Re: Comment on XSD 1.1

Mukul Gandhi wrote:
> From what I could gather from your current proposal, you seem to say that:
> We have some good Schema languages like RelaxNG and Schematron, other
> than XSD. You are suggesting, that we shouldn't move from XSD 1.0 to
> XSD 1.1 but instead develop a new schema language say, X which has
> best of all worlds. I totally agree with this. 
No, "the best of all worlds" is nothing like my proposal.  

And suggesting that this should have been raised for XSD 1.1 is a bit 
rich: it was explicitly taken off the table for XSD 1.1 wasn't it?   The 
reason I am raising this to the TAG rather than the W3C WG (the 
individuals in which I respect, despite my frank comments)  is that I 
think the committee process has manifestly broken down, for whatever 
reason. A committee that still, after 8 years, has not even been able to 
*start* looking at what is clearly the most pressing issue is clearly 
disfunctional to the point that its product cannot be given the benefit 
of the doubt. XSD 1.1 may even entrench the problem rather than make it 
easier to dig out of, but XSD 1.1 should only be done in the context of 
getting a simplified XSD. 

What James Clark said* in 2001 at the W3C plenary in Boston when he 
pointed has proved accurate: "little short of a disaster." XML Schemas 
Structures is "overly complex, yet lacking in expressive power."  What 
other standard had to have workshops to try to figure out why 
implementations were so ratty? What other standard has a W3C TR listing 
the parts that are safe to use for one of the leading applications?  
What other standard has had such constant negative comment?

We need something simpler.

And since the issue-by-issue approach has failed, I propose a method 
based on a practical agenda, so that the WG has enough direction that 
there is a possibility of a result: first make a standard (say, XSD 
Lite) that is the intersection of XSD and RELAX NG, and that pretty much 
sits between the hull and kernel provided by the W3C Databinding 
patterns TRs, building on RELAX NG semantic and XSD syntax. Put that 
out, recommending it as the schema language of choice for small, fixed 
documents and small projects especially a new target for databinding.

Then fold whatever minor syntax changes this requires into XSD 1.1 which 
can then go ahead. If I had my druthers, XSD 1.1 should be reformulated 
in terms of adding type, reference and assertion apparatus to XSD Lite 
(i.e. on top of restricted RELAX NG semantics) before being released, 
but that certainly would be a large editorial effort.

The XML Schema WG has rather laid itself open to the charge that it is 
pandering only to the needs of those who want value-adds and tweaks, to 
the XQuery community for example, at the detriment of the rest of us. 
But the needs for XQuery and typed XSLT are niche requirements, they are 
simply not mainstream requirements even if you only look at the 
commercial uses.  It is great if XSD 1.1 turns out to be the perfect 
technology for this niche.

Cheers
Rick Jelliffe
* http://www.xmlhack.com/read.php?item=1097

Received on Wednesday, 20 May 2009 11:39:13 UTC