Profiling, diff and change tracking best practices?

[Lech Rzedzicki <xchaotic@gmail.com>]

Hi all.

I am at a fortunate stage where we are redesigning our XML schema so
that it fits our requirements better.
To give you an idea of the XML we're dealing with, it's loosely based
on DocBook and used for multi-channel publishing.
Some frequent scenarios include updating XML with new content,
comparing versions, different languages, sending diffs to tranlation,
but also producing slight variations depending on the output. Tracking
changes (by being able to see what's been added and deleted) is also a
nice to have feature.
Basically what I aim to put in place is structures to help with these
function that are not too verbose to overwhelm editors, yet powerful
enough for 'future' scenarios.

My initial thoughts are to employ xml:id attributes on block-level
elements and add a set of attributes for each facet of profiling,
possibly reusing DocBook attributes such as condition, version,
audience, but my fear is that it won't powerful enough in the future.

I would love to hear your general thoughts on best practices in this
area of managing XML content and specifically on:

1. How low should we go with id's on elements? My main concern here is
making diffs as easy as possible and possibly identifying chunks of
xml that are as small as possible, making translation cheaper. On the
other hand should I be bother at all about the performance, since all
the documents are size-limited to a book size of ca 1000 pages(a few
MB of XML)?
2. Use a possible verbose set of elements/attributes on the elements
directly or use a meta-attribute that links to an attribute/element
set in a secondary file? (less verbose but more complex)
3. Are 'add' and 'remove' sufficient change tracking marks to cover
all scenarios? (I think any more complex edits such as update can be
built up from those two)?

I really hope I can get some good feedback from you and thanks in
advance for that,

Lech Rzedzicki