Re: devfs

Grant R. Guenther (grant@torque.net)
Tue, 13 Jan 1998 18:31:16 -0500 (EST)


I haven't been following this thread too closely. I hestitate to complicate
things any further, but I should at least mention some of the issues that
are presented by the parallel port IDE devices. I suspect that PCMCIA
and potentially USB devices will also need some consideration.

Parallel port IDE devices have all the same variables as normal IDE and
ATAPI devices:

- type: IDE disk, ATAPI disk, ATAPI CD-ROM, ATAPI tape, ATAPI generic
- master/slave
- lun (for some ATAPI devices)

But, they also have some additional variables of their own:

- the parallel port to which they are attached
- the "protocol" used
- a "chain ID"

The port is fairly obvious, the other two require a bit more explanation.
Most of these adapters are designed to be inserted into the cable path
between the parallel port and a printer. They "wake up" when specific
patterns appear on the data and/or control lines. Otherwise, all signals
are passed through transparently. A consequence of this is that several
devices, using different protocols, can share a single port, since they
will only respond to their own wakeup codes.

But to complicate things further, some of the protocols support their
own chaining concept, so that, several devices of the same protocol can
be separately addressed on the same parallel port. Thus, these three
variables are independent. If I count correctly, that's one level
deeper than the SCSI hierarchy:

type/port/protocol/chain/unit/lun/partition

I suggest that any naming hierarchy should be flexible in the number
of independent variables and their interpretation.

--------------------------------------------------------------------------
Grant R. Guenther grant@torque.net
--------------------------------------------------------------------------