- From: <noah_mendelsohn@us.ibm.com>
- Date: Thu, 12 Aug 2004 10:04:29 -0400
- To: "Michael Kay" <mhk@mhk.me.uk>
- Cc: "'Rob Harrington'" <robot252@yahoo.com>, xmlschema-dev@w3.org
Michael Kay writes: >> I find the syntax extremely unmemorable I do too and I fought hard against this and other similar strangeness, much of which came late in the design process. For what it's worth, my recollection is that a fair amount of the current deeply nested syntax was justified by the view that our language should be able to do a reasonably good job of validating its own syntax. Many of the options that are simpler for users would involve co-occurrence contstraints in the .xsd syntax; for example, you might find that a baseType attribute would be allowed only if a derivation by restriction or extension were specified. The decision was made that since the language was not good for enforcing such constraints (and perhaps because some members of the WG considered such usage to be sub-optimal markup), we went for the clumsy, deeply nested and "unmemorable" syntax that we have today. I think there were also those who felt that ease of hand authoring is a secondary goal, as is "terseness" [1], and that most schemas would be created automatically by tools. I have never agreed with that view. Indeed, I have always thought that our deeply nested "no co-occurrence constraints" syntax was one of the most damaging decisions we made. With a more straightforward syntax, the language would appear quite a bit more simple and straightforward for many purposes, and that an 80/20 subset would be very easy to learn and remember. No doubt some subtleties and unfortunate complexities would remain for power users. Anyway, that's the history as I remember it. Noah [1] http://www.w3.org/TR/1998/REC-xml-19980210#sec-origin-goals -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 --------------------------------------
Received on Thursday, 12 August 2004 14:11:00 UTC