Re: [xml-dev] ID/IDREF is evil

[Michael Kay <mike@saxonica.com>]

> 
> My most radical idea was to extend the notion of VI (Value Indicator) from just '=' to a set of characters, according to the major categories of declared values.  Thus instead of name="value" for all attributes, we could have had name@"value" for IDREF/S, name&"value" for ENTITY/IES, name%"value" for notations and name*"value" for NMTOKENS, leaving name="value" for CDATA only.  And an ID attribute wouldn't need a name or surrounding quotes at all, if we tacked it onto the GI with a separator, like '!'.     
> 

Similar idea to FtanML which freely borrows from JSON:

married=true (boolean)
size=23.4 (number)
name="John" (string)
dimensions=[12, 14] (array of numbers)
colours=["red", "green"] (array of strings)
id=#ABC123 (ID) (not in the published FtanML paper, because we never worked out the details, e.g. how to mark references)

If you're changing the syntax, I don't see any merit in retaining XML's weird set of attribute types.

Michael Kay
Saxonica