N1861

ISO/IEC JTC1/SC18/WG8

Document Processing and Relating Communication—

Document Description and Processing Languages

Report of the SGML RG: Discussion Notes

At its meeting May 20-24, WG8 discussed the following areas:

1. ISO 8879:1986 states that "an attempt to redefine an entity is ignored, but is not an error." However, multiple declarations of element types, attribute names, notation names, and short reference map names all are errors. WG8 has recommended that ISO 8879 be revised to permit an individual associated element type to appear in multiple attribute definition list declarations, and an individual attribute name to be specified multiple times within the attribute definition lists associated with an individual element type.

   Although time did not permit full design or establishing consensus among the experts present, more consistent treatment of redeclaration was discussed at this meeting. A proposed recommendation is that all attempts to redeclare an entity, element type, attribute type, short reference map, short reference use, or notation name within a document type declaration would be ignored and not considered errors. Link type declarations would be treated analogously.

   Ignoring attempts at redeclaration allows an individual document to override declarations in an external document type declaration subset. While there are situations in which users may wish to modify the effect of an external document type declaration subset in this way, there are also situations in which the designer of an external document type declaration subset wants to prevent users from doing so. Therefore, WG8 discussed an enhancement that would allow indicating that redeclaration of a particular item was indeed an error. Possibilities included both optional new parameters of individual declarations and the ability to mark a block of adjacent declarations as being unmodifiable.

2. While agreed that despite wide-spread use of SGML, a situation in which the differing opinions would produce different results has not yet arisen, experts disagree on the interpretation of names and parameter literals that appear within the concrete syntax parameter of the SGML declaration (in particular within naming rules, reserved name use, the delimiter set ). Do character representations that appear explicitly within such names and literals represent characters in the syntax reference character set or characters in the document character set? Do numeric character references within literals refer to character codes in the document or the syntax reference character set?
3. At end of incomplete discussion, WG8 had come close to arriving at consensus on recommending the following additions to SGML:
   • A document type name to DOCTYPE declarations separate from the name of the document element
   • A new declaration (or specialization of notation declaration) for relating a document type name to a public identifier that acts as a universally-unique name for a document type as an abstraction, possibly distinct from any SGML syntactic expression of that document type (i.e., the public ID for an external declaration subset)
   • The ability to declare document and subdocument entities that must conform to specified document types and/or architectures
   • The ability to specify data attributes to SUBDOC/DOCUMENT entities
   • Specification of multiple namespaces within an SGML document (proposed by the Japanese member body). This possibility has interesting but non-trivial interactions with the SUBDOC/DOCUMENT proposal.

   WG8 also discussed the issue of semantic distinction between documents and subdocuments and whether or not this is a distinction that should be codified in SGML.

4. An inheritance mechanism between element types simplifies declaration of related types. This mechanism would allow a derived element type to inherit the following properties of base element types:
   • A derived element type inherits all attribute definitions of its base element types.
   • A derived element type inherits the content model (or declared content) of its base element types.
   • In an SGML instance, a derived element type could be used instead of the element type defined in a content model.
   If this proposal is accepted, it should be possible to state which of the three a given derived element type inherits. In a tentative syntax, base and derived element types might be declared as follows:

         <!ELEMENT common-attr O O DUMMY
             -- "DUMMY" is new declared content to show the element never
                 occur in an instance. --
         >
         <!ATTLIST common-attr
                     id         ID         #IMPLIED
                     notation NAME      #IMPLIED
                     ...
         >
         <!ELEMENT list ...
             -- base element type --
         >
         <!ELEMENT orderd-list [list common-attr] ...
             -- derived element type. inherits "list" and "common-attr" --
         >
         <!ELEMENT para1 (list _ #PCDATA)
         >
         <!ELEMENT para2 (orderd-list _ #PCDATA)
         >
   

   Corresponding element instances might then be marked up as shown below:

         <para1><list>...</list></para1>                     --- OK.
         <para1><orderd-list>...</orderd-list></para1>    --- OK.
   
         <para2><list>...</list></para2>                     --- ERROR.
         <para2><orderd-list>...</orderd-list></para2>    --- OK.
   
         <orderd-list id="xxx">...</orderd-list>
             --- attributes of "common-attr" could be used.
   

   This proposal will be discussed in future meetings in conjunction with related issues, such as attribute definition list enhancements and architectural forms.
5. WG8 discussed a proposal for extending the syntax for marked section declarations to allow an "If ... then ... else" mechanism such as is used in programming languages. It was concluded that there is no need for this extension since the requirements that motivate it have already been met with standardized mechanisms within HyTime, DSSSL, and General Facilities and the accepted proposal described in N1605 to allow boolean expressions within marked section declarations.