Re: Semantics [was: Re: Well-formed (was: Re: F2F Proposed Resolutions Draft Updates)]

Hi Joe,

At 19:16 17/06/2005, Joe Clark wrote:

>>It is true that SGML does not define well-formedness, but if you say that 
>>a well-formed document is essentially "one that can unambiguously be 
>>parsed to create a logical tree in memory" (Jon Bosak, at 
>>http://www.isgmlug.org/n3-1/n3-1-18.htm), then you can apply this concept 
>>also to SGML.
>
>OK, so let me understand this: The Working Group is contemplating issuing 
>a vague and counterfactual guideline based on one person's blog posting,

The new success criterion is not based on Jon Bosak's article; if you had 
that impression, that is entirely my fault (i.e. the wording of my response 
to Gez Lemon). I tried to identify criteria for well-formedness "after the 
fact".

We now have 2 SCs at GL 4.1 L 1: one for SGML-based formats and one for all 
other formats. We may consider splitting the first one into two SCs: one 
that requires well-formedness for XML-based formats and one that requires 
something else for non-XML SGML-based formats, but we're still struggling 
to define this "something else". If we want to stick to the terms of the 
SGML standard, we could require that "Non-XML SGML-based delivery units are 
formatted according to the SGML declaration of their specification or to 
the Reference Concrete Syntax if no SGML declaration is defined." We might 
then add a note saying that this does not require (type-)validity. The XML 
specification does not define well-formedness by means of an SGML 
declaration but in Extended Backus Naur Form; this is why I propose to 
split the first rule instead of treating XML as a special case of SGML.
An SGML declaration defines such things as the character set and the 
characters that can be used for delimiters (e.g. <, >, </); it defines a 
"concrete syntax". The Reference Concrete Syntax is a concrete syntax 
defined in the SGML standard.
(Note that SGML parsers are not required to detect or report errors.)

Regards,

Christophe Strobbe



-- 
Christophe Strobbe
K.U.Leuven - Departement of Electrical Engineering - Research Group on 
Document Architectures
Kasteelpark Arenberg 10 - 3001 Leuven-Heverlee - BELGIUM
tel: +32 16 32 85 51
http://www.docarch.be/ 

Received on Monday, 20 June 2005 11:15:21 UTC