Semistructured data is a generic term for data that does have structure
information, while not being tabular or very tightly restricted.  XML and
HTML data is the most prominent examples for this.  You normally would not
use this term e.g. for database tables (which for example do not allow
nesting of entries).

While this application currently only supports XML, the algorithms should be
able to process other semistructured data as well.

The approach used here is usually much slower than other well-known xmldiff
applications, however it produces better results in many "tricky" cases.
You say that other xmldiff applications try to do a syntactic diff, whereas
xmldiff tries to do a semantic diff.

WWW: http://ssddiff.alioth.debian.org/