[sane-devel] ADF support in SANE API

[Alexander Pevzner]

Hi Jürgen,

> nice to hear from you in person, I’m quite impressed by the work you have done in sane-airscan.

Thank a lot! :-)

> At some point, I wanted to include WSD support in AirSane but had to give up because Windows was never
> willing to create a device node for the announced device, even if the metadata xml closely matched that from a real scanner.
> So I concluded Windows has a whitelist for devices it is willing to accept, or some other undocumented mechanism designed to obstruct efforts like mine, or maybe I simply made a silly mistake that I was unable to see, despite I spent quite some time on this.
> Do you have any insights into that complementary side of your code?

I don't think there is a whitelist of devices. Most likely, you've made 
a silly mistake, which is very easy to make with this ugly protocol.

Although formally WSD is fully documented, the quality of this 
documentation is exceptionally poor. Written text sometimes contradict 
with examples from the same text. Very similar things may have quite 
different syntax in two different but similar contexts (for example, 
<a:To>URL</a:To> vs <a:ReplyTo><a:Address>URL</a:Address></a:ReplyTo> in 
the message header).

All implementations are very fragile. A minor mistake may easily crash 
the device. And seems at devices side there are only 2-3 implementations 
actually exist, and all vendors use them by license.

Note also, at the devices side implementations are sensitive to XML 
formatting. XML, formatted into the very long single line works best of 
all; easy to read XML, formatted into multiple lines, not always 
understood by devices (this is true for eSCL too). Probably, the same 
defect may exist on a host side.

And another thing, MessageID should really be different all times you 
send a new message. They really drop messages with duplicated MessageID 
(and on my device the memory where received MessageIDs are saved, 
eventually overflows, so discovery stops working for several minutes, if 
used extensively).

> OK, that sounds reasonable. So I have to report scan:AdfState as ScannerAdfLoaded all times, except once after scan failure.
> BTW, would you bother to omit scan:AdfState for scanners that don’t have an ADF? I’ve noticed that Vuescan complains about an empty ADF when used with AirSane, even though the scanner has no ADF. This would suggest that scan:AdfState == ScannerAdfLoaded is not optional but expected for all scanners, with or without ADF?

This is what HP Deskjet 2540 series returns:

	<pwg:Version>2.0</pwg:Version>
	<pwg:State>Idle</pwg:State>
	<scan:Jobs>
             . . .
         </scan.Jobs>

So there is no AdfState here (this device has no ADF).

-- 

	Wishes, Alexander Pevzner (pzz at apevzner.com)