Lists Home |
Date Index |
Elliotte Rusty Harold wrote:
>At 4:10 PM +0000 11/23/03, Michael Kay wrote:
>>Others may disagree, but my perception is that Xerces is bigger, slower,
>>has more features and fewer bugs.
>I'll second that. By my count, Xerces-J currently has one bug in
>basic XML parsing. (It doesn't report default attribute values from
>the internal DTD subset in invalid documents.) Its nearest
>competitor, the latest Oracle beta, has about ten such bugs. All
>other parsers I've tested are worse.
I think the Xerces parser is excellent. Unfortunately the DOM
implementation is tainted by their use of "deferred node expansion" by
default. This works reasonably well for larger documents but adds a
hefty overhead to building the DOM for small documents. It also has the
unfortunate side effect of making read-only DOM access unsafe for
multithreaded use (since reading the nodes will actually alter their
internal representation). I haven't checked if Sun left this enabled for
the JAXP bundling of Xerces, but assume they did.