[jdom-interest] Interesting XML benchmarks (inc JDOM)

[Tom Bradford]

Jason Hunter wrote:
> Remember, Xerces is on its third release and has been heavily
> performance optimized.  JDOM's just starting to have a set API where
> it's worth doing performance work.
> Here's our philosophy:  First define an API where it's possible to be
> thin and fast.  Then implement that API thin and fast.  If you tested
> what's in CVS versus beta6 I think you'd see a dramatic improvement.
> The ArrayList change alone shaved significant amounts of memory and

Any API is possible of being thin and fast.  Including interface based
factory driven APIs.  The arguments that I had always heard were that
factories are slow and awkward and interfaces lead to more problems than
they solve, along with the importing problems, etc.  It seems like all
of the initial design for JDOM was driven by assumptions of performance
and ease of use, rather than extensible design, which was looked
negatively upon.

> build time.  I'm sure there's more to be done, and anyone with interest
> is welcome to look for bottlenecks.

I know.. I have.

> It's no different recently than it's ever been, for better or worse.  We
> try to maintain the original vision, but are always happy to look at
> innovative ideas.  Much of the value in the API comes from contributions
> from this list.

My experience is that, in the past, anyone who's ever suggested patterns
and practices that were out of line with the authors' original vision of
JDOM was met with immediate resistance and condescension.  This HAS
changed in the past couple of months, where now, a lot of those ideas
that you initially turned your back to are finding their way into the
API.  Some of these ideas are even coming from Brett now.

Ok, I'm done.. Jeez, try to give someone a complement.

-- 
Tom Bradford --- The dbXML Project --- http://www.dbxml.org/
We store your XML data a hell of a lot better than /dev/null