Ilya Zaslavsky, Ph.D. (1995, University of Washington; 1990, USSR Academy of Sciences) is an Asst. Research Scientist at the San Diego Supercomputer Center, UCSD, specializing in Geographic Information Systems, spatial analysis and syllogistics, and Web mapping interfaces. Before joining SDSC he was a Staff Scientist at San Diego State University, and prior to that, an Assistant Professor at Western Michigan University, Department of Geography.

Richard Marciano, Ph.D. (1992, University of Iowa) is a Research Scientist at the San Diego Supercomputer Center, UCSD. He has also served as a parallel processing and environmental science consultant / researcher. He holds advanced degrees in Computer Science, Electrical Engineering, and Avionics Engineering. His research interests include: integration of GIS and historical data, parallel processing of spatial analysis, information mediation of geospatial data repositories, and geospatial markup languages.

Amarnath Gupta, Ph.D. (1994, Jadavpur University, India) is an Asst. Research Scientist at the San Diego Supercomputer Center, UCSD. His research interests are in multimedia and spatiotemporal information systems, heterogeneous information integration, and scientific databases. Before joining SDSC he was a scientist at Virage, Inc.

Chaitanya Baru, Ph.D. (1985, University of Florida) is a Senior Principal Scientist and technical leader of the Data Intensive Computing Environments group at the San Diego Supercomputer Center. He is a principal investigator on several XML-based projects related to digital libraries and information integration. Prior to joining SDSC, Dr. Baru led several IBM development teams, and prior to that he served on the faculty of the EECS Department at the University of Michigan, Ann Arbor.

The World Wide Web is a successful global infrastructure, and XML (eXtensible Markup Language) emerges as its data interchange model. In this paper, we present the XML-based information integration framework developed at the San Diego Supercomputer Center, as it extends to spatial information integration and enables global spatial data interoperability. Within this framework, sources of geographic data and services are “wrapped’ in XML conversion software, and thus expose their schemas and capabilities to a special mediating middleware, as XML documents. The mediator dispatches user queries to distributed heterogeneous data sources, and assembles query results. Selecting XML as the data interchange syntax provides for enough freedom in structuring geographic data, at the same time ensuring that different local standards can be made interoperable within a common extensible framework.

We describe the main components of the XML-based mediation architecture: (1) a network of “XML-wrapped” geographic sources and services, (2) the information mediation middleware that supports global resource discovery and query, and (3) an end-user interface capable of map rendering on a Web client. This represents work in progress, as reflected in how the various components are described with different levels of detail. We conclude with the outline of a research agenda for effective spatial data mediation on the global scale.

Global spatial data infrastructure is emerging through various national-level and international efforts, as a means to answer global environmental and health challenges, support international telecommunication, commerce, and human development, stimulate economic growth and productivity (Coleman and McLaughlin 1997, Holland 1999). Making local geographic datasets available internationally, and establishing a common interoperability framework over shared data interchange protocols are important components of this undertaking. As distributed geolibraries (“Distributed Geolibraries…” 1999) offer numerous advantages over stand-alone geographic databases, institutional and technical problems of geodata sharing and interoperability have become, over the last several years, the focus of international research and infrastructure efforts (Onsrud and Rushton 1995, Masser 1998). Technical aspects of this research agenda have been discussed at several international conferences and workshops focused on GIS interoperability, and several spatial data interoperability testbeds have been also developed (most notably, within the Digital Earth Initiative (“Digital Earth”, 2000) and OGC’s Web Mapping Testbed (OpenGIS Consortium, 2000)).

