Lists Home |
Date Index |
On Fri, 2005-11-04 at 17:20 -0800, Michael Champion wrote:
> Thanks, I'd heard rumors about this presentation but hadn't seen it. Really
> thought provoking ... especially if contrasted with Adam Bosworth's almost
> diametrically opposite position expressed in
> The way I (perhaps crudely) would characterize the positions are that Adam
> thinks that XML/RDF is too complex to be useful, so "worse is better"
> things like HTML for text and RSS for data will become the mainstream
> technologies that are more derived from XML than really applications of
> generic XML. Dana agrees that XML is too hard for mainstream developers to
> process, so we need better tools and some fixes to XML to make it more
> I find myself more in Dana's camp than Adam's, but believe that XML/XQuery
> features will be absorbed into programming languages rather than vice versa.
> I do like her points about making XML more graph-friendly with links as
> first class citizens, but the implications could get messy ....
I'm surprised to hear you say that. That is also my position, as I
think I've yelled often enough (sometimes at you ;-) ).
* I think that XML can and should be processed with as much
declarativity as possible. I think this is a fundamental facet of
data-driven app design .
* I think that XQuery gets declarativity right, and almost everything
else wrong. I think it will always have a healthy niche, but I don't
believe it will ever trump application languages for apps processing.
* I think that the best features of XQuery are already becoming
available in general-purpose languages, and that XSLT 1.0 was the key in
bringing this about, not XSLT 2.0 and XQuery. In particular, XSLT 1.0's
declaration through patterns and trigger functions (templates) is a
powerful mechanism that is migrating beyond that DSL (e.g. my work on
Python/Amara, some work I've seen in Java from Eric van der Vlist and
others, and if I'm understanding them properly, Comega and XLINQ).
* I think that RDF can be thought of as a declarative tool for semantics
to go with declarative tools for XML syntax.
* I think that RDF gets the declarativity right, but is losing its
chance to get everything else right. It should be focusing on making
simple assertions dirt simple for folks to sprinkle around, and only
after we get a critical mass of such assertions should we worry about
sophisticated model issues.  I think that in emphasizing a model
that only professors can understand, RDF is making a similar error to
XQuery (XQuery people claim that their model is not over-emphasized, but
I don't believe it for a second, and watching Mike Kay talk about XSLT
versus XQuery heavily reinforced my impressions).
What's amazing to me about the self-inflicted wounds in XQuery and
(increasingly) RDF is that the Web domain is so ripe for what they are
offering right now, from Web feeds everywhere to the tagging/folksonomy
mania. If they could just KISS, they'd be no-brainers, but they're
anything but. For my part I continue with RDF because I still have some
belief that there is a super-simple subset, untainted by Semantic Web
big-think, and that I can maybe use that for lightweight formalization
of the semantics that I want to express in XML (preferably leaving
ugly-ass RDF/XML out of the matter as much as possible). I never took
up with XQuery for a moment because I could never perceive such a
right-sized kernel at its core.
> Her points about declarative processing of XML are particularly interesting,
> but contrast them with Anders Hjelsberg's discussion of the evolution of
> mainstream languages to be more declarative in
So the the presentation tells me that I must have IE5+ or Netscape 7+.
I have Firefox 1.0.7, what gives? (If I get past the warning, I'm
afforded a blank page.) I guess there's a warning against
any idealized notion that we can move beyond browser wars to
considerations of the sorts of matters Florescu, Bosworth and Hjelsberg
would like to dwell on. Bah!
Uche Ogbuji Fourthought, Inc.