W3C home > Mailing lists > Public > xml-uri@w3.org > May 2000

Re: Syntax and semantics

From: Jonathan Robie <Jonathan.Robie@SoftwareAG-USA.com>
Date: Wed, 17 May 2000 07:17:32 -0500
Message-Id: <4.3.1.0.20000517060450.00ad7f00@127.0.0.1>
To: xml-uri@w3.org
At 02:59 PM 5/16/00 -0500, Paul Prescod wrote:

>I don't see a definition of semantics in your quote from Grune. It's
>like obscenity. One short-hand definition might be "anything not
>specified in the grammar."

That's an unusual definition of obscenity <grin!>

I *do* believe that the grammar should be all I need to determine the names 
and structure of an XML document. It must be fast and easy to find the 
names of elements and attributes. Fancy stuff comes later.

Grune's definition of "semantics" might differ from the one you give, but 
that's largely irrelevant, as long as we can agree that the grammar should 
be sufficient for determining the names and structure of an XML document. 
As I understand it, Tim uses the term "semantics" in conjunction with "The 
Semantic Web", which is related to semantic networks and RDF. Grune defines 
"semantic clause", rather than semantics, and discusses semantic clauses in 
connection with attribute grammars and affix grammars. All of this stuff is 
fancier than I want if all I need is the name of an element or an attribute.

Both RDF and attribute grammars rely on lower-level systems that establish 
names. This requirement is not limited to software systems that deal with 
semantics.

>The question is not whether XML w/namepaces should have semantics --
>XML does. The question is which semantics it should have. Clearly we
>want "tree-building" semantics. XML also has "link-defining semantics"
>(ID/IDREF) and vocabulary conformance semantics (DTDs).

As you note, "semantics" is a term that is used differently by different 
people. Grune would probably call these "structure" rather than 
"semantics". We do need to find the scope of a namespace declaration to 
resolve names, so understanding the tree structure is needed for name 
resolution.

Regardless, it should be fast and easy to find the names of elements and 
attributes. If we forget this, we're sunk.

Jonathan
Received on Wednesday, 17 May 2000 07:15:38 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:32:42 UTC