Inserting 'id's to improve "diffing".

One useful functionality to support is to be able
to "diff" between various versions of a document.

Without help, it can be hard to figure out when
some things "line up"; a straight ASCII-based 'diff'
could really struggle to identify common lines.

It seems to me that inserting 'id=' values with pseudorandom
hash values into, for example, each paragraph, MIGHT make it
easier to perform a document diff.  Experts can tell me
whether or not that's really true.  But if that's true,
then that should be noted somewhere in the specification;
I didn't see any discussion of this.

--- David A. Wheeler