RE: [xml-dev] Patent on streaming evaluation of XPath

["Michael Kay" <mike@saxonica.com>]

> This patent troubles me. It is sits squarely on a key 
> requirement for high performance XML processing. I think it 
> will trouble some of you, too. If anybody understands how 
> prior art can be submitted for this (these pages are unclear 
> about whether the patent is granted), I would very much like 
> to hear from you, and so would the USPTO...


There's certainly been a vast amount of published work on streaming XPath
evaluation, including published papers from within Oracle that substantially
pre-date this patent application.

I'm afraid I really don't understand the patent system or its arcane
language well enough to know how broadly or narrowly defined the claims of
this patent are. Presumably if it's very broad, covering almost any way of
doing streamed XPath evaluation, then it's invalid because of prior art;
whereas if it's very narrow, then an independent implementation of streamed
XPath evaluation is unlikely to run foul of it. 

In practice of course that's not the issue. Patents like this are not
designed to be defensible in court, they are designed to put the frighteners
on people and to exhort protection money.

Saxon has implemented streaming of the subset of XPath used in XSD integrity
constraints since 2004. I'm sure other XSD validators like Xerces and MSXML
must have done the same. But again, I don't know enough about the patent
system to know whether the existence of a product counts as prior art - my
layman's understanding is that a product isn't enough, there has to be a
publication.

Saxon's streaming XPath is very different in detail from the approach
described here: though at one level, I guess all approaches to streamed
XPath evaluation are in some sense isomorphic. But I've no idea whether
isomorphism has any meaning in the patent world. That's one of the problems
with software patents, and one of the reasons they shouldn't be allowed.

The patent doesn't seem to discuss the question, which I would have thought
rather crucial, of how large a subset of XPath it can handle.

Michael Kay
Saxonica