Lists Home |
Date Index |
1/7/2002 8:39:30 PM, Ronald Bourret <email@example.com> wrote:
>Maybe I haven't been reading this thread (or the DOM spec) closely
>enough. Is this saying that DOM 2 doesn't automatically insert namespace
>declarations as needed? That is, users have to add xmlns attributes by
Yup. The DOM essentially presents a *syntax* view of an XML document.
(Remember, it preceded the InfoSet, Namespaces, and XPath by some
time). This is an EXTREMELY frequently discussed issue in the DOM WG.
The dominant view has always been that DOM user essentially has to do
what someone editing text by hand has to do -- figure out where the
namespace declaration attributes should go and put them there.
It is indeed a "big honkin' mess" but it's not clear in my mind whose mess it
is or who can clean it up. I must confess that after several years of
discussions of how to define an API that is simultaneously compatible
with XML syntax, DOM Level n-1, the Namespaces Rec, robust when things
are added to or moved around in the tree, minimally inconvenient to XPath users,
implementable without armies of programmers and testers ... I simply curl up
into the fetal position and whimper softly when the subject comes up.
Anyone who has some great ideas how to do all this (within the additional
constraints of W3C politics, of course) can be an Invited Deity in the
DOM WG as far as I'm concerned. The best we've come up with so far is
the "superdupernormalize" in DOM L3, along with some parse-time options to
throw away stuff that the InfoSet doesn't represent and some methods to
present an XPath-like view of a DOM tree.
When the going gets tough, the tough refactor?