Sorry Joe, but on every HP 3000 I saw with a card reader (about eight, all
in the state of Virginia), it was LDEV 5, following the standard set by
FORTRAN of unit 5 being the card reader, unit 6 being the printer.  I
believe that's also where the standard of LDEV 7 being tape came from.

While the HP 3000 card reader (HP 2893A) was usually a spooled device, it
could also be run "hot" (unspooled).  There were a number of programs
floating around which ran the reader hot and parsed the "!JOB" card
themselves, looking for, evaluating, and removing parameters which were
meaningful only to the reader program, then writing the individual jobs to
temp files, and streaming the temp files.

Those reader programs were used by academic sites to do such things as
add/change ";TIME=", ";PRI=", and ";INPRI=" parameters to the job card
based upon the account the job would run under, the user name the job would
run under, or whatever criteria the management deemed necessary.  In our
case only administrative jobs and faculty jobs ran in the DS queue -
student jobs ran in ES queue.  All faculty and student jobs had their
";INPRI=" adjusted according to the value they had placed in their ";TIME="
parameter.  Ten seconds or less and you got INPRI=8, 11 to 20 seconds and
you got INPRI=7, etc.  If you wanted CPU minutes of time, you got INPRI=1
and the operator had to intervene for your job to run.

The Series II didn't have much main memory, and it sure didn't have much
disc space, so often you HAD to run the reader hot so you could read and
process the deck (student registration, cafeteria style, as others have
mentioned, with tray after tray after tray of cards).  Spooling was not an
option, as you then had to have twice as much disc space.  Also, if (in the
early days) the job blew up, you could go back to the card reader, look at
the last few cards read, and perhaps see the cause of the problem.

There were also card reader/punch devices for the HP 3000, but the only one
of those I saw was connected serially to a MUX port (2400 baud) and so had
a twenty-something LDEV number.  The reader/punch was very slow, very
unreliable, and was not recommended by the computer center which owned it,
while the dedicated reader (HP 2893A) was 600 CPM, quite reliable, and
well-liked.

The HP 2893A differed by one card (which made it much smarter) from the HP
2892A which was the card reader on the HP 2000 systems.  Add the one card
to the HP 2892A and you had an HP 2893A (which happened one time when HP
ran out of rubber feed wheels).  Both the HP 2892A and HP 2893A were Data
Products readers with HP nameplates on them, and both were very reliable as
compared to other card readers.  The most common repair was replacing the
rubber feed wheels.  They used the differential interface, and their cable
looked just like that of the old HP 2617 printers, so the big problem with
moving cables around under the floor was making sure you had the right cable.

We had 1200 CPM readers on our CDC Cyber 720, but they were down more than
they were up, and had so many mis-feeds and read-checks that they were
almost useless.  We ended up using the HP 2000's 600 CPM card reader (HP
2892A) to read in batch jobs for the Cyber, using a BASIC program to spool
them to the HP 2000 disc (HP 7920), then using another BASIC program to RJE
the batch input files to the Cyber.  Remember, the HP 2000 was a BASIC-only
interpreted machine, which in its largest configuration (32 user) had only
64K bytes of RAM in a 21MX processor for the system processor, and 64K
bytes of RAM in another 21MX processor that served as the IO processor
(terminals, printers, RJE datacomm).  BASIC programs (and their data, and
their control blocks, and their file buffers) had to fit in just over
10,200 words (16 bit) of memory (10,212 rings a bell).  Despite all the
BASIC programs involved, the slower read speed (600 CPM vs. 1200 CPM), and
the 9600 baud communication line between the HP 2000 and the Cyber 720, we
could get more than twice as many cards read and processed by the Cyber 720
per hour by reading them through the HP 2000 than by reading them through
the Cyber's own 1200 CPM reader.

Anyone remember PTAPE?

John

At 5/19/99 06:09 AM , Joe Geiser wrote:
>Hint from a long-time user:
>
>The card reader was used, just like it was on older IBM platforms such as
>the 360 and 370 models, to submit batch jobs.
>
>The answer did come out earlier - it was LDEV 10
>
>Cheers,
>Joe


--------------------------------------------------------------
John Korb                            email: [log in to unmask]
Innovative Software Solutions, Inc.

The thoughts, comments, and opinions expressed herein are mine
and do not reflect those of my employer(s), or anyone else.