[OSGeodata] Re: [Geotools-devel] On the catalog tutorial

Jody Garnett jgarnett at refractions.net
Mon Sep 25 22:29:09 EDT 2006


Stefan F. Keller wrote:
>
> Jody,
>
> Could'nt follow all details of your post but you hit one of the core 
> problems of modeling geo-information and services, namely 
> relationships, especially mutual relationship between these.
>
Yes that is a core problem, but one that is very accessible. I always 
get stuck on modeling associations between resources. You have been 
correctly pointing out that I miss out on some of this information (like 
"source").
>
> To me geo-information and services have a many-to-many relationship.
>
Check.
>
> When there is a containment relationship with data access services 
> containing geo-information (which is OK for an API) and a set of 
> containers you obviously can compute reverse associations to 
> re-establish this many-to-many relationship in a libary behind an API.
>
I am afraid we were not that nice, we simply return a list of handles 
that satisfy the query, the set of handles returned will all refer to 
the same data. But each handle (as they come from different services) 
may offer different abilities (from WFS data access, to WMS service side 
portrayal).
>
> As you know, I am trying to reach consensus about a minimal 
> information metadata model (to all: I'm still awaiting comments or 
> counter proposals to http://tinyurl.com/kfkyv). So I like to give my 
> 0.25 pennies to this open problem:
>
Will review now, I am afraid I am some weeks behind in reading email 
lists that are not part of my core functions :-)
> Having the use case of minimal metadata exchange - not the API - in 
> mind, I'm strongly proposing a containment relationship of 
> geo-information containing zero to many data access services (data 
> binding protocols).
Nice way to slice it, no argument here. May wish to extend it to refer 
to other associated information, so in addition to "data binding 
protocols" you can refer to associated "intrinsic" resources, such as 
FeatureType, Style and different flavours of metadata.

Sorry if that was not clear:
- my road collection
-- WFS data binding: 
http://localhost/geoserver/wfs?REQUEST=GetCapabilities+SERVICE=WFS+VERSION=1.0
-- WMS data binding: 
http://localhost/geoserver/wms?REQUEST=GetCapabilities+SERVICE=WMS+VERSION=1.3
-- ISO19115: ... could be an intrinsic entry in GeoNetwork containing 
traditional ISO19115 metadata
-- SLD: http://localhost/geoserver/wms?REQUEST=GetStyles+Style=road_style

Aka leave it open so you can go beyond data binding later ..., not the 
reference to the original source ISO19115 meta data for applications 
that can actually care.
> Consider also the fact that geodata can exist without data access 
> services but data access services can't.
We do have some odd balls, Web Processing Service, Feature Portrayal 
Service which seem to leave by data chaining alone.
> In addition, I'm making a difference between data access services (as 
> mentioned above) and 'services on it's own' like filter or label 
> placement services. Here I think there is no solution out yet because 
> these 'services on it's own' wouldn't contain associations to specific 
> geodata (= data access services) but would try to describe kinds and 
> structures of geodata they can process.
Point taken, to translate into OGC OWS speak the services you described 
have no "resource type" that they are responsible for managing. This is 
why a FeaturePortrayalService is not a subset of WMS, FPS has no 
resource it is responsible for managing.
>
> So to me, when looking for the best approach to auto-discover 
> and publish geo-information, I'd like to ask the following questions:
>
> * What's easier for webcrawlers: find data sets (or metadata about it) 
> or services?
So far services seem to be easiest, it is what Paul managed to do.
> * What do data providers more likely do (if they do): register data 
> sets or metadata about it containing some data access services or 
> services (containing many views to data sets)?
So far they seem to grab geoserver, fill in a little bit of information 
as they configure it to publish their information.
> * And finally: What is easier to describe in a minimal metadata model: 
> data sets or data access services?
Services seem easier, their is simply less information needed to 
describe them. When describing data there is a wider range of 
information that may be useful (and indeed the usefulness depends on 
what client application is trying to answer the question).

Cheers,
Jody




More information about the Geodata mailing list