Caution using XML Schema backward- or forward-compatibility as a versioning strategy for data exchange

Roger's point about the distinction between validation and processing is
entirely on-point. If the term "compatible" only means "causes limited pain
to whoever maintains XML validation code" then its not a very useful word.

I think that both "semantic" and "compatible" are problematic terms. What
is "compatibility"? Is it breakage to a system or pain experienced by a
user? Is it something else? Defining "compatible" too narrowly and too
technically makes the word useless because it causes too many changes to be
defined as non-compatible, and presumably we need the identification of
"non-compatible" changes to drive some action. The usual use of the term
compatible suggests that validation is more important than processing.

There is a change case that has not been talked about, and that is where a
schema is used as the basis for another schema or is incorporated into
another schema. This has a different set of sources of pain than changes to
XML documents because the impact of a schema change may not alter the
documents that are produced but is likely to impact the schemas that
incorporate it. It seems that there is a different (and more useful) set of
meanings of the term "compatible" in that use-case.

Greg