What do you know about tapes...?

Carl S. Gutekunst csg at pyramid.pyramid.com
Fri Aug 5 17:29:19 AEST 1988


In article <254 at execu.UUCP> dewey at execu.UUCP (Dewey Henize) writes:
>The question is, though, are we going to run into this same level of [tape]
>incompatability with various *nix machines?

Generally, one of the beauties of Unix tape handling is its incredible sim-
plicity. And, the biggest headache is its incredible simplicity....

A Unix raw tape device is very much a "you asked for it, you got it" device.
You can write anything you want, byte-for-byte, in any size block, up to the
limit of the buffers on the tape controller. Inter-record gaps get written
after each write(2) call, and an EOF mark gets writen when the file is closed
(if you had it open for write). You just can't get simpler than this. 

The other edge of the sword is that if you want to do anything "standard,"
like an ANSI format tape or plain ol' 80-block-40, you have to write an
application to do it for you. The operating system won't help. The Unix dd(1)
utility is supposed to assist in this sort of thing, although I've found that
for any given new tape format, I have to write a C program to manipulate it.
Someday someone write a really good tape handler for UNIX, and get rich....

Differences are found in the controller's buffer size, which determines the
maximum record size. You need at least 10K for the tar(1) utility to work, and
the smallest I've seen on any Unix system is 20K. Pyramids allow up to 30K
with the old IOC tape controller, and 64K with the TPE. Suns write up to 61K
blocks on their Archive 1/4" tape; I dunno about their 1/2" 9-track interface.

The other difference is how the physical end-of-tape is handled. Most UNIX
systems botch this horribly, something like writing a record across the gap
(good) but not allowing you to read it (bad). If this is important, ask your
salescritter. (I am told Pyramid does this "right," meaning the same as most
commercial DP machines, but I don't recall what "right" is. I do know that
multi-volume cpio works flawlessly, which is not true on many other Unix boxes
I have used.) 

<csg> 



More information about the Comp.sys.pyramid mailing list