On Tue, 2 Sep 2003, James Strachan wrote:
> > Note that my votes are not strongly cast... so if we do end up
> > debating XMLBeans vs betwixt vs digester then I'm going to
> > retract my vote and push for just dom.
> >
> > Note I still have not heard any real reason not to use dom - so
> > it is the obvious choice if we can't agree on anything else.
>
> Which dom :) W3C, dom4j, EXML, JDOM, XOM...
I have to say, we've got a specific set of schemas, and we're only
going to write this code once. Many of the XML mappers have kind of left
a bad taste in my mouth. If we use DOM -- the DOM that's included in JDK
1.4 -- then we have guaranteed compatibility, no library bloat, and great
speed.
That said, it's not quite true. We will need to adjust the
Geronimo DDs over time. But my experience has been that all the tools
need to be customized to operate the way we want, whether it's to change
generated class names or handle the mix of attributes and content in some
of the J2EE tags. I personally find it as easy to write the code to pull
stuff out of a DOM tree as to write the Betwixt or Digester rules, and
far easier to debug as well.
The one remaining obstacle is XPath, and we just have to decide on
the library bloat issue. It wouldn't be terribly hard to write the
minimal XPath processor we need (operating on the POJOs), but JXPath does
it already... Plus, it debugs well.
Aaron
