Santa Barbara, California. Dec.3-4 1997
Mission Report
0. Distribution List
R. Winter
J. Meyer-Roux
G. Schmuck
A. Annoni
Heads of Units of SAI
R. Peckham ISIS
U. Boes DG III
M. Littlejohn DG XIII
D. Riese GISCO
1. Subject
The first ever international conference devoted to interoperability in GIS.
2. Participants
About 80 academics, representatives from OGC (OpenGIS Consortium), NIMA, USGS, NSF,
From JRC: Sargent (SAI)
3. Aim of the Meeting
Academic conference: presentation of papers and plenary discussion. Attemt to decide the academic research agenda in this area. Selected proceddings will be published in book form in July 1998.
4. Report of the Meeting
This conference on Interoperability in GIS is the first in what may be an important series. Because of the strong European presence at this conference, it is likely that there will be a follow-up conference in Europe, probably in Zurich, in 1999.
There were 2 full days starting at 08:00 with up to 4 parallel sessions. The existing pre-meeting abstracts and some full papers are available on:
http://bbq.ncgia.ucsb.edu/conf/interop97/
Many of the more intriguing papers came from individuals pursuing individual research topics, but there were also institutional presentations given by the following organisations:
· The President of the OGC: The OpenGIS Consortium.
· A representative of the US national civil mapping agency: USGS.
· The head of research grants in GIS at the US National Science Foundation.
· The head of GIS at the US National Defense University
· A representative of the US military mapping agency: NIMA
· The Director of the Large Scale Information Systems Lab. at the University of Georgia.
which, for a conference with 88 attendees, gives an idea of the importance and timeliness of this meeting.
Apart from the addresses presented in plenary; including the final summing up and discussion on the conference as a whole, I attended presentations, took notes and participated in discussions, of the following papers:
· Interoperability and Spatial Information Theory, Andrej Vchovski (Zurich)
· A Specification Language for GIS Andrew Frank (Vienna)
· GeoToolkit: Opening the Access to Object Oriented GIS Oleg Balovnev (Bonn)
· Interoperability of Geographic Information: Spreadsheets and Virtual Environments Pedro Pereira Goncalves (Lisbon)
· The Geospatial Interoperability Problem: Lessons learned from the GeoLens Prototype Cliff Behrens (Bell Labs., USA)
· The Open Geographic Datastore Interface (OGDI) Paul Morin (Ottawa)
· GEOLIB: A Software Component for GIS Tools Donatas Kverdaraukas (Tolouse)
· Implementing the OpenGIS SImple Features Interface Scott Morehouse (ESRI, USA)
· IRIS: A tool to support data analysis with maps Gennady Andrienko (German National Research Centre)
· Information Brokers for a Web-based GIS Ian Finch (Liverpool)
· The Use of Functional Programming in Specifying and Testing Interoperating GIS Werner Kuhn (Munster)
· Geospatial Modelling: A case study for a statewide land information strategy David Pullar (Brisbane)
· Using the Internet: an OpenGIS Prototype Frederico Torres Fonseca (Belo Horizonte).
Note that just two of these were from the USA, and three from Australia, Canada and Brazil. The remaining eight were from Europe -including one of only two papers on OpenGIS standards implementation.
There were also some papers on the implications of geospatial modelling, especially enviromental modelling, for interoperation and also specific needs for Transport GIS. I was not able to attend these because they were in parallel sessions.
OO GIS, Web-Based GIS and Universal Server GIS
There were a great many papers using object-oriented GIS, distribute computing and web delivery of geoinformation; as a side aspect to the central topic of the paper. GeoToolKit (Bonn University) was the leading OO system, NovaGIS (Nove Lisboa) used OLE/COM, SDBC (Seoul) used CORBA, GeoLens (Bellcore) metadataproject used RDF/XML and web delivery, Prodabel (Belo Horizonte) and SIGAL (Laval) used Java and web delivery. A prototype using KQML (a NASA-developed distributed expert system query language) in conjunction with Java and web-delivery showed strong promise for metadata negotiation (Liverpool).
ESRI (Scott Morehouse) presented a paper on the use of Oracle's Spatial Data Option with OpenGIS and SDE.