Advances in spatial data infrastructure and GIS interoperability are to a large extent driven by the development of the World Wide Web, client-server architectures and distributed processing. The WWW itself clearly shows an example of a successful global infrastructure enabling international telecommunication and commerce. Its success derives from universal acceptance of transfer protocols, easy procedures for content publishing and access, availability of free or inexpensive web client and web publishing software. The design of spatial data infrastructure architectures, both local and cross-national, relies heavily on this example. However, high degree of government involvement, and varying national strategies regarding production and dissemination of spatial data, introduce several important differences. In many countries production and sharing of spatial datasets remains a government prerogative (Masser, 1998). National-level spatial data infrastructure efforts in the US (FGDC, 1994), European countries (Masser, 1998), Australia and New Zealand (Holland et al., 1998), etc. are focused primarily on coordinating access to government-produced data. Different amounts of investment in spatial data infrastructures have led to wide differences in quality and currency of spatial datasets between the developed and developing countries. An increasing number of participants in global spatial data interchange, with varying needs and capabilities, explains the special concerns for the extensibility of the framework, the relative autonomy of particular datasets, and the variety of interfaces targeting particular user groups. In this situation, principles and strategies adopted in a single country with a successfully developing national SDI, may not be entirely applicable to the global scale, as global spatial infrastructure will likely be more than a simple sum of national infrastructures (Masser, 1998). We believe that a hybrid approach, integrating governmental and non-governmental efforts, allowing for easy inclusion of additional data providers in the system, and based on emerging Web data interchange standards, is necessary. What would be the technological basis for such an approach? This paper attempts to answer this question, by exploring an XML-based mediation infrastructure for global spatial data interoperability.

Extensibility, relative autonomy of infrastructure nodes, universal access to heterogeneous data sources from a variety of portals, reliance on Web standards are the principles behind the mediation framework described in this paper. Leaving institutional, political and economic considerations beyond the scope of this discussion, we focus on technical aspects of spatial data mediation. The paper is structured as follows. In the first part, we describe various strategies for data interoperability, spatial data interoperability in particular. The goal is to explore how these approaches can be scaled to the global interoperability context. The second part presents the details of our proposed XML-based spatial information mediation architecture, developed as part of the SDSC MIX (Mediation of Information using XML) project. We will focus on research and technical problems of spatial mediation on the global scale, and describe the main components of the interoperability architecture. In the third part, we present our work on XML-based map interfaces for the end-user that allow for building virtual maps from XML files distributed over the Web. The last part summarizes the benefits of spatial data mediation for the emerging global economy.

As offered in (Gore 1994, “The Global Information Infrastructure…” 1994) the five basic principles for creating global information infrastructure include encouraging private investment, promoting competition, providing open access, creating a flexible regulatory environment, and ensuring universal service. The main features of an underlying technology to answer these challenges are:

• Extensibility and modularity. The infrastructure design should allow for easy inclusion of additional relatively autonomous sources of geographic data and services, including ad hoc and field data, independent and private data providers (not just governments), pluggable conversion services, etc. In a sense, the evolution model of the infrastructure should follow the growth and expansion pattern of the WWW.
• Reliance on universally accepted market-driven Web protocols and standards. Voluntary compliance with Web standards and joining the infrastructure should be the basis for the broadest non-discriminatory and low-cost access to data sets and services of the global network (“The Global Information Infrastructure…” 1994, p. 10).
• Limited capital investments, i.e. maximal reuse of existing legacy system for spatial data.
• Built-in ability to resolve semantic, as well as format/platform conflicts between different datasets and services.
• Support for extensive metadata descriptions, including dataset schemas, data quality information and source capability descriptions, which enable search agents, catalog services and spatial query engines to navigate and compare different sources.
• Integration with other global infrastructure elements (including non-spatial relational databases, national statistical collections, etc.) through the use of common standards.

Different logical approaches to data interoperability described below, provide different answers to these challenges.

Main logical architectures for data interoperability

Recent reviews of GIS interoperability and integration efforts can be found in Abel (1998), DeVogele et al. (1998), Laurini (1998). GIS standardization efforts have resulted in adoption of various spatial data interchange standards, different in the U.S., France, Germany, Canada, and other countries (“Geographic Data Exchange Standards”, 2000). While government-level standardization efforts have not been successfully coordinated in most cases, the market place have embraced a different variety of commercial interchange formats.As regards spatial data interoperability architectures, the efforts fall into two major groups: data warehousing, and mediator-based systems. These approaches significantly differ in the degree of “tightness” of the infrastructure components, and extensibility. The data warehousing approach, as described by Voisard and Juergens (1999), for example, implies accumulation of spatial data in a few well-defined and tightly connected data stores, where information integration is “pre-computed”. While efficient for a relatively small number of core spatial datasets, this approach is not readily extensible to a larger number of datasets with semistructured and ad hoc data. Mediator-based systems, alternatively, are constructed from a large number of relatively autonomous sources of data and services, communicating with each other over a standard protocol and enabling “on-demand” information integration. The 3-level architecture of a mediator-based systems includes a "foundation" layer (heterogeneous databases with wrappers), a mediation layer (which supports exchange of queries and results between wrapped legacy data sources and applications), and an application/user interface layer (Wiederhold, 1992).

