On Fri, May 14, 2004 at 07:31:03AM -0700, William Skaggs wrote:

3) In devel-docs/parasites.txt, a standard parasite called "rendering-intent" is defined. I think this is based on a misunderstanding. Rendering intents come into play when you convert an image from one colorspace to another -- often the best way of doing the conversion is a function of the intended use of the result: printing, CRT viewing, etc. That is, a rendering intent is a property not of an image but of an image transformation.

I've two comments on this: first some image format store a rendering intent, and this setting should be preserved, and the natural place to do it is a parasite.

And second, rendering intent is a property of a image transformation, but there needs to be a way to select the actual image transformation, so the rendering intent _is_ a property of the image just like a image comment is.

Think of it that way: an image using pseudocolour should probably be rendered with wide colour spread. This is a property of a the image, i.e. exact colours are not important. The rendering intent is used to select, out of a set of colour transforms, one that best preserves this aspect of the image.

Hi,

"William Skaggs" writes:

1) When the jpeg loader finds EXIF data, it creates a parasite called "jpeg-exif-data". EXIF data is not specific to jpeg files, though, and having differently named parasites for "tiff-exif-data", etc, makes things unnecessarily difficult, as far as I can see. Can this be renamed as simply "exif-data"?

Is this data really exchangeable between file formats? If so then this should probably be changed.

2) If libexif is available, then the jpeg loader creates a jpeg-exif-data parasite regardless of whether the file actually contains exif data; if not, the parasite is just a bare header. Would it be preferable to not create a parasite in this case?

That sounds like the right thing to do.

Sven