> Authors can create a location address element that pairs the object in
> question with a unique local identifier. There are three kinds
> of addressing:
> a) by name (using ID attributes)
> b) by position in a coordinate space (the GIF example)
> c) by semantic construct (e.g. the 3rd polygon on the left)
My suggestion is that the position will be
1) A character number N (from the begining of the document)
2) An (x,y) pixel inside image number N in the document.
If there is a well defined syntax for this, it can be used in a URL
to point to any location in the document.