> > The basic idea is simple. Video information is 3D in comparison to imagemap
> > which is 2D. So we only need to return x,y, and t coordinates in stead of x
> and y
> > in the imagemap case.
>
> I think the "x,y" coordinate for imagemaps was a quick hack that should be
> upgraded; it's very limiting. Extending it to video would be even more
> limiting.
Another thing worth looking at is HyTime FCSLOCs
(finite coordinate system locations). Those can
be used to define rectangular ranges in any number
of dimensions:
<link id=video1 rel=embed href="video1.mpg">
<fcsloc locsrc=video1>100 110 0 250 0 200</fcsloc>
would link to the area (0,0) -> (250,200) of
frames 100-110 of "video1.mpg", for example.
Of course, FCSLOCs can't specify polygonal regions,
so we'd still need a NOTLOC (application-defined
location notation) for those.
(IMHO, HyTime defines a lot of useful concepts
but almost no useful features. I don't think it's
worth the effort to add full HyTime support to
browsers, but many of the basic concepts and
mechanisms are definitely worth looking at.)
--Joe English