Lists Home |
Date Index |
My XBIS performance results can be viewed at
http://xbis.sourceforge.net/performance.html The tests include 3
specific medium-sized (100-200K) documents of very different types, a
pair of large (2.9 and 4.0M) documents, and several collections of small
documents. The documents cover the full range from narrative style (the
XML specification with entity definitions brought inline) through
conventional data style (your own periodic table example) through SOAP.
None of the documents involve any external entities or even extensive
internal DTD subsets, and no parser validation was used.
The tests also compare XBIS with both Piccolo and Xerces parsers,
running on both Sun and IBM JVMs. I haven't checked each individual
input file in detail, but offhand it looks like for every document XBIS
is *at least* 4 times as fast as the faster of the two parsers on each JVM.
I'm not sure what you're looking for in the way of "convincing
evidence", Elliotte, but you're welcome to download the XBIS code from
the site and try it out yourself.
Elliotte Rusty Harold wrote:
>At 10:59 PM -0700 9/23/03, Dennis Sosnoski wrote:
>>It's clear that transformations of XML into alternative
>>representations can result in substantial increases in performance
>>over conventional text parsers (at least for Java), and can also
>>result in substantial reductions in document size.
>People keep claiming that, but I've yet to see convincing evidence of
>it. At most, I see benchmarks based on one or two documents with very
>particular structures. I've never seen any data for these binary
>formats based on a broad range of XML documents of varying sizes.