> I very much agree that change marks based on authored markup 
> would be the best solution. For a very technical audience I 
> am nevertheless interested in process (b), at the XML level. 
> I am well aware of the complexity of a Longest Common String 
> algorithm combined with inserted/deleted/moved elements plus 
> constraints based on id attributes... Is the process (b) you 
> mentioned an open resource?

I believe Norm Walsh's diffmk tool was used. I don't know where the
definitive source for this is.

Regards,

Michael Kay
http://www.saxonica.com/
http://twitter.com/michaelhkay