>
> Well it was a miracle it worked in the first place!
> 1) Linux thinks it is the only initiator on a SCSI bus
> 2) W95 + W98 most likely make the same assumption
> 3) The SCSI scanners I've played with have particularly
> bad target implementations prone to lock up at the
> first sign of an irregularity.
Yes and no (to 1 and 2). Nothing in SCSI necessarily precludes
you from having several initiators on the same bus, provided
that that the ID and termination rules are obeyed. Wether the
firmware on the target can handle this situation is an entirely
different issue. I do agree though that it's a miracle that the
target's firmware (the HP 5p) didn't crash when it received
an inquiry from initiator A with the scanning transaction in
progress to initiator B. Clearly though, it does have some
issues (if you want to see a corrupt image, check out
http://mccoy.penguinpowered.com/scanner). It does however
crash when two "Scan Window" commands arrived from different
initiators.
I've used the sharing trick on many systems over the years.
I've shared tape drives, disk drives and other (like scanners)
perepherials. Tape drives tend to behave like the scanner,
that is only one of the connected systems can use it a time.
Disk are usually more robust and treat the command/reply as
a queue, storing the initiator ID along with the request
and sending the reply in a atomic fashion to the correct
initiator (but, I would not swear that is is true for all
disk drives. In my experience this did work on Seagate SCSIs).
There are other issues with disk drives, mostly having to
do with filesystem metadata consistency problems, but
you can generally mount different partitions of the same
drive to different hosts. Note that storage area networks
are really the same thing, with software to control the
state of the metadata explicitely between hosts.
You can also mount the same partition several times to different
hosts in read-only mode.
SCSI has even been used as a high bandwidth "network" of sorts
between several machines (a topic that reappears on LKM
occasionally), but obviously that requires special software
to operate the controller in target mode.
> Interesting that you tracked down the problem. To my knowledge
> the SANE program doesn't poll the scanner. It is possible
> that a polling strategy could be used to detect a user
> pressing the scan button on those scanners thus equipped.
No, SANE doesn't. There are utilities though (daemons) that will
monitor the "button" and launch SANE when a "push" is detected.
I think one is called "hpbutton", but I'm not 100% sure on that.
> Doug Gilbert
>
--Jens
--
"We are Microsoft. UNIX is irrelevant. OS/2 is irrelevant. Openness
is futile. Prepare to be assimilated."
-- prs@turing.org
-- (quoted by Eric Berggren)
--
Source code, list archive, and docs: http://www.mostang.com/sane/
To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
This archive was generated by hypermail 2b29 : Thu Aug 24 2000 - 06:20:48 PDT