HyTime supports this...
for that matter, there were people using an SGI reality engine (I think) at
the WWW conference in Chicago last October, demonstrating a virtual
reality addition to Mosaic, where you could see a tabletop with an apple
and a pear sitting on it. The surface tumbled over & over, and whenever
a fruit was visible, if you clicked on it, an appropriate link was followed.
So it's coming.
Audio and textual annotations to video certainly make a lot of sense.
``If at any time you would like to buy any of our cars,
simply click on the model you wish to purchase during this
infomercial. Of course, if you click on the front bonnet, you
will receive a 5% discount right here...''
``The series begin to converge when zeta falls negative;
click anywhere on the surface to zoom in to that region.''
Other data visualising applications spring to mind...
BUT:
I don't think that HTML 2.0 has to support them.
I don't think there is working code distributed widely enough for
this to be something that can go in 3.0 either -- or am I wrong?
Lee