RE: [xml-dev] Interesting mailing list & a rare broadside

["Dare Obasanjo" <dareo@microsoft.com>]

The W3C XML Schema recommendation allows for doing (1) since xsischemaLocation and xsi:noNamespaceSchemaLocation are just "hints" but doing (2) is the kind of thing that is typically called a "non-standard extension" and given that schemas should be interoperable as much as possible, I'd be hesitant to encourage validating processors to augment the REC in such a manner. 

	-----Original Message----- 
	From: Henry S. Thompson [mailto:ht@cogsci.ed.ac.uk] 
	Sent: Sat 6/8/2002 11:52 AM 
	To: James Clark 
	Cc: Dare Obasanjo; xml-dev@lists.xml.org 
	Subject: Re: [xml-dev] Interesting mailing list & a rare broadside
	
	

	I'm not sure your answer follows from Dare's, but I'll let him answer
	that.
	
	Let me try to clarify the overall situation.
	
	When the WG were designing this part of the REC, we were confronted by
	two competing requirements:
	
	 1) document authors should be able to specify the schema they
	 authored to and expect to be validated with;
	
	 2) application authors should be able to specify the schema they want
	 incoming documents to be validated with.
	
	We identified four major sources of schemas in general:
	
	 1) Direct provision by the application;
	 2) Specification by the application user ("on the command line", as
	    it were);
	 3) Via namespace names interpreted as URIs and dereferenced;
	 4) Specification in the instance via xsi:(noNamespaceS|s)chemaLocation
	
	We concluded that it would be a mistake to give _any_ of these
	cast-iron primacy, and wrote the REC so that any the above could be
	preferred to any other, by application design choice and/or runtime
	user choice.  We noted that a general-purpose schema validator would
	do well to provide for user choice. So far, I'm not aware that much
	choice has been given -- all the validators I'm aware of operate
	pretty much in the same way, i.e. 'command line' schema documents
	override schemaLoc and/or namespace name-derived schemas.
	
	I think general-purpose processors could and should do better (and I
	include XSV here), in two ways:
	
	  1) Provide a switch that says, in effect, "don't listen to the
	     instance _at all_";
	  2) Provide a parameter specifying the desired top-level element
	     or type.
	
	I also expect that non-general-purpose processors operating in
	mission-critical environments will operate the "don't listen" strategy
	as  a matter of course.
	
	ht
	--
	  Henry S. Thompson, HCRC Language Technology Group, University of Edinburgh
	          W3C Fellow 1999--2002, part-time member of W3C Team
	     2 Buccleuch Place, Edinburgh EH8 9LW, SCOTLAND -- (44) 131 650-4440
	            Fax: (44) 131 650-4587, e-mail: ht@cogsci.ed.ac.uk
	                     URL: http://www.ltg.ed.ac.uk/~ht/
	 [mail really from me _always_ has this .sig -- mail without it is forged spam]