[sane-devel] GUI for sane configuration (again...)

Henning Meier-Geinitz henning@meier-geinitz.de
Mon, 17 Jun 2002 20:22:20 +0200


On Fri, Jun 14, 2002 at 03:23:24PM +0200, stef wrote:
> 	I came across the last discussion of GUI for sane configuration. I agree that
> a backend should be good enough to work out of the box. But in some cases, it just
> can't happen. For instance the umax_pp backend cannot detect properly which model is
> attached. Even UMAX provided windows drivers can't: they detect my scanner as a 
> Astra 2000P or 1220P depending on the last version of the driver installed. More,
> IEEE1284 probe information is still 'UMAX Astra 1220P' for 2000P models (with
> bogus class 'IMAGE' by the way).

Are they different in how you need to "talk" to them? Or is it just
the name?

No way to determine which device it is? For the Mustek USB scanners,
there is a way to determine their type or to be more exact, their CIS
even if you don't know the USB identification. There are three types
of sensors, one with 300 dpi, one with 300/600 dpi (can be switched or
something) and one with 600 dpi. So you just scan a selected area and
detect the sensor by this result. Not nice, but works. Maybe a similar
way to distinguish between other scanners exist.

> 	Second, due to the use of parallel port, the backend is affected by parport
> bugs (either software or hardware). Or has some bugs that only appear on some 
> hardware, and that can hardly be corrected since I've no docs for these scanners. So,
> sometime direct hardware access fails and you have to use ppdev character device,
> and sometime, it is the opposite.

Yes, but I don't really thing a GUI can help much in this case. The
user has to consult documentation on what "ppdev" or "direct harware
support" means anyway. At least he has to make sure it's supported by
the kernel or to learn how to interprete error messages.

> 	Third, since umax_pp scanner probing may confused any device on the parport,
> it is commented out in 'dll.conf'. But many people don't read carefully the docs, or
> forget about that and can't get sane working.

People who don't read docs will get lots of other problems on Linux
and any other system. I don't wan't to make it harder for them than
necessary but uncommenting one netry doesn't seem unreasonable for me.

> 	So, for these reasons, the umax_pp backend just can't work out of the box,
> and hand tweakings of conf files is needed. No problem with me, but there is a large
> class of use who'd rather point and click than edit a couple of files. 
> 	I'd rather spend some time in helping write such a GUI, than endlessly send
> the same emails about how to configure the scanner. 

Ok, do it. Nobody can stop you from programming and having a choice is
always a good idea. But don't be surprised if you need to send the
same emails about your gui then :-)

Which users do you target with a gui? Probably not the ones compiling
source code? So maybe it's better to help distributors to make their
SANE configuration tools more powerfull. I know at least SuSE, who
does some tweaking of config files with yast.

The problem with a gui (nedd it to be graphical) isn't just the work
for the program, but also the constant work for updating it for all
the backends.