Appendix 3: Position Papers

Arne-Jurgen Berre, David Skogan

Position statement abstract

The position being argued for is that the work on interoperating geographic information systems should be based on the general IT work in the areas of systems integration, schema integration, semantic interoperability and multidatabase systems. In particular for the important area of semantic interoperability we advocate for a possibility to support a language specification approach, based on concepts from schema integration mapping languages such as EXPRESS-X, in addition to a dictionary approach for semantic translators. Further research is required in this area in order to apply and specialize these concepts to the GIS domain.

The problem of structural and semantic interoperability between data from different information communities is similar to the problem of schema integration in multidatabases. There is a set of well known structural conflicts that might arise [KS91], such as synonyms, homonyms, data representation conflicts, data unit conflicts, data precision conflicts, data quality conflicts, default value conflicts and integrity constraint conflicts. In addition to try to map equivalent objects on a one-to-one syntactical basis, one can define a semantic measure for equivalence of objects. A taxonomy targeted on defining semantic proximity has been developed by [SK92] and mapped to object-oriented models in [EK95]. Semantic proximity is defined as a function between 0 and 1, based on context, abstraction, domain and state of the objects. The taxonomy distinguishes between semantic incompatibility, semantic resemblance, semantic relevance, semantic relationship and semantic equivalence.

Examples of approaches that addresses structural and semantic interoperability can be found both in database mapping languages, and in conceptual schema modeling languages such as EXPRESS. EXPRESS-X [EX97] is currently in the ISO standard development, based on a unificiation of two previous mapping languages, EXPRESS-M [EM95] and EXPRESS-V. A description of ODL-M for the object-oriented databases standard from ODMG is given in [KI96]. The problem of semantic interoperability is being addressed through support for a semantic proximity function.

The use of a mapping language is in particular feasible when a conceptual schema language has been used in the specification of a generic feature model and corresponding application schemas. Both the ISO/TC21 and CEN/TC287 standards advocate such an approach, while the OGC/OpenGIS standard is only working through a dynamic API.

In this position statement we argue for further addressing research in the area of semantic interoperability, in particular through extending work on mapping languages for schema integration, from the area of multidatabase systems.

[EX97] "EXPRESS-X", Mapping language based on EXPRESS-M and EXPRESS-V in progress in the ISO STEP/EXPRESS community

[EM95] "EXPRESS-M Reference Manual", CIMIO Ltd, August 1995

[KI96] "ODL-M - A Mapping Language for Schema Integration in Object-Oriented Multidatabase Systems" MSC-thesis, Steinar Kindingstad, University of Oslo/SINTEF, August 1996

[KO95] "Semantic Proximity in Multidatabases", MSC-thesis, Espen Koren, University of Oslo/SINTEF, July 1995

[SK93] "Multimodels for GIS", MSC-thesis, David Skogan, University of Trondheim, December 1993

[BE93] "An Object-Oriented Framework for Systems Integration and Interoperability", A.J. Berre, Phd-thesis, University of Trondheim, August 1993

[SK92] "So far (Schematically) yet so near (Semantically)", Amit Sheth, V. Kashyap, IFIP DS-5, Semantics of Interoperable Database Systems, Australia, November 1992

[KS91] "Classifying Schematic and Data Heterogeneity in Multidatabase Systems", Won Kim, Jungyen Seo, IEEE Computer, (22)3, December 1991

Arne-Jurgen Berre, David Skogan
SINTEF Telecom and Informatics
P.O. Box 124
Blindern, N-0314 Oslo, Norway
{Arne.J.Berre | David.Skogan}@informatics.sintef.no
http://www.informatics.sintef.no

Ling Bian

The most recent accomplishment of OpenGIS will have a significant impact on a broad range of disciplines. Environmental modeling requires geographic data and geoprocessing functions; thus it is an inseparable part of the interoperability mission. Integrating environmental models and GIS involves issues that are well beyond the concerns of interface standards, and should be first addressed at a conceptual level.

