On Thu, Sep 21, 2000 at 03:34:12PM +0200, Oliver Rauch wrote:
> Where is the problem?
>
> It had to defined and it really does not matter in which way.
It is completely inconsistent. If 1 means black in 1bit, why does it
mean white in 8bit? How about 16bit? What does it mean for color
channels? What other horrors will I find when I have time to write
a more complete set of tests?
> Does anyone really think it would be a good idea to change an existing
> standard in such a point? I don`t think so.
Chaper, section, subsection reference to the STANDARD please Oliver,
or do you mean "I don't want to change XSane at this point" ? I do not
see anything about this mess in the standard, just a clean clear
description of the various color depths.
> Where is the sense to discuss about it?
Well, it looks broken and is definitely inconsistent. "We always did
the wrong thing" is not a good excuse for doing the wrong thing,
especially when we're so early in the life of SANE. This will catch
out everyone who tries to implement the spec from scratch (not by
cutting and pasting code as most have so far)
As far as I can see this behaviour comes from using PNM as a testbed
for SANE (and a very inconsistent and bad testbed it was too, as all
who examine the resulting code can see for themselves) and it isn't
codified anywhere in the SANE documentation.
I would prefer that it be corrected in the implementations and
clarified in the spec. If not, the spec must be changed to indicate
how existing implementations do it, with a footnote explaining that
this is a result of a ridiculous historical artifact.
If this list comes to a consensus I will arrange for one or the other
to be in my next draft.
Nick.
-- Source code, list archive, and docs: http://www.mostang.com/sane/ To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com
This archive was generated by hypermail 2b29 : Thu Sep 21 2000 - 14:39:49 PDT