In this section, we focus on the two main already cited works
addressing what we have called ``static semantics'' of Web sites and
based on a context-dependent computation : the work of
PCR99 from the software engineering
community and Webmaster [van Harmelen and van der Meer1999] from the AI community.
We can compare our method with the one used by Psaila and
Crespi-Reghizzi that use attributed grammars to semantically evaluate
Web documents. It has been shown that Natural Semantics is a
generalisation of attributed grammars [Attali1992], and that some
Typol programs can be compiled into attributed grammars.
However our method is not limited to XML documents, as we can also use
it for HTML documents (like WebMaster) and, at the difference,
we do not need to modify an
actual DTD to apply it as we do not use XML attributes.
Even if our goals are very closed to those of WebMaster, our approach is
quite different. We are closed to traditional software engineering,
trying to take Web documents as input and producing error messages or
warning related to them, whereas WebMaster tries to classify documents
following the type of documents then the type of errors
that have been detected in them. With our approach, the output lists all the
errors and their point of occurrence
for each Web document, that, according to our point
of view, corresponds more to the needs of Web sites designers.
Our method is based on a more powerful expressiveness of semantics
representation: Webmaster addresses only semantic constraints
between pages whereas we can address both syntactical and semantic
constraints, being able to specify any semantic constraint between any
entity such as pages but also any fragment of pages.