Many integration frameworks have been proposed or implemented in the past few years (Chou and Ding, 1992; Nyerges, 1993; Abel et al., 1994). These efforts have alleviated the difficulties encountered almost daily in using GIS for environmental modeling. This endeavor also sees its limitations. First, these proposals have always involved multiple options, from the lower level, simple data transfer to higher level, complex coupling. There has not been a more focused solution because higher levels of integration are always associated with higher development burden. The newly developed OpenGIS interfaces provide transparent access to heterogeneous GIS data sets. This development will partly free users from the integration labor and make it affordable to choose more desired, higher level of integration. This effect will help narrow the "multiple choice" down to more focused solutions.

A second problem still remains that, even for a higher level integration, the aforementioned integration strategies are limited to "per-model" solutions. An integration system, often including a user interface and a shared database, developed for one model is not portable to another. This problem stems from the fact that many environmental models are closed, monolithic systems. Most of them have no capability to communicate readily with either GIS or other environmental models. The diversity of these models makes it impractical to develop specifications for every GIS–model and model–model interface. This has been a long-standing problem in an era when multiple data sets and multiple models are required to solve environmental problems. Evidently environmental models must "open up" in order to achieve a full integration with GIS and between models themselves.

Moreover, one of the high-level integration options calls for implementing environmental models in GIS or vise versa. The former is more often seen for the benefit that the models can use GIS data models and languages directly. Rewriting mathematical equations in AML or map algebra type of languages is a typical attempt. This option can achieve reasonable results for a limited type of models, such as simple empirical models that use black box approaches or simple physical models whose parameters are temporally invariant and spatially homogeneous. For a majority of physically-based models used in hydrology, atmospheric science, and increasingly in ecology, executing mathematical equations directly by GIS languages is an impractical choice because GIS languages cannot perform at the level of computer programming languages traditionally used for such tasks. Similarly, conducting spatial functions in environmental models is equally inefficient. The ideal integration needs to be worked in a middle ground.

It is necessary to understand GIS and environmental models at a conceptual level before technical solutions are sought. GIS and environmental models differ in their representations of the world. GIS focuses on descriptions of space and relationship between spatial features. Environmental models aim at descriptions of dynamic processes of phenomena. This space–process difference determines the distinction in abstract models and languages used by GIS and the models (Maidment 1996). While environmental models use mathematical languages to model the dynamic aspect of the world, GIS languages are designed primarily for spatial operations. Reflected in integration practice, the role of GIS in physically-based process modeling has not been much beyond "front ends" (pre-processing spatial data to prepare model input) and "back ends" (visualize model output spatially). This difference should be well respected and kept. Instead of forcing one into the other, the two representation models should be linked in a common framework.

Object orientation may provide such a framework that links the space- and process-oriented abstract models (Raper and Livingstone, 1995). The design and implementation of object systems outlined in Cook and Daniels (1994) are adopted by OpenGIS specifications. They can be extended to set the framework of linking GIS and environmental models. The fact that environmental processes occur in geographic space helps establish an essential model of the link. At the specification level, mathematical operations a re-executed on spatial fields or features. The spatial fields or features may be defined as objects and they possess properties (e.g., geometry-topology, location-time, or non-spatial attributes). These objects can exert or receive operations, spatial or process-based. Events execute the operations and trigger the state change for the objects. While this framework defines the nature of the link between GIS and the models, more questions arise at the implementation level.

If both the spatial and process representations can be appropriately implemented as spatial objects, spatial operations, and process operations, they would likely or should be implemented in different languages most efficient for the implementation. The linkage between them should be able to interface the difference. Kemp (1993, 1997a, 1997b) elaborated an interfacing strategy that went a step further. It provides intelligent match between the spatial and process representations. The interfacing syntax can be implemented in computer programming languages so that the process models can directly call for the appropriate data models and spatial operations. The development of OpenGIS specifications helps realize this strategy that the process models can access directly the standardized GIS data and operation components. However, this is still a one-way solution. On the other side of the interface, environmental models need to be opened.

