> I don't think that's a very good assumption. IMHO, if we do want to
> go towards autoconfiguration, we should go all the way. That is, we
> should try to get rid of the configuration file completely. Anything
> else will only increase user confusion.
If we have /proc/scsi/scsi, there is no problem with that, a backend
(or a sanei_xxx) just reads it and looks if it supports a scanner
listed there.
But if there is no /proc/scsi/scsi, it is not a good idea to scan
all possible scsi-devices, especally if there are dozens of devices.
I am thinking about making a configuration tool that scans the scsi-bus
for scanners and processors (hp) and generates the config files
with VENDOR="XXX" MODEL="YYY" ID=Z
for the backends, so the user does not have to modify the *.conf-files.
That tool could set the permissions for the /dev/... devices if it
is called with root-permissions.
It could be done with a nice graphical frontend, where you can
enable and select network-functions and SANE-servers etc.
We could make a group "scanner" and edit the /etc/group file with it.
If someone changes the hardware-configuration, he has to call the
configuration-program again.
Bye
Oliver
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com