Compression may or may not be related to density (which is what the
bits 2/3 are to select for as well as compression- you could modify).
It's possible that a future enhancement would be to use an extra device
bit to select compression or not.

Currently, the use of mt is preferred. A tape has to be inserted
because a mount session determines the lifetime of the compress/nocompress
option. A mount session is determined on the open of a non-control mode
device. In any case, enabling/disabling compression doesn't work for
a large number of drives when you don't have media inserted (for example,
some DATs require the tape inserted because they encode on the media
itself whether it's compressed mode or not- and because once you write
a DAT you're stuck with that format until you degauss the tape has
to be inserted etc....)...

What I suggest you do on this is to file a send-pr/rfe to ask for
a 'compress/nocompress' bit to be reserved for tapes and then the
requisite work in the tape drive (and the MAKEDEV update and man
page update) can be done that will attempt to honor the intent.




On Tue, 27 Oct 1998, Simon J. Gerraty wrote:

> The consensus from my amanda question was to disable compression on
> the tape drive and let amanda do it....
> 
> So I'd like to setup a tape device that will always turn off
> compression.
> 
> The man page for st(4) says:
> 
>      There are several different `operation' modes. These are controlled by
>      bits 2 and 3 of the minor number and are designed to allow users to easi-
>      ly read and write different formats of tape on devices that allow multi-
>      ple formats.  The parameters for each mode can be set individually by
>      hand with the mt(1) command.  When a device corresponding to a particular
> 
> does this mean that I can create say:
> 
> crw-rw----  1 root  operator  14, 5 Oct 26 10:44 /dev/nrnst0
> 
> and use mt -f /dev/nrnst0 compress 0
> 
> to associate no-compression with that device permanently? (until
> next reboot).  The fact that mt fails unless there is a tape in the
> drive suggests perhaps not.
> 
> Can this be encoded in the quirks table?
> 
> Otherwise I guess I can hack amanda's pre-dump checks to do the deed.
> 
> --sjg
>