[sane-devel] Description file, Epson, and wildcards

[Olaf Meeuwissen]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Chris Bagwell wrote:
> Hi all,
Hi Chris,

> I'm reviewing Fedora patch set and notice an issue that I'm not sure
> how to handle within Sane.  Has this been discussed in past?
> 
> It looks like all 3 Epson backends can detect SCSI scanners based on
> scsi type being "processor", vendor ID being "EPSON", and the
> model/product name having 1 of 5 prefixes.  This is good to know for
> next step.
> 
> On Fedora, it seems its a common problem that hal or udev gets the
> permissions for scsi Epson scanners wrong and users can only scan as
> root.  They generically fix permissions for scsi "scanner" class but
> not "processor".  Their work around is to edit entry for specific
> scanner in the epson2.desc file and a line as follows each time a user
> reports a problem:
> 
> :scsi "ESPON" "Exact Model Name Reported" "processor"

The 2010-03-01 git snapshot has

 $  grep :scsi doc/descriptions*/ep*.desc
 doc/descriptions/epson.desc::scsi "EPSON" "SCANNER GT-7000" "processor"
 doc/descriptions/epson.desc::scsi "EPSON" "Perfection1200" "processor"
 doc/descriptions/epson.desc::scsi "EPSON" "Perfection636" "processor"
 doc/descriptions/epson.desc::scsi "EPSON" "Perfection1200" "processor"

already.  Are you talking about other models?
Of course, the epkowa and epson2 backends ought to do the same (insofar
they support these models), especially since the epson backend has been
deprecated.

The sane-desc tool from the same sources ought to output the right
udev/hal/fdi/whatever rules.

None of this stuff is in the latest sane-backends release (1.0.20), so
that's probably why everyone is patching.  :-(

> Once thats done, sane's tools will generate a
> libsane.fdi/libsane.rules that will set permissions correct.  Sane is
> not providing enough details right now it seems.  First option to help
> end users is to start providing that much details.
> 
> Mostly, the epson2.desc values are driven by epkowa developers from
> their epkowa.desc.  Would it be possible for epkowa developers to
> provide a ":scsi" line for all SCSI devices?

Yes.  No promises on a release time line but we could send a diff to the
list and ask someone with the necessary privileges to commit.

> Second option is to provide some kind of wild card support.  For example:
> 
> :scsi "EPSON" "Perfection*" "processor"

Probably not a good idea to begin with and maybe not very helpful
considering that

> [t]his would be pretty difficult to convert into hal and udev yntax
> though. Its also will cause some confusion when a Perfection 636S rule
> matches a Perfection 636.

> A third option is to provide some sort of hard coded/hand written wild
> card rules that is added to libsane.fdi and  libsane.rules for these
> special cases.  This is probably the easiest.  It looks like at east
> Epson and HP could benefit from this.
> 
> Do people think this is worth addressing?

Anything that makes out of the box scanner device detection and setup so
you can scan without root privileges is worth addressing.

Hope this helps,
- --
Olaf Meeuwissen, LPIC-2           FLOSS Engineer -- AVASYS CORPORATION
FSF Associate Member #1962               Help support software freedom
                 http://www.fsf.org/jf?referrer=1962
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkuMo80ACgkQt5qrxaZLMnLmpwCfewAiu0pv4R6SjFSa5DyhOdCN
ybUAni0H79NrdCChK6xQ0Wo8oBjn/xGb
=czsv
-----END PGP SIGNATURE-----