Re: wisdom of the ancients, was Architectural implications of locate mode I/O

On Wed, 03 Jul 2024 22:17:59 -0600, John Savard
<quadibloc@servername.invalid> wrote:

On Wed, 3 Jul 2024 23:03:24 -0000 (UTC), John Levine <johnl@taugh.com>
wrote:
>

That is pretty amazing.  Did NASA have a plan for reading
those tapes, or did they not realize that normal tape units
need record gaps, or what?

>
It would be interesting to find out the real historical answer to that
question.
>
But I can guess at a possible answer. This is before COTS became an
acronym. They felt they needed to record those tapes that way, and
they figured that reading them should be no big deal, even if existing
hardware didn't support it.
>
Another possible approach from the one described (which could
potentially lead to errors if data on the tapes was repetitious),
would be to connect two computers to the stream of data coming from
the tape drive, with them synchronously handing off the responsibility
for reading the data back and forth. I'm sure they could have thought
of that back then.
>
But they really shouldn't have used standard 9-track tape for this.
Something like DECtape, with random-access capabilities, would have
been more appropriate. Getting DEC to make a higher-performance
DECtape drive with vacuum columns may have been a problem, though.

Actually, even given the technology of the time, NASA engineers
probably figured that solving the problem was trivial on existing
computers. For example, like this:

1) Channels on IBM mainframes were programmable computers in their own
right, you could write channel programs. Thus, a Control Data 6600
with Peripheral Processors is not a requirement.

2) Since the channels are programmable, it should be easy to tell them
to do this when reading the tape:

Just read the data continuously into a circular buffer which is at
least three times the size of a normal tape record, ignoring
overwrites.

Issue an interrupt, to let the computer know a record's worth of data
is available, whenever a record boundary in the circular buffer is
passed.

3) The computer hands off the incoming records alternately to be
written as records with gaps to two other tape drives.

4) After this process finishes, take those two tapes, and turn them
into two tapes with the first and second halves of all the records,
instead of one with odd records and one with even records, through the
use of a computer with four tape drives. This is just standard data
processing stuff, which any computer that can to a tape sort can do.

John Savard