Re: Syntax and semantics

XML has precious little global semantics.

But precious little XML, as used, has no post-syntactic semantics which 
is proper to the genre or dialect built using XML for streaming, for 
structuring, and for namespacing.  The work of the names is not done
unless the proper semantics of the markup is honored.

>
>                           Re: Syntax and semantics
>                                       
>   From: Jonathan Robie ([1]Jonathan.Robie@SoftwareAG-USA.com)
>   Date: Tue, May 16 2000
>
>>  * In what sense does XML not have semantics? Isn't the interpretation
>>of less-than symbols and ampersands as an annotated, tree-structured,
>>information set the "semantic content" of XML? Can any useful language,
>>or meta-language, or meta-meta-language be entirely devoid of semantics?
>
>I suspect that the term "semantics" is used differently by different
>people. I understand the relationship between tokens, structure, and
>semantics pretty much the way Dick Grune described it in his "Parsing
>Techniques, A Practical Guide":
>
>To me, namespaces are used to disambiguate names, and are used in the
>process of tokenizing. XML DTDs or schemas determine which sentences are
>legal and describe the structure of those sentences. Semantics are not
>contained in XML per se, but in systems that use XML, including RDF, but
>also including many of the everyday programs that munch XML.
>

The names also bear connotations that link them to the semantics.  These 
connotations may be formalized in the type definitions and constraint 
definitions in a schema, or they may be informally expressed as in 
XHTML.  But they are there.  The dialects comprise not just the structure,
but also the proper constraints on interpretation, that are bound to 
structure and to plaintext content using XML and names in XML as the
glue.

XML 1.0 does not specify what the semantics is, other than containment
hierarchy and a few things like IDREF.

But XML applications that are completely free of proper semantics, that 
is to say semantics systematically bound to all instances of the markup 
names and patterns, is a vanishingly small subset of the uses to which 
XML gets put.

Anything that touches naming and name evaluation participates in the 
lifeline for that semantics.  This includes namespacing.

It is not "leading the Web to its fullest potential" to relegate the
semantics to the private purview of the processors.  The dialect needs
to disclose semantics so that the family of processors is open.

This is particularly important for namespaces, which is a way to create
new dialects from elements of old.

Al

Received on Wednesday, 17 May 2000 19:25:03 UTC