Questions on Reading from Exabytes made on a VAX, on a Unix system.

[Shigeki Misawa]

	Hi. I tried posting this before, but I must have done something
wrong, so I am trying again.

	In the near future, I will be attempting to read Exabyte tapes
written by a VAX and/or a custom multiprocessor, on a Unix box. Since
the system reading the tapes is different from the system writing the
tapes, I trying to determine what problems I will encounter. I hope that
people on the net could help me with a couple of points.

1) First, I expect that the exabytes will essentially be black boxes
that accept a stream of bytes which are stored in some local (i.e. inside
the black box) , as well as some control information like flush the
buffer to tape.

2) I expect that if no commands are issued to the exabyte drive, an
incoming stream of information will be locally stored until the
local buffer is filled, at which point the local buffer is flushed
out to tape.

3) I would expect that the buffer flushing would be transparent to the
operating system.

4) I might expect that the drive might put some information at the
end of the buffer or put a physical gap of some sorts on the tape
at the end of a buffers worth of information.

5) I would expect that the drive might also insert information relating
to an end of tape should it be commanded to do so by the computer. what
this end of tape mark might signify is not clear.

6) I would expect that the operating system of the computer would be
responsible for putting any structure on the stream of data stored
on the tape. This includes "record structures", "file structures", and
"directory" structures.

7) Point 6 would mean that within the users data that is stored on
tape, additional operating system information will be interspersed
to mark "end of records", "end of files", etc.

8) Point 7 would tend to imply that a tape written on a VAX may not
be easily read on a UNIX box, unless these operating system artifacts
were standardized.

9) From my minimal understanding of the Unix OS, I would expect that I
would be able to get around the problems raised in 7 and 8 by reading
directly from the tape device (however this is acomplished. I would
expect this would be accomplished through the files associated with
mounted device) one character at a time and them, if I know what the
various OS artifacts are, I could selectively filter them from the
data stream.

	I would expect that some, if not all the expectations stated
above are either wrong or not quite correct. I would appreciate any
information or references that would help me clear up any misconceptions
I may have about this system.

	As a final note/question, I have been told that in 9 track
tapes, a computer may specify a variable length physical record. I was
told this to mean that the computer can write out a stream of bytes
out to tape, at which point a physical end of record mark is written
out to tape. Let me postulate that the tape drive system is only
capable of reading data from tape only in chunks delineated by this
end of record mark. If this were true, I would expect that I would be
able to generated 9 track tapes on one machine that would not be
readable on another machine. I would expect that such a situation
would occur if this record is read into a buffer that is too small or
cannot be read out by the computer system fast enough. Is this
possible?

	I would like to thank you in advance for any tidbits of information.

Thanks.

Shigeki Misawa UCB Physics Department.