This architecture offers several advantages for global spatial data infrastructure. Obvious pluses are scalability and modularity. It is the responsibility of individual participants in the system to adapt the existing interchange standards and protocols, and failure to do so will not destroy basic global interoperability. Another advantage is that, responding to user queries, mediator only retrieves and combines individual query results from participating data stores, without retrieving and combining raw data. In addition, the use of a semistructured data model at the mediator enables modeling of sources with no structure or implicit structure. Examples of such semi-structured mediator-based systems include TSIMMIS (Papakonstantinou et al, 1995, 1996), DISCO (Tomasis et al, 1998), and Information Manifold (Levy et al., 1996). In the spatial data interoperability context, mediation-based approaches have been described by Shimada and Fukui (1999) and Bishr et al. (1999), as well as in our previous work (Gupta et al., 1999, 2000).

Emerging web standards for geographic data

Recommended by W3C (W3C, 1998a) in 1998, XML (eXtensible Markup Language) emerges as the new Lingua Franca for data interchange on the Web. It provides for semantic tagging of information, and supports semi-structured data. Several XML-based languages have been proposed for both 2D vector rendering, and for encoding geographic data and GIS projects. Among the 2D vector graphics languages, the Scalable Vector Graphics (SVG) and Vector Markup Language (VML) formats provide a mechanism for encoding graphic primitives for rendering in a Web browser. VML (W3C, 1998b) is implemented in Microsoft’s Internet Explorer 5 and is the graphics interchange format within the Microsoft Office 2000 suite. Geography Markup Language (GML), in its 0.5 version as of January 2000, provides a set of semantic tags for encoding coordinates of OpenGIS simple features (OpenGIS, 1997). Open GIS Consortium deserves an important credit for its efforts in developing this standard, as well as standard GIS interface specifications for Simple Features, Catalog Services, and Grid Coverages. An example of an XML markup language implemented in a commercial product, AXL (Arc XML) is developed as part of the soon to be released ArcIMS system and provides a mechanism for different geographic services to declare their schemas and capabilities in a uniform way (ESRI, 2000)

Global interoperability testbeds

Several global interoperability prototypes are being developed, including OGC’s Web Mapping Testbed (WMT) and the Digital Earth (DE).

The Web Mapping Testbed, Phase I, is the first of OGC's planned Interoperability Initiatives, which involve sponsors and participants. Federal agency and corporate sponsors provide funding and a set of objectives related to geoprocessing interoperability. WMT has so far demonstrated two types of applications: Web Mapping Clients, and Web Mapping Servers. The Clients create requests that satisfy the Web Mapping Protocols.The main interfaces are the GetMap protocol (which identifies one or more layers to be displayed), the GetCapability Protocol (which allows the client the ability to discover the abilities of a server), and finally the GetFeatureInfo protocol (which allows the client to uncover the attributes of a displayed feature). Web Mapping Servers interpret requests that conform to the WMT protocols and generate appropriate objects that are returned to the querying clients.

The Digital Earth initiative is meant to develop a virtual representation of the planet that allows an individual to explore and interact with vast amounts of natural and cultural information gathered about the Earth.While WMT specifications are part of the foundation of DE, other Digital Earth prototypes include applications such as TerraVision, GeoDE, Web Image Spreadsheet Tool, Global View from Space, Earth Today, GeoView, and Digital Earth Workbench. At the time of writing, both WMT and DE are exploring XML-based approaches to spatial data interchange.

