[
Lists Home |
Date Index |
Thread Index
]
Bob Foster wrote:
> Fastest doing what? Does the parser need to implement some API in some
> language on some set of platforms? Does it get to define its own API
> and pick a language and platforms?
Excellent question.
I'd leave the language open to the competitor. I'd leave the API open
too, since the aim is to encourage experimentation. However, it would
make benchmarking easier if the processor used an existing API: less
harness code to write. Also, except for SAX and DOM, there is such a
lack of standard APIs that mandating any particular ones seems useless.
The other issue is whether you want to have one contest for
XML-to-stream and a second contest for XML-to-tree. Personally I'd go
for just XML-to-stream to keep it within the reach of single
developers. (XML-to-tree could be a subsequent competition.)
On your issue of N(O^2), I think Rick is referring to a potential
problem of XML-to-Tree implementations, or maybe XPath etc traversers,
rather than XML-to-stream processors necessarily.
Cheers
Rick
|
