[Geowanking] RDF vs GML in Open Source vs Google

[Chris Holmes]

Quoting Chris Goad <cg at mapbureau.com>:

> Hello Chris (Holmes),
>
> > I'm not sure _why_ I want to inject information into the innards of
> a
> > GML file?  I think I'd prefer to visualize them, maybe modify some,
> and
> > maybe perform some analysis.  If I run my analysis, finding
> patterns in
> > crime statistis and desert turtle migrations, then I'd create new
> > geographic information as a result, no?  I wouldn't inject it back
> into
> > the original files.
>
> Ok,  in this scenario, as you are working along with this melange of
> geo and
> turtle-specific data, are you maintaining the geo data separately in
> GML,
> and turtle specifics in some other format?  Of course this is
> workable, but
> fails to standardize the way that the non-geographic content is tied
> to the
> geographic content. If this isn't what you mean - if you mean GML to
> be the
> central representation for all of the data, geo or not,  then we do
> have
> what I was calling "injection" of other kinds of data into GML (even
> if not
> the original files), requiring a turtle application schema.
Ah ok.  I don't _think_ GML was ever meant to be much more than a
transport language, at least I only think of it as transport.  To be
converted to your local gis, or to display on the web, ect.

> With  the automated machinery that you refer to,  application schemas
> no
> doubt  become much less of a burden. But the direction being taken
> here is
> one of absorbing every application area that impinges on geography
> into GML;
> ie GML is inevitably being positioned as a general data
> representation, in
> effective competition with RDF.  I am convinced that RDF  (+ OWL,
> SPARQL,..)
> is the  better technical candidate for this general task, because it
> is
> simple, clean, and built from the outset  with this generality in
> mind. The
> better technical candidate does not always win however!
Ok, _this_ makes a lot of sense to me.  And I see what you're saying,
GML trying to do too much, because it ends up having to, with all the
different communities, defining all their own domains, and doing so
with something that wasn't truly designed for the task.  We've recently
been getting into addressing specific domains with GML, and if RDF is
really all it's cracked up to be, then I see how it's the superior
choice.

I guess what would truly convince me (like make me a convert) is seeing
real RDF applications doing the types of mixing that I hear so much
about.  Like I think I'm convinced on the theory, but have trouble
seeing the practice, how it's widely useful.


>
> >   If you'd like you're more
> > than welcome to hack GeoServer to return RDFGeom2d as an output
> format,
> > and I imagine MapServer would probably let you do the same
>
> Yes, this sounds like a good direction - thanks.
Let me know if you need help, I can definitely point you in the right
directions.  Anything I can do to encourage good open standards - we
went with GML and WFS since it was the best when we started, and
they're naturally linked.  But if we can get an RDF one with a wide
community behind it that is really technically a lot better, then I'm
happy to promote it as much as we can.

best regards,

Chris

>
> -- Chris (Goad)
>




----------------------------------------------------------
This mail sent through IMP: https://webmail.limegroup.com/