Opening environmental models should aim at communicating not only between GIS and models but also between models themselves. The development of OpenGIS specifications sets a precedent for how this could be achieved. Developing componentware in a distributed computing environment seems to be an ideal approach and consistent with the development in computing industry. Long before the success of OpenGIS, there had been many calls for developing standard module libraries and data exchange formats for integrating GIS and environmental models (Moore et al., 1993; Kemp, 1993, 1997b; Leavesley et al., 1996). These calls were from both GIS and modeling communities. The standard module libraries can be developed for either spatial operations or process operations.

Developing componentware is feasible and appropriate for opening environmental models. The dynamic processes of the physical world contain a series of specific processes through time. The mathematical models that represent these physical processes normally consist of a series of algorithms corresponding to the specific processes. Some of the specific processes are common to different models. For example, evaporation process may be a common component shared by atmosphere, surface hydrology, and soil moisture models. Leavesley et al. (1996) developed a module library with standard module structure so that users can select and link modules for a particular modeling purpose. Although their work was not language-, platform-, or GIS independent, the strategy can be used to implement componentware for environmental models.

The components should be compatible to GIS and between themselves, and they should be reusable, extendible, and retrievable from a distributed environment. Object orientation may be the most appropriate implementation approach (as opposed to developing the conceptual framework mentioned previously). Environmental models can be implemented in terms of objects and operations (although these may not be an exact one-to-one mapping of variables and algorithms used in the process models). This implementation allows development of operation libraries. In the libraries, the operation components are reusable and extendible whenever necessary to meet specific modeling needs.

Object-oriented design is the most appropriate approach known for attaining these goals (M eyer, 1987). Furthermore, the environmental components should be retrievable in a distributed environment. Object oriented design allows cataloging the component types and attaching "metadata" to the types so that users can identify and locate appropriate components.

Opening environmental models is not an easy undertaking. It requires research at several different levels, from establishing conceptual framework to technical implementations. The institutional challenge may be greater than the technical ones. OpenGIS specifications are made possible by the efforts of private sectors. Environmental models, especially those in hydrology and atmosphere sciences, were developed or endorsed by federal government agencies. A full collaboration from modeling community is the premise for any further progress.

Ling Bian
Department of Geography
State University of New York
Buffalo, NY 14261-0023

Constraint-Based Interoperability of Spatiotemporal Databases

Jan Chomicki, Peter Z. Revesz

Very large temporal, spatial and spatiotemporal databases are a common occurrence nowadays. Although they are usually created with a specific application in mind, they often contain data of potentially broader interest, e.g., historical records or geographical data. By database interoperability we mean the problem of making the data from one database usable to the users of another. Data sharing between different applications and different sites is often the preferable mode of interoperation. But sharing of data (and application programs developed around it), facilitated by the advances in network technology, is hampered by the incompatibility of different data models and formats used at different sites. Semantically identical data may be structured in different ways. Also, the expressive power of some data models is limited.

Temporal and spatial databases share a common characteristic: they contain interpreted data, associated with uninterpreted data in a systematic way. For example, a temporal database may contain the historical record of all the property deeds in a city. A spatial database may contain the information about property boundaries. Moreover, as this example shows, spatial and temporal data are often mixed in a single application.

In this research, we propose that constraint databases (Kanellakis et al. 1995) be used as a common language layer that makes the interoperability of different temporal, spatial and spatiotemporal databases possible. Constraint databases generalize the classical relational model of data by introducing generalized tuples: quantifier-free formulas in an appropriate constraint theory. For example, the formula 1950 <= t <= 1970 describes the interval between 1950 and 1970, and the formula ((0 <= x <= 2) AND (0 <= y <= 2)) describes the square area with corners (0,0), (0,2), (2,2), and (2,0). The constraint database technology makes it possible to finitely represent infinite sets of points, which are common in temporal and spatial database applications. We list below some further advantages of using the constraint database technology:

