[Gimp-print-devel] Native CUPS drivers vs. GhostScript drivers/Color adjustment

[Till K. <til...@gm...>]

Michael Sweet wrote:
> cups-o-matic and foomatic are not part of CUPS, and (to be honest) we
> don't like to recommend them to users because they aren't a complete
> CUPS driver solution - some things aren't implemented due to the
> way things are wrapped, no color profile or RIP support, etc.
>

The missing color profile support is due to GhostScript not implementing 
this for all the legacy drivers (as "cljet5", "cdj550", ...) which 
GhostScript contains and Foomatic integrates. It would be a very dirty 
hack if someone would write a color correction pre-filter with 
PostScript both as input and output. The best solution would be if 
GhostScript would have general color profile support, for all drivers, 
not only for the "cups" device. Is it perhaps even possible to modify 
GhostScript that the color profile stuff which now acts only on the 
"cups" device will act on all devices? This would improve the usabilty 
of all the legacy drivers (which noone will convert to CUPS drivers) a lot.

Another problem of the legacy drivers is that they do not have clean job 
canceling (as the CUPS drivers probably have). Here GhostScript would 
also need further development (as catching a "Cancel" signal, then 
shutting down the job) but this would also need work in every driver.

What do you mean with RIP support? As I know GhostScript is also a RIP 
and so good old LPD already used a RIP.


Unfortunately, color profiling with native CUPS drivers is not very well 
known and also not very easy for the users.

For color adjustment there is now only the text mode tool 
"cups-calibrate" which comes with GIMP-Print, even if it works with 
every native CUPS driver (as IBM's Omni for example). Here it would be 
better to have it in CUPS (and tell about it in the docs) and perhaps 
even a second implementation of it in the web interface of CUPS, then 
more users would find and use it and then one could probably also 
collect there profiles to make the defaults of the drivers better.

"cups-calibrate" itself must be made more ergonomic:

- The different sample lines from which one chooses the best value 
should not touch each other, as the black and yellow line in the first 
pass and the three colour lines in the last path. Otherwise one is 
influenced by the naighboring lines. one or two cm distance would help 
very much.

- It is very difficult to see what is 25 % gray and 50 % gray. I would 
show the samples for 50% in front of a 1-mm-rastered chessboard of black 
and white squares (imagine every "O" in the following is black, every 
space is white:

     +--------------------------+
     | O O O O O O O O O O O O O|
     |O O O O O O O O O O O O O |
     | O O O O O O O O O O O O O|
     |O O O +------------+O O O |
     | O O O|            | O O O|
     |O O O |   Sample   |O O O |
     | O O O|            | O O O|
     |O O O +------------+O O O |
     | O O O O O O O O O O O O O|
     |O O O O O O O O O O O O O |
     | O O O O O O O O O O O O O|
     +--------------------------+

If the sample is 50% gray, from a longer distance the frame and the 
sample look like having the same colour.

For 25% you take the following pattern:

     +--------------------------+
     | O   O   O   O   O   O   O|
     |   O   O   O   O   O   O  |
     | O   O   O   O   O   O   O|
     |   O  +------------+   O  |
     | O   O|            | O   O|
     |   O  |   Sample   |   O  |
     | O   O|            | O   O|
     |   O  +------------+   O  |
     | O   O   O   O   O   O   O|
     |   O   O   O   O   O   O  |
     | O   O   O   O   O   O   O|
     +--------------------------+

I think, this would make the calibration much easier and much more accurate.

   Till