RE: [xml-dev] Interoperability [long]

[bob mcwhirter <bob@werken.com>]

> Sorry, I was more crabby and obscure than usual yesterday; I had just
> returned via a red-eye flight from a DOM WG meeting in which we took yet
> another whack at namespace normalization and in matching up the DOM data
> model with the XPath data model.  The combination of DOM, namespaces, and
> XPath is likely to sell quite a bit of Prozac in years to come, I fear.
> 
> In hindsight (and in my personal opinion), the DOM probably should have
> tried harder to directly expose the prefix that a node was associated with
> when it was created or parsed; the XPath data model is somewhat cleaner in
> this regard.  On the other hand, by a strict reading of the Namespace spec,
> we concluded a couple of years ago that this was un-necessary and
> potentially inefficient. 

Anecdotal story...

Folks using Jaxen pretty much *always* get confused the first time
they create an XPath like /foo:bar/baz:cheese, and then get nothing
selected from their document.  There is definitely a lack-of-understanding
of how XPath ns-prefixes map to Document ns-prefixes.  Since pretty
much everything is prefix-oriented, no one realizes that you have to
bind prefixes to uris in the context of the XPath itself, not just
the document. 

An example of the prefix-orientedness of the world, many articles and
books define the URL for 'xsd' a single time, and then all examples
throughout the book only ever use the 'xsd' prefix.  Folks get hung up
on the prefixes, and forget there's a ns-uri behind the scenes.)  
Likewise with 'soap' or whatnot.  I think the literature, at lease 
this year and next, could do a better job of emphasizing that the
prefix really isn't the data, but the URI-behind-the-prefix is
what's important.  Prefixes are just syntactic sugar, really.

Just my $0.02.

	-bob