1. Wide spectrum of data models. By varying the constraint theory, one can accommodate a variety of different data models. By syntactically restricting constraints and generalized tuples, one can precisely capture the expressiveness of different models.
2. Broad range of available query languages. Relational algebra and calculus, Datalog and its extensions are all applicable to constraint databases. Those languages have well-studied formal semantics and computational properties, and are thus natural vehicles for expressing translations between different data models. Also, constraint query languages may be able to express queries inexpressible in the query languages of the interoperated data models, augmenting in this way the expressive power of the latter. (This is more a practical than a theoretical contribution. We simply mean that if, for instance, we have a TQuel database, then translation to a constraint database with dense order constraints allows querying by Datalog, a query language which is more expressive than TQuel. Similar comments apply to several other spatial and temporal data models in use.)
3. Decomposability. The problem of translating between two arbitrary data models, which is hard, is decomposed into a pair of simpler problems: translating one data model to a class C of constraint databases, and then translating C to the other data model. Also, by using a common constraint basis, we need to write only 2n instead of n(n-1)/2 translations for n different data models.
4. Combination and interaction of spatial and temporal data within a single framework. This is an issue of considerable recent interest, for example in the ESPRIT Chorochronos project.

In this paper we address the issue of application-independent interoperability of spatiotemporal databases. We show that the translations between different data models can be defined independently of any specific application that uses those models. We distinguish between data and query interoperability. For the former, it is the data that is translated to a different data model, while the latter concerns the translation of queries. The constraint database paradigm is helpful in both tasks. For data interoperability, constraint databases serve as a mediating layer and translations between different data models are expressed using constraint queries. For query interoperability, it is the constraint query languages themselves that serve as the intermediate layer. In an actual implementation, the presence of a mediating constraint layer may be completely hidden from the user.

We show below two scenarios in which data interoperability may be useful in practice.

SCENARIO 1: The user of a data model Mod2 wants to query a database D1 developed under a data model Mod1. He translates D1 to a Mod2-database D2 (using constraint databases as an intermediate layer) that he can subsequently query using the query language of Mod2. (As a practical matter, if a user is interested in a query Q2 in Mod2, then only the part of the database that is relevant to the query needs to be translated.)

SCENARIO 2: The user of a data model Mod1 wants to augment the power of the query language of Mod1. For example, this language may be unable to express recursive queries. However, such queries can be formulated in an appropriate constraint query language. Thus whenever the user wants to run such a query on a database D1, he first translates D1 to a constraint database, runs the query in the constraint query language on it (using a constraint query engine), and translates the result back to Mod1. (note that interoperating query results is an often neglected aspect of database interoperability.)

We report here on the preliminary results of this NSF-funded research project. We have studied the interoperability between the two-dimensional spaghetti spatial data model (which we believe to be representative of a large class of spatial data models) and linear arithmetic constraint databases. The move to spatiotemporal databases has turned out to be tricky: we are still in the process of defining an appropriate temporal extension of the spaghetti data model. While constraint databases are clearly an appropriate formalism for specifying the translations between different data models, current constraint database engines are too slow to compute the translations. This suggests the need for developing efficient algorithms for data translations, whose correctness can then be checked against the constraint-based specifications.

Jan Chomicki
Dept. of Computer Science
Monmouth University

Peter Z. Revesz
Dept. of Computer Science
University of Nebraska - Lincoln

Organizational and Technological Interoperability are Intertwined in Geographic Information Infrastructures: Evidence from Sociological Theory and Empirical Study.

John D. Evans

Introduction

Sharing geographic information is often seen as either a technological problem or an organizational one, each with quite distinct research thrusts: whereas some may seek to build, say, data-translation software or navigational tools, others may tackle such matters as institutional inertia or intellectual property. These focused research efforts are valuable in their own right; but in an unsettled, rapidly changing technological and organizational context, sharing geographic information is rarely a purely technical problem or a purely organizational one (Evans and Ferreira, 1995). For instance, technical innovations such as interoperable interfaces may only affect information sharing in organizations that are encouraging their members to pursue cooperative approaches to their work. Conversely, the "inertia" that slows use of outside geographic data may in fact be a quite sensible response to difficult data-coordination problems tied to the constraints of current technology and to the complexity of the data itself.

To understand and guide the growth of interoperable information sharing, it's helpful to consider organizational and technical interoperability as interdependent, moving targets. As the next two paragraphs summarize, this perspective finds support in leading sociological theory, and has proven valuable to the study of inter-agency geographic information infrastructures.

