Lists Home |
Date Index |
Dennis Sosnoski wrote:
> The problem, which I've expressed more than once, is to compare the
> performance for the alternatives of using text XML vs. some post-parse
> representation of XML documents. For the reasons given in my earlier
> email I'm chosing to base my timing comparisons on the parse event
> stream. This is in my opinion the fairest comparison to all
> approaches, since essentially any application working with XML is
> going to be using the output from a parse, not the raw document text.
> SAX2 is the most widely used parser API for Java, so that's what I'm
> using as the common event stream.
As you may remember from previous discussions, my approach in its ideal
mode IS working with the 'raw document text'. I think that this aspect
of my approach is different from what everyone else is proposing even
though we may share other strategies. Except when converting to/from
XML 1.0, esXML has no parsing and no serialization for a 'native'
application (i.e. one that operates in the most optimal, esxml-aware
mode). That doesn't mean you can't have a standard DOM or SAX
interface, but that esDOM is likely to be much more efficient.
To benchmark other approaches against esXML, one would have to look at
the overall process of input/access, create/output, or
input/process/output because the most direct analog to parsing and
serialization with esXML simply results in raw reads and writes of large
(tunable) blocks of data. I am trying to optimize the overall process
in a holistic way, a bit outside of the box in some ways.
By analogy, comparing raw SAX and DOM isn't fair because DOM is building
a data structure. You would have to compare SAX+data structure to DOM,
and even better SAX+data structure+test work vs. DOM+test work to get
meaningful data for a decision between the two. (I'm imagining a
business application here.)
We had a long discussion about meaningful test methodology in Santa
Clara, but I'm not sure that it was really captured as there was little
agreement. Many good points were made, IMHO.
> - Dennis
firstname.lastname@example.org http://www.hpti.com Personal: email@example.com http://sdw.st
Stephen D. Williams 703-724-0118W 703-995-0407Fax 20147-4622 AIM: sdw