Re: [xml-dev] Victory has been declared in the schema wars ...

[Rick Jelliffe <rjelliffe@allette.com.au>]

Michael Champion wrote:
>
> I have a few specific questions:
>
> - Is this groundswell of RELAX NG adoption coming from anywhere 
> besides XML geekdom?
>
> - What about XSD 1.1 (which clarifies some of the nastier corners of 
> the spec and addresses the very limiting lack of occurrence 
> constraints)? Is there any awareness of / demand for *it* outside 
> geekdom? And, ahem, to what extent is this “RNG – Mission 
> Accomplished” meme an attempt to strangle it in the cradle? :-)
>
> - Are the non-geeks who consume RNG schemas doing so by translating 
> them to XSD for production use (or using the non-normative XSD 
> variants of the standards)?
>
> - What about the people who use XML for “data” rather than 
> “documents”? Is RELAX NG just another fault line in that dochead / 
> datahead schism, or are people finding ways to bind objects and 
> databases to XML via RELAX NG as well as with XSD?
>
I hope Michael gets beyond "geekdom" and "dataheads/docheads" some day 
soon. They are cliches and less relevant. And they are dismissive, 
especially of system integrators. We are all XML Geeks now. Maybe we are 
all dataheads and docheads too. Even where there has been no 
interbreeding, there has still been a lot of furtive groping under the 
covers.

I've emailed Michael separately with information about a good-sized 
government project I know of that uses RELAX NG happily.
>
> - Do the paying customers of the world REALLY hate XSD that much more 
> than they hate XML itself, especially namespaces and the APIs? I get 
> the impression that “it sux, but it sux less than not having a data 
> interchange standard” is about the most love we can hope for from the 
> teeming masses who don’t subscribe to xml-dev.
>
Yes, everyone hates XSD. The thing is that XSD has become substantially 
*more* complicated in recent years, as Web Services have come along. Not 
only do people have to understand XSD, but their schemas may also 
partially be specified using wsdl:message syntax. (This lets you specify 
what the soapenv:Body element must contain, because the official SOAP 
schema is open. XSD's poor capabilities to support openness seem to be 
the root problem here.)

I teach XSD courses a lot. People like the courses and can do what they 
need afterwards, and are impressed by a lot of features, but no-one 
loves XSD.

> - Given the rather sad track record of “The Right Thing” technologies 
> in the eternal battle against “it sucks but you have to use it anyway” 
> technologies, why exactly should we “tear down the shaky towers with 
> XSD in their foundation”, as Tim put it? 5 years ago I wholeheartedly 
> agreed, but at this point it seems likely that XML itself will get 
> buried in the rubble if the XSD towers are pulled down.
>
My personal opinion is that XSD is a bad foundation not only because is 
so difficult and broken (and Michael's XSD towers are the ones vendor's 
have constucted: ultimately users say "that's your problem" and will 
vote with their feet with solutions that are easy and workable and 
reliable), but also because grammars are a bad foundation. Schemas 
should be based on paths, which are more straightforward to understand 
and implement, and which can do most of what XSD can do. (I have a blog 
article at XML.COM on converting content models to paths, so don't give 
me any crap about not being able to use paths for data binding.)

But if you have to use grammars, using the intersection set of features 
of XSD and RELAX NG where possible (see that XML.COM survey on which 
features actually get used in schemas) is still the best policy, where 
possible. I recently converted the Microsoft Office Open XML schemas 
(Office 12) from XSD to RELAX NG (working on an updated version this 
week) for ECMA preparation for ISO standardization, and I don't recall 
finding anything that couldn't be converted: some things are easier than 
others of course...swings and roundabouts.

Cheers
Rick Jelliffe