive responds
	    no: prompt for a floppy	<-- max delay slightly over 10 sec

>  Since
> the majority of drives respond correctly, it does not make sense to
> compromise systems with good drives for the few with bad ones (bad meaning
> badly behaved).

Compromised!  The only delay this affects is the time to display the
hand, and that delay only occurs if the following is true
  there is not a floppy in the drive
      AND
  there is a 2090A card
      AND
    there is not a bootable ST506 drive
	AND
    there is not a bootable SCSI drive

Realistically, how often is that going to happen?  That low probability
is what you are balancing against the cost of having a number of drives
which will not boot with the current firmware.  It looks to me like about
30% of hard drives will not boot because of this spec.  The 2090 users's
guide (with the addendum) lists 8 that do and 4 that don't autoboot.
If I include my data point (Miniscribe 3180S) that works out to 5/13=38%
of known SCSI drives which will not autoboot.

>  This problem has been (sort of) addressed on the A590 and
> A2091 by having a "Seagate" jumper on the board.  If this jumper is
> installed, much longer timeouts are allowed when cold booting the system.
> 	steveb@cbmvax.UUCP (Steve Beats)

I bought the A2090A because it was supposed to autoboot.  Now it appears
I need to buy the A2091 to autoboot.  I hope you understand my frustration.
--
	Charles Brown	charles@cv.hp.com or charles%hpcvca@hplabs.hp.com
			or hplabs!hpcvca!charles or "Hey you!"
	Not representing my employer.