In describing the relation between technologies and organizations, Markus and Robey (1988) emphasize an emergent perspective, focused on the interactions between organizations and technology, in contrast with both a technological determinism (in which technologies are presumed to have known, inexorable effects on organizations) or a social strategic choice (in which technologies are seen as inexorably shaped by people's intentions and actions). Barley (1986) examines these interactions as they unfold over time, and invokes structuration theory (Giddens, 1984) to trace the ongoing, recursive influence between an organization's structure (i.e., its rules and resources) and the behavior of its members, as change is triggered by new technologies. The structuration perspective is sensitive not only to the effects of group norms, rules, and broader trends, but also to the influence of people acting unpredictably within and on these forces.

Applying the structuration perspective to information technology, DeSanctis and Poole (1994) emphasize the "intertwined" nature of technological and behavioral patterns, and Orlikowski (1992) proposes a useful view of technology as a malleable structural property of organizations: that is, a set of rules and resources that enable some actions, while constraining others, and that are in turn shaped by those actions over time. Within a structuration perspective, organizational intentions alone cannot give rise to a given technology, nor can a technology have a fully predictable effect on organizations. Rather, in every phase of a technology's existence—its conception, design, deployment, use, evaluation, and modification—the human actors involved mediate both causal effects in unpredictable ways.

Furthermore, in this perspective, both technological and organizational change are considered normal and ongoing: particularly in the case of large information networks, this implies "organic, yet systematic" change over time (Spackman, 1990). Designers of information systems often make the more-or-less tacit assumption that organizations are static—that structural changes are abnormal and reach an equilibrium. Conversely, organizational thinking tends to accept technologies as artifacts with stable features and a fixed role. However, particularly as seen through the structuration perspective, social structures undergo constant change, and information technology itself is an element of that social structure, enabling some actions, constraining others, and itself shaped by those very actions over time (Orlikowski, 1992).Within this perspective, the technical design of an information sharing infrastructure is ineluctably tied to its ongoing implementation and use within an organizational context.

The cyclical, dynamic perspective provided by structuration theory is conceptually pleasing; and in addition, it provided a fruitful model for an empirical study of interoperable geographic information sharing infrastructures (Evans, 1997).

A recent case study of three inter-agency geographic information infrastructures (the Great Lakes Information Network, the Gulf of Maine Environmental Data and Information Management System, and the Pacific Northwest StreamNet and its predecessors) shows the value of the perspective described above. In seeking to describe and understand these cases, traditional one-way factor models of technology's impact on organizations, or organizational impacts on technology, inevitably led to "chicken-and-egg" dilemmas, accounted poorly for change over time, and blurred the roles of individuals, groups, and broader societal trends. Instead, the structuration perspective elucidated a cycle of influence similar to that described by Orlikowski (1992), with mutual influences between organizational, technological, and policy/planning structures, and the actions that people perform on and within those structures.

These cyclical, dynamic patterns of influence provided new insights into the growth and change mechanisms evidenced in the three cases. First, rather than postulate direct influences between constructs like technology, organizations, or policy, this model sees all of the influences as mediated by human actors enabled and constrained by these constructs. For instance, as technological standards influenced what people could do, some of these people chose to create new partnerships; in so doing, they sometimes found themselves empowered or slowed by broader laws. Second, this model proved fruitful in understanding the evolution (or stagnation) of the three inter-agency efforts towards interoperable information sharing: in particular, it reconciled the free-will choices of particular "champions" with the influence of their evolving social and technological context. Third, this model suggested any number of levers for perturbing existing behavior and guiding it towards a particular target, while making it clear that interoperable information sharing infrastructures are less a set of fixed, interlocking technical and organizational components than a chosen direction, or even a style, of evolution through an uncertain future. Although a broad set of levers can be pulled to affect sharing, collaboration, or consensus, their influence on outcomes is uncertain and only temporary. Thus, any solutions considered should be conceived as packages of mutually-influencing technological and organizational features, and as pathways of not-fully-predictable growth and change over time.

In summary, a "holistic" view of technological and organizational interoperability in concert, rather than in isolation, is persuasive for the study of inter-agency geographic information systems; it can be rigorous thanks to the structuration model, and has proven useful through its insights into recent empirical findings.

