CKMY craziness...

[Saem Ghani]

On Sunday 21 November 2004 23:53, Boudewijn Rempt wrote:
> On Saturday 20 November 2004 18:42, Saem Ghani wrote:
> > Though, I wonder if this can be done with layers?  A simple wetness layer
> > could be used in conjunction with each pixel layer?
>
> That was the other possibility -- but there is really nothing that a
> watercolour paint layer has in common with an 8-bit rgb layer. It's not a
> simple matter of adding information to rgb, but rather the entire
> composition of a pixel is different.

Yup I think I have a good grasp at what's being aimed at now.

> any type of 2-dimensional data in a memory efficient way coupled with
> interpreters for that data.

> That's very close to the original plan we developed this summer -- an if
> you create filters and other pixel-mangling code as templates on byte, int,
> float, then you don't have to agree on a single data size, but can offer
> any datasize that is useful.

Perhaps, QUANTUM isn't particularly bad.  With some massaging, basically 
ripping out some assumptions from the code.  We could make the transition 
fairly easily -- I hope.  If the object that stores the image knows the size 
of each pixel in bytes, by way of knowing what dataformat (READ: total bytes 
to describe a pixel in the image) it has, then we're set.  If we want to test 
bounds, we basically query the dataformat.  It keeps everything fast, and all 
the code that interprets the pixels will only have to worry about whether 
there are the correct number of bytes and then accessing and casting them 
appropriately.  Converting between various formats, can be done via CIE, 
which should get rid of all the gamut problems, except for a special path to 
RGB for the sake of performance when preping for display.

I hope I haven't done a full circle in my thinking and basically restated what 
the original intent was, and that this idea is both good and original.