The goal of the spatial mediator system we describe is to give a user the ability to issue a single query that would access multiple geodata sources to retrieve different pieces of the result and then assemble these pieces to provide a composite response to the user’s query in a seamless manner. In (Gupta et al., 1999) we described the architecture of a system that accomplishes logical integration for geographic information stored in a GIS system and georeferenced images stored in an image database management system. The capability description language for spatial data sources is described in (Gupta et al., 2000)

Functional architecture of the Spatial MIX system

Our information integration framework is called MIX, or Mediation of Information using XML (Fig. 1). It achieves information integration by using a two-part middleware between the information sources and the end-users’ application. The first part is called a mediator. It accepts a user request, breaks up the request into small fragments according to the capabilities of the sources and delegates the request-fragments to the appropriate sources. When the sources process the requests and return the results, the mediator integrates the results and sends the combined information back to the user.

The second part of the middleware is called a wrapper. The task of the wrapper is to translate a request from the mediator’s language to that of the information source and transform the results provided by the information source back to the mediator’s language. Acting as a proxy of an information source, the wrapper communicates with an information source in its native language/API, and communicates with the mediator in a commonly agreed language. In this way, “wrapping” each information source into the translation software makes the protocol diversity of particular sources manageable.

The specific feature of our MIX framework is that elements of the system exchange messages in XML format, the emerging Web data standard. User queries relayed to an application mediator are also expressed in an XML-based query language. The XMAS (XML Matching and Structuring) query language (Baru, 1999) developed as part of the MIX project, allows object fusion (e.g., combining a map layer reference from one source and a layer reference from another source into a new composite object) and pattern matching on the input XML data. Additionally, XMAS features powerful grouping and order constructs for generating new integrated XML “objects” from existing ones.

We have extended the MIX wrapper-mediator architecture to perform spatial data integration (Fig. 2). If the application mediator detects that a query evaluation requires accessing geographic sources, it delegates a particular query fragment to a spatial mediator. This happens in two situations: either variables declared in a particular XMAS query match data elements from a schema exported to the mediator by a geographic source, or the predicates used in a query are contained in a capabilities description of a geographic source. Using the query fragments supplied by the application mediator, and information about source capabilities, the spatial mediator determines an optimal query evaluation plan. This plan is formulated in a common language understood by GIS source wrappers, which translate the query fragments into the language of a particular source, initiating data retrieval or processing at each of them. According to the query execution plan, responses to query fragments are exchanged between sources, so that eventually a virtual map is assembled and sent to the user application. For example, in our experimental wrapping of ArcView sources described in (Gupta et al., 1999) XMAS requests were translated into Avenue code fragments by the ArcView wrapper; the wrapper also handled ArcView output generating a virtual XML document with query results and sending it back to the spatial mediator. Details of spatial mediation and XML-wrapping of geographic sources are presented below.

XML-wrapping of geographic sources

Heterogeneity of geographic sources may arise for a number of reasons, including differences in projection, precision, data quality, data structures and indexing schemes, topological organization (or lack of it), set of transformation and analysis services implemented in the source, etc. As mentioned above, GIS wrappers are designed to overcome the problem of source heterogeneity by maintaining an internal model of the GIS source mapped into a basic set of generic geometric primitives and operations shared by most current GIS. If a specific GIS contains some special function not included in the basic set, this function’s description can be exported from the wrapper to the mediator thus making the wrapper-mediator system functionally extensible. Expressing an internal model of a GIS in this set of primitives allows the wrapper to translate XMAS queries into syntactically well-formed queries in the language of the underlying GIS, and then export the results returned from the GIS as a well-formed XML document.

Since a wrapper maintains some “geographic knowledge” expressed through the geometric primitives and the model of the underlying GIS source, it acts as a “value-added” translator. Wrappers not only export the raw schema of the information source, but add their own organization and “knowledge” in it. This additional information aids the process of integration, because it allows the spatial mediator to know which data elements from two spatial sources should be associated with each other to produce a correct response to a query.

Spatial mediation in the middleware

The task of the spatial mediator is to parse the spatial part of the query and generate an evaluation plan. There are three parts in this process. The first part (source selection) entails fragmenting the query between information sources and the determination of the order of execution of each fragment. In a general case, the spatial mediator browses the schema information exported by GIS wrappers as an XML DTD, and dynamically evaluates the minimal combination of sources that satisfy the query attributes and predicates. Compared to source selection based on an explicit set of mapping rules (views), this general approach delivers a more efficient query evaluation plan when there are several competing sources (i.e. sources supplying similar and possibly redundant data). This is clearly a more appropriate strategy for global interoperability context, with a large number of data sources whose schemas are not available a priori.

