[OSGeodata] discovery requirements

Kralidis,Tom [Burlington] Tom.Kralidis at ec.gc.ca
Tue Aug 8 15:49:42 EDT 2006


Comments prefixed by TK: 


________________________________

	From: Stefan F. Keller [mailto:sfkeller at gmail.com] 
	Sent: 05 August, 2006 7:58 AM
	To: geodata at geodata.osgeo.org
	Subject: [OSGeodata] discovery requirements
	
	
	Tom suggested stepping back: One of our requirements was that we
not only want to register a static resource but also web services (e.g.
WxS). My dilemma is that services are not really the resource users are
looking for! What's really the goal of dissemination is the geographic
data, e.g. raster maps or the geographic feature object sets or the
coverage data. 

TK: I agree about the importance of data.  Having said this, the
dissemintation of the data is (additionally and increasingly) put forth
via web services (i.e. no static links).  So our next step would be to
step back again and thinking of requirements in terms of defining who
the "users" are.
	 
	Either a harvester must identify this geographic service and try
to request it for more information based on the indicated 'proprietary'
protocol (WxS GetCapabilities, SOAP, etc.) or every single geographic
data resource needs to be registered once for every service, be it a
'WMS layer' or a 'WFS feature set' etc. But isn't this 'duplication' the
same with file formats, like data delivered as GML, SHP, DXF (similar to
DOC, PDF, HTML). So: Isn't this a possible solution? 

TK: I'm not sure what you mean.  I would envision a plethora of web
services (OWS instances, static GML/SLD/WMC docs, standalone ISO19115
metadata, etc.).  From an enterprise level, I wouldn't publish any
static resources if they were being provided through web services.  So
these services would be published to a catalog once, and their
layers/feature types/covereges (within them) would be registered as
same.


	-- Stefan

	P.S. I tried to model this as value 'WxS service...' (tbd.) in
Dublin Core (DC) and found attribute/DC element Type appropriate (see
http://www.geometa.info/rappiinfo/wiki/index.php/OSGeodataMetadataModel
<http://www.geometa.info/rappiinfo/wiki/index.php/OSGeodataMetadataModel
> ). In terms of metadata records the two variants mean, that in case we
register 'service resources', we have one record entry per WxS
containing many hidden(?) geographic data resources. I prefer to
register one 'data resource' record (per data provider) where Type
values are repeated (in DC every attribute can be repeated AFAIK), one
entry for every available file format and one for every service (WxS).
If this seems impracticable, there exist also hierarchical sets in DC. 

	This dilemma becomes even more evident when different SLD's
offer different 'views' on the same data or when future conversion
services exist! But still: To IMHO there are too many open questions
regarding service description (quality rating as Paul mentioned?) and
service chaining. This does not mean that registering these types of
services is prevented in the solution I suggest. 
	






More information about the Geodata mailing list