Lists Home |
Date Index |
> At 02:30 PM 7/6/2002 -0400, Jonathan Robie wrote:
> >For data that is physically represented as XML, there already *is* a
> >layering - XML text is parsed to create an Infoset, which is validated to
> >create a PSVI, which is mapped into the XML Query data model. Do you
> >suggest a different layering? What layering would you prefer, and how
> >would this relate to the XML Query data model and typed operatoins?
> That's easy. Layer XML Query on the Infoset, drop the PSVI, and you have
> half a chance of producing something likely to have a long and useful
I'd like to suggest a different take on why there may have been such a hue
and cry about XQuery/Xpath2/XML Schema. I think it has a lot of explanatory
power. See what you all think.
Up until recently, most experienced XML practitioners probably felt that
they could write any part of the XML processing chain themselves if they had
to or wanted to. Anyone could write a parser (I'm not necessarily talking
fact here, but perception), and most people probably though they could write
an XPath processor if put to it. XSLT would be harder, but still one or a
small group could picture themselves doing it. We have Saxon from Mike Kay,
for example, and 4xslt from Uche's gang.
Going along with this, if you needed, say, XPath for Ruby or Curl, you could
imagine writing it if no one else got around to doing it. In this lies a
great sense of freedom. You could do it yourself, and especially you could
do it without needing the products of big companies.
Now comes XML Schema. This does not look like a one-person project (Yes, I
know, XSV - I said it was about perception). This is a big deal, if you
could even understand the Rec well enough. And to most people it probably
does not seem like an interesting job, either. To mix XML Schema into XPath
for XPath2 also seems like a great burden. XQuery seems too much to tackle,
too. SOAP toolkits - they are coming out of much bigger efforts.
If this is close to the mark, the resentment and fear comes from a perceived
withdrawal of the previous freedom. Any unclear or complex feature of, say,
XML Schema, will tend to trigger the reaction. So no amount of explanation
about any one issue can settle anything, which seems to be what we are
seeing here. The threads just keep circling around and repeating the same
Now personally, I do not see myself creating another RELAX NG validator.
But the next best thing to actually being able to build something is for it
to seem so clear and simple as to be quite understandable. So we keep
hearing how simple and elegant RELAX NG is. XML Schema can never be
anywhere near as understandable. If we can keep XPath2 simple - keep the
@#$@! PSVI out - we can still feel like we can do it ourselves if we have
I am not suggesting that there are no technical or architecture issues that
influence the discussion. I am suggesting that there are hidden aspects to
the debate, and theat they are fairly powerful.