> Typically, tapes are designated as [e][n][r]CtU, where
> 
> 	C	is the controller type; 'w' for an IDE tape; 's' for a
> 		SCSI tape; 'x' for a tape on a Xylogics controller
> 		(if you happen to be stuck with a SPARCheater or Sun 3).
> 		I'm sure others will be dreamed up in the future.

wt's are not IDE, they're tapes on a qic-02 or qic-36 interface. The
`w' in both cases seems to come from "western digital". I guess with
the advent of atapi, we might start seeing IDE tape drives
someday. blah.

> 	t	Constant.  Stands for 'tape'.
> 
> 	U	Unit number.  0 is the first tape device found, 1 is
> 		the second, etc.  It's probably good to nail down the
> 		targets and units in your kernel config.
> 		Historically, the density was controlled by the unit
> 		number, adding 8 to the unit for each level of density
> 		up to a maximum addition of 24.

Does netbsd still follow this convention? (I hate the convention, btw,
but I do have one tape drive i might at some times like to access in
other-than-default density)

> 	[e]	This is a new precedent and I'm not sure what it means
> 		(yet).

"rewind and eject on close" NetBSD (perhaps *BSD) is the only system
I've seen with this feature. It's nice, since it eliminates a race
condition between the completion of your dump program and an "mt
rewof" where some other process can open the tape drive and overwrite
(or read, for the security-conscious) your backups. This is sadly
necessary since Unix doesn't have very advanced tape handling.

> Unless I Missed Something [TM], tapes don't reference wt0[a-h] because,
> wouldn't you agree, it's silly to restrict a tape to eight files, one
> of which encompasses the entire tape...?

Actually, some tapes, notably DAT tapes, *can* be partitioned, albeit
only into two sections. It would be really nice to have an interface
to this feature; one particularly nice application of this would be to
put a tape label and index in the first partition, and data in the
second. The index can (theoretically) be updated after the data has
been written.

--
Dan.