Standards and Interoperability
Most standardisation activities are not proper subjects of academic research. Standardisation is only appropriate for mature areas where the essence of the problem is understood and where only the accident of divergent history needs to be overcome.
Andrew Frank showed that integration is a multi-layer problem. The bottom issues are accident and subject to standardisation. There are research issues in the top laters. Progress, however, is from the bottom up. We manage to effectively standardise one layer every two years. Problems effectively solved at the bottom include ASCII/EBCDIC, Ethernet, TCP/IP, SQL, CORBA, COM, HTTP/HTML etc. These now form the environment within which we work, whereas in the past they were percieved as problems in their own right. Some "progress" is retrograde for some operations, e.g the abandonment of mathematically-precise relational entity relationship modelling in favour ob object-oriented design now means that some operations are inexactly described.
Central Theme
The major result from the conference was a new appreciation of the essential need for semantic representations and metadata communication in order to communicate effectively.
Semantic communciation involves moving geodata between contexts, so some formal means of representing such contexts is required before any thought can be given as to how to translate the data, given a description of both source and target contexts.
Max Egenhofer raised these questions:
· When are two things the same ?
· When are things similar enough to integrate them ?
· When can things be integrated ?
· What semantics are in common and what is domain-specific ?
· What solutions for interoperating GIS will have an impact on domains beyond GIS ?
Although not obvious from the titles of the papers, many discussions turned to the difficulty of making independently generated geographic datastores interoperate or present a common interface ("federate"). This invariably involves bridging slightly different meanings of almost but not quite identical concepts which may be identical in the real world, but which have operationally different constraints within each specific GIS. Thus the papers on metadata and the extensive lecture by Amit Sheth (Georgia, USA) on Federated Databases were extremely to the point.
Andrew Frank said that effects of formal GI semantics will be to:
· decrease the size of GIS systems.
· increase useability (cheaper, more elegant)
· improve education (reduced self-inconsistencies, reduced pointless incompatibilitites between GISs)
He also noted that the largest software systems on the planet are Oracle Financials and SAP's R3. These are not remotely interoperable and noone expects them to interoperate. The conclusion is that in some areas interoperability is not a problem because it is not necessary, therefore academics should not try to solve problems that do not affect anyone in practice.
This central issue was so strong that one attendee even raised the question of whether spatial data would fade into the background as the 'defining aspect' of GIS, to be replaced by semantics and metadata. The reasoning was that in the past we were constrained by technical difficulties in representing and indexing geopmetricl and spatial reality. These are now largely solved for most people's practical problems, but we are constrained by how we can deal with a specific type of semantics. If a discipline is defined by the problems it addresses and the techniques it uses, then this is the apparent direction. There was minority suport for this as a valid point of view.
Academic Directions' Conclusions
The academic teaching impact was not neglected. Derek Reeve asked:
How do we get these semantic concepts into out GIS courses ?
How will GIS teaching change ?
Should GIS move from geography to computer science ?
Should GIS move from geography to business schools ?
Who needs to learn GIS ? Software developers or geographers ?
Who does not need to learn geographical analysis ?
The attendees left the conference with an appreciation of the scale of the problems faced and new insight into a central issue which had not been so widely known beforehand.
The work of Kuhn and Frank on specification using the GOFER language takes a formal and rigorous approach aimed directly at the central issue: definining what is meant in a GIS and how it works so that another GIS can use the information. GOFER appears to be a way in which we can begin to develop "unambiguous and portable statements of meaning" can be made for both data and operations, and also for contexts.
Andrew Frank stated that the current frameworks and tools require improvement before some research can begin, e.g. we have no bases of measurement or even derived or proxy metrics for some things we wish to measure.
Once conclusion was that many of the intrinsically difficult (essence) aspects of problems facing GIS interoperability would not be solved by the GIS community. This is because they are generic problems which will be solved by computer science research and software technology, e.g. metadata communication languages, formal specification of constraints, and semantic interoperability architectures. This implies that the GIS community must keep up to date and be ready to adopt new technologies as soon as they begin to be usable, e.g. the RDF/XML standard from the W3 consortium which will be available outside W3 only in January 1998.
Philip Sargent
Space Applications Institute
TP 950
Philip.Sargent@jrc.it