[Geojson] : Metadata ?

[Sean Gillies]

Christopher Schmidt wrote:
> On Tue, Oct 02, 2007 at 08:35:44PM +0200, Guillaume Sueur wrote:
>> Aren't you afraid that a data from which you don't know the origin, 
>> definition, reference scale, precision and so on is a kind of ... useless ?
> 
> Nope.
> 
> To take another example: 
> 
> I have a shapefile from MassGIS. It comes with attribute data. However,
> other than the datatypes, I don't know much about it: what is it of?
> what do the attributes mean? etc.
> 
> However, there is a wonderful webpage where I can look all this
> information up. It tells me who maintains the data, what the attributes
> mean, what the projection is, etc.
> 
> Does this mean that the shapefile is useless? Not to me -- it just means
> that I need additional metadata before I put it to its full use.
> 
>> I work on several governement projects in France and the mainstream is 
>> to reference and validate the data accessible on the web. 
> 
> A sane policy.
> 
>> If GeoJSON 
>> aims to replace WxS protocols I think there's something to consider 
>> here. No ?
> 
> GeoJSON is a data exchange format -- *not* a protocol. In the same way
> that HTML is a data format -- the protocol it rides on top of is HTTP.
> HTTP offers additional information about the data -- content-type,
> encoding, language, etc. A protocol for doing exchange of GeoJSON data
> would need to mimic these types of behaviors as they apply to geo. 
> 
> Depending on the protocol, this might be a reference to the schema in
> the HTTP headers. It might be a "schema_url" property on the
> FeatureCollection. It might be any number of things -- all above the
> level of the Feature Exchange mechanism that is GeoJSON. 
> 
> There are a couple people out there with thoughts about schemas for JSON
> data. I'm not among them. I've never used an XML schema -- I've even
> tried a couple times, and failed. I don't see GeoJSON's lack of schema
> as any more limiting than the lack of a schema in shapefiles: the
> benefit is that JSON data is extremely extensible, so if you want to
> exchange that data in the format, or alongside the format, or in a
> completely seperate format, you have all those options available to you.
> 
> It's important to realize that GeoJSON is not a format. It's a geometry,
> and feature exchange language. Describing what a feature is is a great
> candidate for another level of specification -- one I have no need for
> or interest in, unfortunately for those who like my heavy-handed
> dealings with specification creation :)
> 
> Regards,

+1 on auxiliary schemas.

James Clark's TEDI

http://blog.jclark.com/2007/04/do-we-need-new-kind-of-schema-language.html

seemed interesting, but there's been no follow-through.

Cheers,
Sean