John D. Evans, Ph.D.
MIT Dept. of Urban Studies and Planning
E-mail: jdevans@mit.edu
Phone/Fax: 617-734-1512
Mail: 1404 Commonwealth Ave. #4
Brighton, MA 02135-3722

Semantic Interoperability Issues in the Geosciences

Mark Gahegan

Over the recent past, considerable progress has been made towards interoperability between mainstream GIS packages. However, the interoperability issues addressed so far have been largely restricted to a class of GIS using two dimensions and where the geographic models considered (the conceptual basis of the systems) are quite similar. Within the whole range of the geosciences there are many other types of systems, some devoted to remote sensing, others to the modelling of true three dimensional geological structure. Each have their own idiosyncrasies in terms of conceptual model, data structures and functionality. It is my aim with this position paper to bring into the mainstream some of the issues concerning this wider set of packages, particularly with regard to the needs of the remote sensing and exploration / mining communities of users, whose needs have perhaps been addressed less rigorously than they would like.

Of key importance to the strategic development of open systems are the embracing of the three (and four) dimensional concepts that are available in many geologically oriented systems, including those used for exploration, mining and groundwater modelling. Examples of commercial systems are Surpac 2000, Vulcan and Micromine, although there are many more. The users of these systems represent a very large community whose data translation and interoperability needs are often overlooked. In many respects, interoperability within these systems is just as pressing an issue as with traditional GIS. The systems are extremely diverse with respect to function and role. For example, many exploration companies will ordinarily use two, three, or more completely distinct systems at the same time, to fulfil all their planning and modelling needs (e.g. minesite layout, drill-hole logging, mineral potential mapping). The costs involved in moving data between these systems are enormous, with some companies having to settle for essentially re-entering the data each time it is required in a different system. Consequently, there is a huge potential for improvement, which interoperability could satisfy, and a commitment within many organisations to solve these issues once and for all. Whilst the OGIS reference model appears up to the task, extensions to the OGIS Geodata Model are required to fully support three dimensional objects, or objects with coordinates in the vertical plane such as geological profiles, faults, dykes and drill-holes. Existing packages often operate within a quite restricted semantic framework; the same logical entities being precisely and similarly defined (as far as the user is concerned) across many systems. Some standards exist for nomenclature and meaning, such as defined by the Australian Minerals Industries Research Association (AMIRA) sponsored GEODATA project. Any planned interoperability must embrace the significant progress that has already been made in defining the logical entities in use across existing systems since these are becoming the accepted norm.

Also of critical importance is the need for further progress to better integrate remote sensing activities with GIS. The recent OGF initiative on image formats and meta-data lays a foundation for better integration, but falls short in respect of the mechanisms by which the geographic objects used in GIS are formed from image data. This in turn raises two related issues: the choice of object formation strategies (image segmentation) by which the objects used by GIS are made, and the semantic definition of geographic objects so that their meaning is communicable in some manner. Image segmentation is problematic to describe formally; the algorithms, data and knowledge used will, in effect, determine the objects formed. So, object semantics are partly determined by the abstraction (extraction) processes applied. (Smith et al., 1992; Gahegan & Flack, 1996; 1997).

Alternatively, object semantics may be defined in linguistic or high level terms as shown by Kuhn (1994). It is necessary to ensure that the meaning of data can be communicated along with the data itself, since without a clear statement of the meaning the opportunities for data misuse increase. Both of the above approaches may be required, depending on the origins of the data. If a statement of meaning can be formalised then it is possible to include some software safeguards as part of interoperation functionality. Further details of my research in this area can be found in the accompanying conference abstract and in earlier work (e.g. Gahegan, 1996).

In summary, my position is one of concern for object semantics in the wider realm of geo-information processing. I am keen to be involved in any initiatives to further develop the semantic basis for interoperability in order to improve the quality of data exchange, and the sharing of functionality between systems. I see this area as being one of the current shortcomings with the OGIS Geodata Model which could be successfully addressed as part of an ongoing collaboration. My relevant project experience (shown below) indicates how I am currently active in this area and involved with many industry representatives (both users and developers) from the larger realm of the geosciences.

