Hello --

On Fri, 23 Nov 2001, Dave Thomas wrote:

> Tobias Reif <tobiasreif / pinkjuice.com> writes:
>
> > > I totally disagree. There are plenty of cases where the
> > > same app is responsible for both writing and reading all
> > > the XML.
> >
> >
> > That's the point; in all other cases, a non-conformant
> > implementation is useless.
>
> Perhaps - it also depends to some extend on the failure mode. If the
> library silently ignores what you tell it to do, or silently writes
> bad output, then that's clearly dangerous.
>
> If instead it throws an exception when you use capabilities that have
> not yet been implemented, then you can use it safe in the knowledge
> that you're not going to produce bad output, or incorrectly interpret
> input. Most of the world's XML applications use only simple subset of
> XML to start with. I suspect these will be well served by even
> non-conforming libraries.

We may be talking about two different things:

  1. an XML parser that implements a subset of what the 1.0 spec asks
     of it (and documents its own non-handling of other things);
  2. an parser that does things that violate the spec.

In other words, sins of omission versus sins of commission.

It's the latter I'm principally concerned about (since that has the
potential to completely isolate a given parser from the reality of
XML).  But maybe we're really talking about #1?  In which case... It's
understandable that an XML tool might be released and even used a lot
before it handles everything (as long as it doesn't pretend to, as you
say).  But in the medium-to-long term there's no reason not to aim for
parsers that do handle the full spec.

(Easy for me to say, as I haven't joined the ranks of those writing
them :-)


David

-- 
David Alan Black
home: dblack / candle.superlink.net
work: blackdav / shu.edu
Web:  http://pirate.shu.edu/~blackdav