HAL and scanners

[abel deuring]

Joe Shaw wrote:
> Hi,
> 
> Étienne Bersac wrote:
>> Gnome Scan is based on SANE. Gnome Scan provide a library that wrap sane
>> into a glib-ish API and a library of widget and a collection of
>> applet/plug-in. I wonder if i will be able to drop the glib-ish API if
>> hal allow this. I mean, HAL handle scanner either with sane or whatever
>> you want. If HAL does not, gnome scan only need usual hal device info
>> plus the sane device name in order to use sane. A simple
>> scanner.sane.device_name is enougth.
> 
> The important thing to remember about HAL is that it's primarily used
> for device enumeration. 

Sounds sane (pun intended ;) Aside from HAL's "core job", see my
other mail to list about possible risks, if Sane backends would be
too deeply integrated into HAL.

> The acronym "hardware abstraction layer" is
> mostly a misnomer.
> 
> It's still the responsibility of applications to use the SANE, or
> gnome-scan, or whatever APIs to do the scanning.  HAL hopefully removes
> the necessity on the user to have to provide information about their
> scanner, the device node (if people have to provide /dev/bleh, we lose),
> etc.

My point is this: if you consider HAL as a sort of "hardware
database", it might make sense to add some hints about the
compatibility between a device and APIs that are installed on a
system. While it is nice for me as a Sane developer to know that
Sane is more or less the standard for scanning under Linux/Unix, I
don't want to enforce this near-monopoly by setting "scanner
supported" equivalent to "scanner supported by Sane".

Abel