At 3:13 PM -0600 10/19/06, Richard wrote:
In article
<a06230902c15d8d43781e(a)[131.215.234.40]>0]>,
"John A. Dundas III" <dundas at caltech.edu> writes:
If on the other hand the files are more
important (than being able to
recreate the tape exactly), [...]
What would require an exact tape image, as opposed to files and their
contents?
Tapes where retention of attributes, block sizes, placement, etc.,
are just as important as (maybe more so than) the file contents. Two
things come to mind:
1) a simulator (SIMH, Charon, etc.) where you wanted to work with a
(virtual) tape device. Pat mentioned booting an OS tape; absolutely.
2) the ability to recreate the tape (either on the same media, TK50 I
think was this case, or other media 9-track, TK70, etc.).
If I copy files from tape to disk, I lose all blocking information
(there might be different block sizes on the tape for different
files). I won't be able to recreate the tape from the files
themselves without additional meta information. Boot tapes often
(sometimes?) contain boot data/code outside of recognizable file
headers or marks on the tape. I believe the early PDP-11 tapes were
this way. I can tell you that later PDP-11 (RSTS) tapes are
multi-format, i.e., the first few (maybe dozen) files are DOS-11
format, including the boot file, then the rest of the tape is written
in ANSI format. IIRC, DEC's Unix-11 V7m contained boot information
at the front of the tape, then several tar images separated by tape
marks on the remainder. A tape image is able to capture this sort of
information in addition to the file data itself.
John