[Hp3300backend-devel] Niash backend status update

[Bertrik S. <be...@zo...>]

Hi all,

Here are a couple of my thoughts about the status of the niash backend.

1) Rework of the transfer method layer
Currently each transfer method (USB using kernel module, USB using libusb,
USB using uscanner, libieee1284 transfers) only has an init which both
initialises the transfer method and tries to open the first available
device. I intend to extend that a bit (actually I have already started).
The new init function will report each device that it thinks may be used
to access a scanner, but not open it yet. There will be a new function
("open") that actually opens the scanner.
This way it should be possible to show the actually present scanners
(instead of always reporting a hp3300c even if no scanners are
connected) and it should even be possible to use more than one scanner
at once.

2) Sane 1.0.9 will support both libusb and the kernel module
I think this will simplify a lot of things, meaning we can use the
sanei_usb transfer method for the SANE backend and forget about the
specific kernel module, libusb transfer method and uscanner transfer
method.
The only remaining transfer methods will be sanei_usb and the libieee1284
method.

3) Image still stretched for HP4300c ?
I got an error report that the image scanned using the SANE backend
was still stretched (although I don't have actual images showing this).
I thought I fixed this a couple of version ago, but perhaps it got
broken again in one of the recent releases.

4) Other open issues
* we still don't know how to set the vertical starting position of the
scan for HP3400/4300 scanners
* we don't know how to properly set the vertical resolution for HP3400/
4300 scanners (what is implemented now, is rather tricky)
* there is no coarse calibration done to set the gain of the analog parts
(we still use a set of fixed values, which seem to work quite well for
most people. Although some other people report vertical stripes which is
a typical symptom)
* parallel port access test
I think there is a reasonable chance that the backend will work over the
parallel port, but it's hardly ever been tested.
* timer for the scanner lamp?
It would be nice to have some mechanism that keep track of whether the
scanner lamp has already warmed up or not.

If you there is anything that you can more-or-less independently
work on, please let me know.

All the best,
Bertrik