Mark Gahegan
Geographic Information Science
Curtin University of Technology
PO BOX U 1987
Perth 6001, WESTERN AUSTRALIA
phone: +618 9266 3309
fax: +618 9266 2819
E-mail: mark@cs.curtin.edu.au
Web: http://www.cs.curtin.edu.au/~mark/

Spatially Enabling the Web with OGDI

Kenn Gardels

The use of geographic information is extending far beyond its traditional boundaries of mapping programs, to embrace a broader community of analysts, decision-makers, and interested citizens. With this has come increased interest in locating and using geodata available throughout the network, whether this is the entirety of a global spatial data infrastructure or in the confines of a corporate intranet. These data are of course maintained in a diversity of systems/organizations, presenting vast challenges for data access and interpretation. Geographic information system interoperability across platforms and data formats has proceeded in a series of initiatives addressing conversion tools, standardized interchange formats, application interfaces, and clearinghouses.

Two major forces are at work in forging new demands on technologies for access to information: recognition by system builders and users that applications require complete, consistent, timely information; and expectations in the age of the World Wide Web that information should be easily found and used. While earlier efforts in the field of geographic information systems targeted mechanisms for inter-system data exchange, current efforts in the GIS industry, as exemplied by work by Open GIS Consortium participants, have redirected interest into specifications for geodata and geoprocessing interoperability within distributed computing environments. At the same time, software developers are attempting to take advantage of the proliferation of the Web as a basis for tools for making information available to Web users from specialized servers.

Although the proposed implementations fitting the Open GIS Abstract Specification are based on distributed computing platforms—CORBA, OLE/DCOM, and ODBC—they do not directly solve the problem of general Web-based access to a network of heterogeneous geodata. The Open Geographic Datastore Interface (OGDI) was written as a means of rendering geodata heterogeneity invisible to Web clients. It does so by defining a set of standard interfaces for connecting to datastores, describing a dataset’s organization and structure, extracting geodata objects, and establishing common regions of interest, projections, and coordinate systems. A key component of OGDI is gltp, a stateful network protocol for linking geodata servers and clients. It provides a mechanism for locating a remote data source, specifying its format, and defining the pathname or identification for a dataset. Specifically, it links client requests or queries to a software driver on the appropriate geodata server, and returns results from the server to the client application. The driver interprets OGDI queries and initiates the corresponding data retrieval operation on the native data store. The results are placed in a generic structure that models point, line, polygon, annotation, and raster information, which the OGDI-aware client can then use directly.

Under development now are assembler components between OGDI clients and OGIS servers. Although the OGDI and OGIS data models are not presently congruent, they are very similar in their ways of describing geodata—OGDI type families and OGIS feature types describe the same basic entity types. The assemblers act as proxies to make collections with exposed OpenGIS interfaces available via the gltp protocol for map visualization and modeling by relatively thin clients. Such clients can be standalone applications, downloadable applets, or plug-ins. At a higher level of abstraction, a map view may be rendered via standard html/http mechanisms to naive browsers.

Kenn Gardels, UC Berkeley

Interoperability and Integration: Finding Semantic Agreement

Francis Harvey

Position

One of the more intriguing issues facing geography, GIS, and certainly open GIS environments is integration. Recent GIS research on interoperability opens perspectives on what I will call integrative interoperability. This term reflects the dynamic needs of information integration in heterogeneous environments. While a rather loose use of the term encompasses any action that merges two previously distinct elements, a more exacting definition that calls on the philosophical tradition of British empiricists going back to John Locke, narrows integration to the process of integrating parts essential to the completeness of the whole. Distinguished from essential parts, integral parts are not necessary, but parts without which the thing would not be as complete and entire.

A useful analogy may be made to the human body. If we consider the whole human body, than the removal of arms leaves just a body. Only through the arms does the body possess the anthropomorphic qualities that make the human body an integrated whole. Geographers have sought fulfillment of this ideal for millennium. Most idiographic studies of geographic regions offers a wealth of rich examples from a wide range of approaches seeking to integrate observations of places into a coherent whole. Systematic geography sought to integrate as well in various, more mechanistic, ways. GIS continued this particular development in analytical cartography through the use of overlay to integrate various themes (Harvey, 1996). Now interoperability, complementing these approaches with strong computational and information processing backgrounds, copes with similar issues.

