snip...
> Masks should definitely, positively, absolutely NOT be an attribute in
> HTML! Transparency information is part of the image, not of the document,
> and to the greatest extent possible the addition of more ill-designed
> hacks to HTML to make up for the deficiencies of other formats should be
> avoided. I'd much rather see TIFF being used (or some new form of JFIF
> with embedded transparency info) than have this go in.
Embedded transparency is OK for somethings, but not for everything. One
of the tricks we use is to reuse a single image (our reader is smart
about caching images across pages) with different masks to achieve
"interesting" effects. Obviously this can't be done with embedded
masks. A JPG and several 1-bit GIF masks are much smaller than several JPGs.
TIFF is unusable because it has lousy compression (compared to JPG or
fractals). And a new JFIF format is something the world doesn't need.
As a final point, we intend to ultimately support a full alpha channel,
not just a 1 bit mask. Since not all browsers will support this, why
burden the data channel with useless information by embedding it in the
image file. Better to let the browser choose whether to fetch both the
image and the mask.
> Then again, maybe at this point there really isn't very much in HTML
> that makes it worth saving from becoming an even bigger pile of half-
> assed kludges than it already is.
> (I don't like the <OVERLAY> element either. This is properly the
> province of a simple vector graphics format. If HTML tries to do
> everything itself, it will quickly become as huge, unweildy, and
> difficult to understand as the U.S. Federal Government.)
Flamebait...
- Ron