- From: David McKelvie <dmck@cogsci.ed.ac.uk>
- Date: Fri, 21 Jan 2000 12:30:25 GMT
- To: www-xml-infoset-comments@w3.org
Comments on XML Information Set W3C Working Draft 20-December-1999 Attributes Because there is no concept of a 'Attribute Definition Information Item' in the data model, it would seem possible to have the same attribute name on two different occurances of the same element type, which have DIFFERENT attribute types and default values. This sounds wrong. One solution would to have a new type of information item, i.e. a 'Attribute Definition Information Item' which contained the DTD/Schema info about the attribute and have individual 'Attribute Information Items' refer to it. If you want to allow editing of XML info sets, then it is necessary to know if an attribute is defined as #FIXED, since presumably an application should be discouraged from changing such attribute values. DTDs 'Document Type Declaration Items' don't appear to contain lists of element types or attributes defined in the DTD. If they did it would make it easier for a processor to write out a valid XML file starting from an XML information set. If 'Document Type Declaration Items' contained a list of 'Element Definition Items', this would make outputting DTDs easier. Also if 'Element Information Items' pointed to these 'Element Definition Items', this would centralise the element definitions, and allow one to extend the XML Information Set data model to multiple (perhaps hyperlinked) XML files, which had different DTDs and maybe element types with the same name. It is a pity that 'Element Information Items' don't contain content models, as this makes it harder to validate changes to the XML info set. I would say that it was essential that 'Element Information Items' say at least whether they are element-only or mixed content, since applications that write info sets to XML might want to format them differently. Character content Should this proposal say something about what the character content of #PCDATA that contains unexpanded entity references is? Maybe this is the wrong level, but if the info set doesnt handle this, then we are going to hit problems defining a query language that includes regular expressions over characters, i.e. does re "ab" match "a&unexpanded;b" ? does re "a.*b" ? David McKelvie Definition of equality of info sets Does this standard need to say when two info sets are considered the same? Do they need to be identically the same, the same on the core properties, or is it more complex? The Document Info Item Should the 'Document Information Item' say what peripheral info set properties it includes? Language Technology Group Human Communication Research Centre, Edinburgh University, 2 Buccleuch Place, Edinburgh EH8 9LW, SCOTLAND Tel:(44) 131 650-4630 Fax:(44) 131 650-4587 email: dmck@cogsci.ed.ac.uk
Received on Friday, 21 January 2000 07:30:28 UTC