TI 99/4A initially presented the "wafertape" as an option
(block-addressable-continuous loop).
And, as mentioned, the stringy floppy, somewhat similar to the later
wafertape and somewhat popular with the TRS-80 Model 1 crowd.
The Coleco ADAM used a block-addressable tape, but I don't know what the
read-after-write capability of the HE format tapes was.
There's no reason that a 7- or 9-track tape could not be used as a
random-access device, provided an adequate IRG was allowed for.
Implementing read/write random-access filesystems on tape presents some
significant issues.
The first is the read-after-write physical block length problem, already
discussed.
The second is that extending files gets to be really interesting, unless
file extensions get appended to the end-of-information point.
The third is that it's pretty much impossible to deal gracefully with
tape block flaws that develop some time after the block has been
written. In particular, the usual 9-track method of erasing over the
bad spot and then writing a new block would create a problem by
overwriting already-written data.
Streamer drives are singularly awkward (although as I mentioned, one of
the QIC vendors implemented a "virtual disk" system for the PC (Archive,
Seagate...I don't recall). Looking at the HP 35480 DDS OEM manual shows
that the drive has to construct a new group from old data as well as
new, write a temporary copy, then back up and attempt to rewrite the old
data with the new. Ugly stuff.
--Chuck