In the second part, the knowledge about each source is used to rewrite the query. This rewriting employs a set of “capability” information exported by each source to the mediator (a capability specification scheme for a spatial data source is described in Gupta et al., 2000). The capability information includes, in particular, the set of data transformations supported by a GIS source. This information lets the mediator determine whether additional specialized data transformation services (such as conflation mediator, ontology mediator, accuracy mediator, etc.) need to be invoked in the process of query evaluation. In the third step, mediator sends out the rewritten query fragments to the sources, collects the result fragments from each source and passes them on to the application mediator.


Figure 3. A possible mediator architecture for global interoperability.

Spatial mediation for global spatial data infrastructure will require a more elaborate mediator architecture. We need to recognize, in particular, that mediator architectures for datasets with different “lifecycles” will also differ. For example, explicit source selection rules will deliver efficient evaluation plans for queries accessing a relatively small number of a priori known datasets, such as several core global, regional and national datasets often managed by a data warehouse. For large-scale, ad hoc and field data, a different mediator architecture and different wrapping mechanisms will be more appropriate. Thus, it may be beneficial to separate mediator mechanisms serving different kinds of datasets. An additional consideration for global mediator architecture is that most user queries are likely to be confined to a particular geographic area. This suggests that a hierarchy of regionally-specialized mediator services may prove efficient. A possible wrapper-mediator architecture for this situation is shown in Figure 3. Certainly, the ways different mediators interface with each other remain to be explored.

Query results, returned as map and tabular data from the mediator to the user interface on a Web client, can be rendered in different ways. A common approach implemented in most Internet Map Servers is to assemble the map on the server side (the spatial mediator, in our case), and send it to the client browser as an image file. Another approach is to stream geometric features, in binary format, from map server to a Java client handling vector rendering in a browser (one of the options implemented in ESRI’s ArcIMS 3.0 (2000). Yet another approach is to use a client side 2D vector rendering language, such as VML or SVG, to draw map features directly in a Web browser. In this part of the paper, we describe an implementation of this latter technique, based on VML rendering capabilities of Internet Explorer 5. The VML viewer, named AXIOMap (Application of XML for Interactive Online Mapping) is developed at ELZA Research (AXIOMap, 2000) in coordination with the MIX project. A snapshot of an AXIOMap is shown in Fig. 4.

AXIOMap is a relatively light-weight DHTML client, which allows users to construct multi-layer map presentations integrating geographic data from any number of Web servers. Geographic data are represented as XML files, with relatively straightforward DTDs for point, line and polygon layers. These XML files with geographic information can be produced within a wrapper-mediator system in response to user queries, or exported from stand-alone ArcView wrappers (at the time of writing). The viewer, though only 70K in size, allows users to:

• color maps by areas using datasets possibly retrieved from different Web servers;
• display/hide multiple point, line and polygon layers;
• display, analyze and save attribute information for any object or group of objects;
• construct simple and complex queries and map query results;
• navigate any map by zooming and panning, with user-defined zoom parameters;
• connect with Web sites associated with geographic objects;
• analyze proximity of various objects to one another;
• etc.

Once a map layer is rendered for the first time by the AXIOMap client, it is cached by the browser, and subsequent operations on this layer do not cause additional server round-trips. In our experience, meaningful user interaction with vector map data in the browser can be accomplished with much more modest computational resources compared to the standard model, where a Web map is delivered as an image file. We believe that this prototype of affordable XML-based interactive map publishing may be attractive for the heterogeneous environment of global spatial data infrastructure, where different economic and technical models, and different levels of network communication, should be co-operational.

