Lists Home |
Date Index |
John Cowan wrote:
> Joe English scripsit:
> > We don't need mobile code. We need a reasonable XPointer
> > specification.
> I agree; my reference to mobile code was intended as a reductio ad absurdum.
That's what I figured. But I also think it reached
the point of absurdity about two steps before that,
right when QNames were used to identify (non-W3C) schemes.
> > XPointer was supposed to solve *one* *simple* problem: how to
> > locate part of an XML document and/or a specific position in
> > an XML document. And yes, this _is_ a simple problem. It's
> > been solved dozens of times. We just need a standardized way
> > to do it.
> Fine. You go and devise a method that is both maximally powerful and
> trivially easy to implement.
Sarcasm aside, I could devise one. So could you, and so could any of
the individual members of the Linking WG. Not trivially easy,
but a good deal simpler than the proposed framework and with
most of the expressive power. I suspect you would have, too,
if it weren't for the half-dozen other W3C WGs plus the TAG
applying pressure to add features and do things the "W3C Way".
The XPointer family of PRs look to me like a clear victim of
scope creep. They have left the 80/20 neighborhood and are well
into 5-95 territory -- that last 5% of the functionality that
takes 95% of the effort.
> > But then it gets worse: since you have to identify scheme "foo",
> > what better way, someone must have thought, than to Identify
> > it with a URI? So now you have to say "Using scheme
> > 'http://www.example.com/foo' locate element 'bar'" instead of
> > "Using scheme 'foo' locate element 'bar'".
> If you want to use a scheme that the W3C didn't standardize, yes.
> > We *really* don't need this. Only a rabid URI fetishist
> > would even consider embedding URIs in the fragment identifier of
> > a URL --and embedding them via QNames is just insane -- but
> > apparently the URI fetishists have a quorum, 'cause that's
> > what we got.
> Propose an alternative open system that doesn't involve creating a new
> registry just for the purpose.
First, I don't believe we need multiple schemes to begin with.
But if I'm mistaken and multiple schemes are an inescapable
necessity, an open system just makes matters worse. Henry S.
Thompson cites the interop nightmare caused by potential scheme
name collisions as the reason for using URIs to identify them.
I suspect that this will be an even worse interop nightmare.
How will implementors know which schemes to implement, and how will
users know which schemes are safe to use, if new XPointer schemes
can pop into existence on the Semantic Web at any time?
But if I'm wrong on that count too and (1) an open system is
required and (2) there can be no new central registry, there's
always the NIST identifier collaboration service:
<URL: http://ats.nist.gov/nics/ >
This would be a better solution than URIs on several counts.