[GSDI Technical] Ideas on Mainstreaming SDIs

[Doug Nebert]

At the GSDI-10 conference these were several presentations that 
encouraged the SDI folks to think outside the box and to engage (be 
present) in mainstream Web computing. The support of 'special' catalogs 
and their interfaces arguably may serve a few professionals well but we 
are still invisible to the public and browsers who might like maps and 
data. Ed Parsons, CTO from Google, and a neo-geo advocate had some 
recommendations that may be helpful, as did Chris Holmes from the Open 
Planning Project (author of GeoServer WMS/WFS/WCS). I provide an 
abridged version of some of the points I heard presented at GSDI-10.

Some of the observations:

Keep it Simple. Web users expect intelligent search based on minimal 
input. Engines that return the 'right' ranked responses (and of a large 
number of possible hits) are popular. Advanced query might be available 
but simple query (textbox) must work well enough.

Use the map to help localize search. Instead of requiring the explicit 
definition of a location of interest, allow simple search to be 
constrained by the current geography within-view. It can impose/insert a 
spatial query based on the view extent to help filter the results.

Minimize metadata. It's against human nature to collect so much 
structured information as a prerequisite to publication. The suggestion 
was to lower the barrier to participation by focusing on very few 
structured fields (e.g. spatial/temporal) and allow encourage search 
against keywords, fields, and all indexed words. I still think we should 
encourage the progressive capture of metadata where it is known and 
valuable, and especially where it can be automatically derived/entered. 
If one has full metadata, allow that to be the initial discovered 
resource and to help refine end use of the ultimate described resource. 
The metadata also needs to be presented also in easier-to-understand 
ways perhaps through delivery of progressive detail, to engage novice 
and expert consumers. Clever stylesheets need to be written for 
geo-metatdata to facilitate this.

Expose resources using Web 2.0 technologies. In other words, don't 
invent fringe protocols that keep geospatial SDI communities separate 
from visibility to most web citizens.

- Interface with existing search engines -- expose metadata for data and 
services to mainstream search engines;
- Support catalog search on deep or volatile collections through use of 
Web search APIs such as OpenSearch that have native support in all 
browser toolbars.
- Use RSS (Really Simple Syndication), and in particular GeoRSS, to 
advertise the availability of new resources with a geographic extent. 
Again, browsers and portals have built-in RSS readers that can exploit 
these feeds and keep potential users informed.
- Enable the community with collaboration capabilities (within the 
recognized constraints of government systems security) to include 
member-offered ratings and reviews and linkage of other resource types 
to geography (mashups).

Following these ideas would allow the community to be more broadly 
visible, be accessible through hundreds of running software solutions 
including major search engines, and to more readily be linked through 
existing programming interfaces that would facilitate mashups of all 
types of available information.

How could SDI adapt and promote itself into mainstream GeoWeb?

1. Metadata catalogs could expose GeoRSS (http://georss.org) feeds that 
date-tag all entries. With some searchable metadata elements in the 
RSS/Atom format, these entries could be ingested by commercial search 
engines for broad public access.

2. Metadata catalogs could expose an OpenSearch (http://opensearch.org) 
interface that enables simple styles of search via URL with the response 
being, no surprise, GeoRSS/Atom format. Browser search toolbars can then 
directly access these services with no modification. This would be 
particularly good for large, specialized, or changeable catalogs of 
geospatial data and resources, such as national geo-catalogs. Perhaps 
OGC Catalog Service 3.0 could specify OpenSearch as the "baseline" CSW 
search and retrieval capability that all catalogs must support, though 
they may also provide more advanced search in profiles.

3. Web Map Services are not easily discovered though they may be very 
useful as overlays in map mashups. KML (now an OGC Best Practice and V3 
to be a candidate standard) can wrap references to WMS services and 
layers so they may be discovered and viewed in clients like Google Earth 
or ArcGIS Explorer. I recommend that we devise a way to re-process 
essential information from a WMS GetCapabilities request into a KML file 
with an entry for each layer available. These KML files could then be 
registered with search engines directly or could be referenced by a 
GeoRSS/Atom feed entries, and then be found and displayed by capable 
mainstream clients.

These are some ideas I'd suggest that we pursue, demonstrate, and 
document within the GSDI technical community and related standards fora 
in the coming year. I'd like you to share your ideas and thoughts with 
this list, and if you are not on the list, consider joining: 
http://lists.gsdi.org/mailman/listinfo/technical

Doug.
-- 
Douglas D. Nebert
Staff Geospatial Technology Advisor, System-of-Systems Architect
FGDC Secretariat   Phone: +1 703 648 4151   Fax: +1 703 648-5755