The goal of global spatial data infrastructure is to promote sharing and communication of geospatial knowledge across national boundaries. As in the case of the WWW, its growth and proliferation will likely be a natural market-driven growth, based on shared protocols for information exchange, and relatively autonomous nodes. Incorporating the efforts of governments and non-governmental spatial data providers and users into a flexible and extensible system can be done on the principles of spatial data mediation described in this paper. We highlighted the potential of mediation approaches for global spatial data infrastructure, in particular related to extensibility, modularity, flexibility and reliance on common interchange standards. However, as different kinds of datasets and services (governmental as well as non-governmental information and services, pre-defined core as well as ad hoc datasets, national-level as well as global and regional data, etc.) will need to become interoperable under a common framework, both data warehousing and mediation concepts will likely contribute to the design of GSDI.

While establishing an infrastructure testbed implementing the described architecture for Web-based spatial information mediation using XML is our first priority at this stage, several related research challenges need to be also addressed. These proposed research areas include: (1) development of a comprehensive cost model for source selection in the spatial mediator, that takes into account data quality and data organization, volume of data, transformation needs and capabilities etc., (2) development of specialized data transformation/conversion plug-ins to a spatial mediator; (3) inferring the desired accuracy of data sources based on target data accuracy (i.e. accuracy mediation), (4) resolving semantic and ontological conflicts which are likely to happen in the course of cross-national spatial mediation.

Abel, D. J. (1998) Towards integrated geographical information processing. Int. J. Geographical Information Science 12:353-371.

AXIOMap: Application of XML for Interactive Online Mapping (2000) http://www.elzaresearch.com/gis/vml/vmlmapping.html [Viewed 02-24-2000]

Baru, C., Gupta, A., Ludäscher, B., Marciano, R., Papakonstantinou, Y., Velikhov, P., Yannakopulos, A. (1999) XML-Based Information Mediation with MIX. Proceedings of the SIGMOD’99. http://www.npaci.edu/DICE/Pubs/sigmod-demo99.pdf

Bishr, Y.A., Pundt, H., Rüther, C. (1999) Proceeding on the Road of Semantic Interoperability - Design of a Semantic Mapper Based on a Case Study from Transportation. In Proc. Interoperating Geographic Information Systems, Second International Conference, INTEROP '99, Andrej Vckovski, Kurt E. Brassel, Hans-Jörg Schek (Eds.), Zurich, Switzerland, March 10-12, 1999, pp. 203-215.

Coleman, David, and McLaughlin, John. (1997) Defining Global Geospatial Data Infrastructure (GGDI): Components, Stakeholders and Interfaces. Theme Paper #1, GSDI 97 Conference (Chapel Hill, North Carolina, USA): http://www.gsdi.org/docs/ggdiwp1.html [Viewed 02-24-2000]

DeVogele, T., Parent, C., and Spaccapietra, S. (1998) On spatial database integration. Int. J. Geographical Information Science 12:335-352.

Digital Earth (2000). http://www.digitalearth.gov/ [Viewed 02-24-2000]

Distributed Geolibraries. Summary of a Workshop (1999). Mapping Sciences Committee, National Research Council. National Academy Press: Washington, D.C.

ESRI (2000) ArcIMS. http://www.esri.com/software/arcims/index.html [Viewed 02-24-2000]

Federal Geographic Data Committee (1994). The 1994 Plan for the National Spatial Data Infrastructure: Building the Foundation of an Information Based Society, Washington DC.

Geographic Data Exchange Standards (2000): http://www2.echo.lu/oii/en/gis.html[Viewed 02-24-2000]

Gore, A. (1994). Speech at the World Telecommunication Development Conference, Buenos Aires, March 21, 1994: http://www.iitf.nist.gov/documents/speeches/032194_gore_giispeech.html[Viewed 02-24-2000]

Gupta, A., Marciano, R., Zaslavsky, I., Baru, C. (1999). “Integrating GIS and Imagery through XML-Based Information Mediation”. In P. Agouris and A. Stefanidis (Eds.) Integrated Spatial Databases: Digital Images and GIS, Lecture Notes in Computer Science, Vol. 1737.