In modern geography, the fundamental concept of integration has been tackled in numerous ways. Too wide ranging to review here, I focus on the gap between mechanistic and holistic approaches. In mechanistic approaches, integration is the summing of separate parts. Geographers from holistic backgrounds understand integration as the unification of constitutive relationships. In other words, for the holistic tradition in geography, the whole is more than the sum of the parts (Harvey, 1997a). There have been some attempts to combine these perspectives, but without much effect. Recent work on the construction of scientific knowledge and technology provides some insight that links the constitutive relationships to the separation of parts in mechanistic approaches (Agre, 1992; Callon, 1980; Latour, 1987; Latour, 1993; Pickering, 1992; Star, 1995; Suchman, 1987). This work is especially important because it transcends the mechanistic/holistic dichotomy that has encumbered geography.

Integration remains an unwieldy concept because this dichotomy results in a vagueness and vast range of conflicting interpretations. Understandably, it has remained a qualitative concept. We apply geo-statistical measures to indicate the conformity, simple probability, or in Bayesian logic, the conditional probability, of a combination of different attributes. These offer insights into the similarity of attributes and their spatial arrangements, but don't indicate by themselves whether they are integrated. Determining the integration of geographic entities remains an act of human interpretation. Clearly, to turn integration into a manageable concept in terms of interoperability, it is necessary to address these open questions from an empirical perspective. This work sets out to provide the basis for semantically stable automatic integration processing.

All human activities are dynamic and this work on integrative interoperability rests on a foundation that accounts for the social diversity of geographic activities and the use of geographic information technology. Geographic integration is not the Fordist production of spatial widgits. It is the localized, socially contingent, dynamic process of knowledge production. Dependent on the social groups involved, it is contingent on their acceptance, and subject to their rejection. Situated integration cannot be planned, it is the result of actions in a distinct set of circumstances. Integral to this process, the computing technologies involved in geographic integration and interoperability transform the basic patterns of knowledge acquisition, use, the interaction with humans and machines, and the very actions involved in producing meaning. Models for integrative interoperability need to rest on a conceptual foundation that considers both the humans and non-humans involved.

In the dynamic processes of interoperability, geographic integration is a matter of finding semantic agreement. This is not the technical agreement between exchange protocols, but a situated understanding between people involved that the results of an operation are integrated. They should have guidelines to evaluate their decision-making, but no plan can deal with all the contingencies the integration of geographic information raises.

Linking geo-statistical methods with holistic concepts is fundamental to broaden geographic integration to encompass open, highly heterogeneous computing environments of interoperability. Upon the technical foundations that exist, robust specifications that take into account the semantics of information exchange could be built. This is certainly an objective, not something immediately possible, but requiring research on constraints and possibilities for sharing geographic information. At present my goal is the formulation of guidelines for case-by-case application and refinement. The work I have carried out on boundary objects and geographic information system design touches important parts of these issues (Harvey, 1997b; Harvey & Chrisman, in press). The semantical issues of sharing geographic information are broad and still require much work. The issues connected to integrative interoperability require a rigorous formalization around a concept I refer to as situated integrity. This paper defines this concept and takes a step towards its practical refinement.

Integrative interoperability is described in terms of situated integrity. This concept draws on GIS literature on error and accuracy (Chrisman, 1982; Chrisman, 1987; Goodchild, 1996; Veregin, 1989). On this background, I extend these measures to help quantify integration operations. The linkage of statistical measures with semantics is crucial to developing morphisms and formalizations for open data processing. Furthermore, the concept of situated integrity requires a rigorous description that reflects the integration of multiple social and spatial aspects.

Integrative interoperability in open processing environments requires the due consideration of the situatedness of GIS processing. This integration reflects the dynamics of the social groups involved, at the same time providing the technical basis for new forms of interaction. Considering semantics in terms of dynamic processes is the basis for integrating geographic information in the context of its use.