Re: Options for dealing with IDs

On Saturday, January 11, 2003, 4:39:47 PM, Elliotte wrote:


ERH> At 9:23 PM +0100 1/10/03, Chris Lilley wrote:

>>PG> XML could pick up the #ALL capability defined five years ago in the
>>PG> SGML TC2 [1] that allows a declaration of the form:
>>
>>PG>   <!ATTLIST #ALL id ID #IMPLIED>
>>
>>PG> in the internal subset to add an 'id' attribute of type ID to 
>>all elements.
>>
>>Great suggestion. That would make the 'always declare IDs in the
>>internal subset' option tractable, for common cases. Currently all XML
>>processors are require to read the internal subset, yes?

ERH> Yes, and currently every single one of them would report a 
ERH> well-formedness error on encountering this.

Of course, its not correct XML 1.0 syntax.

ERH> I can't see the need here
ERH> as being worthy of violating XML 1.0 notions of well-formedness.

Well it would not violate the notion but it would add syntax that all
XML 1.0 parsers would break on, and thus have stronger
backwards-compatibility issues (a more 'flag day' approach) than other
options (an attribute is an attribute, even if you don't understand
what it is supposed to contribute to the infoset).

ERH> This is a much more breaking change than any of the other
ERH> proposals,

Agreed and I will note this.

ERH> al of which work within current XML syntax and at
ERH> worse slightly change the definition of validity.

The intent was that they have no effect on the definition of validity,
in that they do not cause validation to happen. Its true that a
validator may find error with what is there, but that is always the
case.



-- 
 Chris                            mailto:chris@w3.org

Received on Saturday, 11 January 2003 13:55:45 UTC