Gupta, A., Zaslavsky, I., and Marciano, R. (2000). Generating Query Evaluation Plans within a Spatial Mediation Framework (submitted for Spatial Data Handling 2000 Conference)

Holland, P., Nairn, A., Baker, G., et. al. (1998) The Development of Spatial Data Infrastructure in the Asia-Pacific Region with Special Reference to Australia. The 10th International Geomatics Conference (Ottawa, Canada, 8-11 June 1998)

Holland, P. (1999). The Strategic Imperative of a Global Spatial Data Infrastructure. Cambridge Conference for National Mapping Organizations, Cambridge, UK, 19-23 July 1999.

Laurini, R. (1998) Spatial multi-database topological continuity and indexing: a step towards seamless GIS data interoperability. Int. J. Geographical Information Science 12, 1998: 373-402.

Levy, A., Rajaraman, A., and Ordille, J. (1996) Querying Heterogeneous Information Sources Using Sources Descriptions. Proceedings of VLDB, 1996: 251-262.

Masser, I. (1998). Governments and Geographic Information. Taylor&Francis: London.

Onsrud, H., and Rushton G., Eds. (1995) Sharing Geographic Information. Rutgers University Press: New Brunswick, NJ.

Open GIS Consortium (1997) OpenGIS Simple Feature Specification For OLE/COM (Revision 1), Wayland, Massachusetts, 1997.

Open GIS Consortium Web Mapping Testbed Public Page (2000). http://www.opengis.org/wmt/[Viewed 02-24-2000]

Papakonstantinou, Y., Gupta, A., Garcia-Molina, H., Ullman, J. D. (1995) A Query Translation Scheme for Rapid Implementation of Wrappers. DOOD 1995: 161-186. http://www.cse.ucsd.edu/~yannis/papers/querytran.ps

Papakonstantinou, Y., Abiteboul, S., and Garcia-Molina, H. (1996) Object Fusion in Mediator Systems. In Intl. Conf. on Very Large Data Bases (VLDB), 1996.

Rhind, D. (1997). Implementing a Global Geospatial Data Infrastructure (GGDI). Theme Paper #2, GSDI 97 Conference (Chapel Hill, North Carolina, USA): http://www.gsdi.org/docs/ggdiwp2b.html[Viewed 02-24-2000]

Shimada, S., and Fukui, H. (1999) Geospatial Mediator Functions and Container-based Fast Transfer Interface in SI3CO Test-Bed. In Proc. Interoperating Geographic Information Systems, Second International Conference, INTEROP '99, Andrej Vckovski, Kurt E. Brassel, Hans-Jörg Schek (Eds.), Zurich, Switzerland, March 10-12, 1999, pp. 265-276.

Telesis Corp. (2000). The San Diego Quality of Life Project. http://www.qolsandiego.net [Viewed 02-24-2000]

The Global Information Infrastructure: Agenda for Cooperation. (1994). Information Infrastructure Task Force, NIST: http://www.iitf.nist.gov/documents/docs/gii/giiagend.html[Viewed 02-24-2000]

Tomasic, A., Raschid, L., Valduriez, P. (1998). Scaling Access to Heterogeneous Data Sources with DISCO. IEEE Transactions on Knowledge and Data Engineering, Vol. 10, No. 5, 1998: 808-823.

Voisard, A., and Juergens, M. (1999) Geographic Information Extraction: Querying or Quarrying? In Interoperating Geographic Information Systems, M. Goodchild, M. Egenhofer, R. Fegeas and C. Kottman (Eds.), Kluwer Academic Publishers, New York.

W3C (1998a). Extensible Markup Language (XML) 1.0. W3C Recommendation, 10-February-1998. http://www.w3.org/TR/1998/REC-xml-19980210 [Viewed 02-24-2000]

W3C (1998b). Vector Markup Language (VML). World Wide Web Consortium Note 13-May-1998. http://www.w3.org/TR/NOTE-VML[Viewed 02-24-2000]

Wiederhold. G. (1992) Mediators in the Architecture of Future Information Systems. IEEE Computer, 25(3):38-49.

Financial support for the SDSC XML-based information mediation project is provided through a variety of funding sources, including the National Science Foundation (NSF) and the National Archives and Records Administration (NARA).

The Quality of Life in San Diego project is conducted by Telesis Corp. We wish to thank David Cleveland, President of Telesis, for his assistance in this project and for providing relevant data.