Tobias Reif <tobiasreif / pinkjuice.com> wrote in message news:<3BEFDF9F.30705 / pinkjuice.com>...
> What exactly to include in the standard lib?
> 

I personally dont see conformance to the various formal W3C APIs (DOM
etc) as
a great priority - I dont believe that there is any intrinsic reason
to process XML
internally using these (particularly given the generally ugly api and
mismatch with
Ruby patterns) and would avoid these unless there was no alternative
(if anyone
has a case where there is a benefit to using these I would be
interested to hear
the rationale) 

A more 'natural' Ruby binding (stream/tree parser and api) with
reasonable compliance
to standards (eg namespaces/encodings) would actually much more useful
for
actually doing anything and I believe that this should be the focus
whatever goes
into the distribution. 

Regarding performance (separate thread) - I would focus on performance
of the
tree parser rather than the raw stream parser as this is the most
useful from
a programatic standpoint - expat (or any native parser) will clearly
be much faster
than a Ruby implementation as a raw stream parser however from Sean's
benchmarks it appears that things are much closer when building a tree
- this
being the case a Ruby implementation looks more reasonable.

I would lean towards REXML (or something similar) as a basis for the
'standard' given the reasonable conformance and 'feel' of the API.

PaulC