OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.


Help: OASIS Mailing Lists Help | MarkMail Help

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Are we losing out because of grammars?

"Thomas B. Passin" wrote:
> Marcus Carr wrote -
> >
> > Eric van der Vlist wrote:
> >
> > > Looks like we are going towards 4 different types of XML "schema"
> > > languages to define:
> > >
> > > - The structure
> > > - The datatypes
> > > - The rules
> > > - The semantic.
> >
> > This is very interesting - I wonder if we might be heading for another split
> > such as the one that saw XSL split into T and FO? The differences are
> > becoming more pronounced, and the term "schema language" is starting to lack
> > definition as a result.
> >
> > Eric (or anyone else), how do you differentiate between rules and semantics?
> Yes, it's really interesting to think about what "semantics" is supposed to
> mean, isn't it?  I'm going to suggest that almost all computerized data
> processing occurs at the syntax and rules level, with virtually no "semantics"
> involved.


> Consider an sql update query.  It has to follow a certain syntax.  There are
> also some rules that are understood somehow.  The database system also has
> some rules to follow that the query itself does not contain and that the query
> writer may or may not have taken into account 0f - like data integrity rules,
> or other stored procedures that get triggered.  There is no "semantics"
> involved.
> I think that in such cases, when we say "semantics" we mean that a human who
> undertands the domain has written rules  and database structures to help cause
> the 'right" things to happen.  But the computer system only knows the syntax
> and the rules.

That's true.

I wonder if semantic should be considered as a direction to follow
rather a target that we can reach.

By semantic, we usually mean the ability of a computer program to
qualify an abstract object -- not so far from the common definition
("meaning of words") -- and we will probably never be able to write a
computer program that will fully qualify this object.

We are making progresses, though.

Allowing meaningful names in SGML and XML has been a step in that

Defining data types with OO features is another one.

The next steps that IMHO we are seeing now are to define classification
systems and (probably more fundamental since we will never define a
single one) ways to express relations between these systems.

I think that this is the commonality between the "Semantic web" and B2B
projects such as ebXML.

> Now, if a computer system may be asked to choose between several courses of
> action based on some other information, perhaps it might mimic what a human
> would do in this case.  Then would we be entitled to speak of "semantics"?  It
> seems that "semantics", at least in these applications, might consist of
> choosing sets of rules, and maybe even syntaxes, based on other information
> which we might call a "context".

IMHO, enabling software taking more and better decisions is one of the
applications of adding more semantic rather than being semantic by

> Of course, choices like these  might be considered to be the result of
> applying higher order rules, but still, additonal information is needed.  In
> this way, "semantics" is certainly a "higher" level layer than syntax or
> rules.   As such, it doesn't seem to be a subject for schemas per se.

I think it is linked to schemas, though.

It's actually one of the purposes of the W3C XML Schema annotation
feature and of the Schema Adjunct Framework (SAF).

It's pretty useful to be able to say that a node with a "date" datatype
is a shipping date and this kind of information belongs to semantic.

And it's also useful and needed to describe this semantic using specific
schema languages...

> Two questions arise -
> 1)Is this a useful view of "semantics" - that it is a means to choose between
> various sets of rules or perhaps syntaxes?

I had seen it the other way, schemata being a hook to add semantic and
semantic needing its own schema languages.

> 2) Do the systems people are envisioning need such a capability?
> Cheers,
> Tom P

Eric van der Vlist       Dyomedea                    http://dyomedea.com
http://xmlfr.org         http://4xt.org              http://ducotede.com