Barbara S. Poore

Federal Geographic Data Committee
590 National Center
Reston, VA 20192

703-648-5971(Tel)
703-648-5755(Fax)

bspoore@usgs.gov
 

Abstract: The development of exchange standards for GIS data and metadata over the past ten years is described as a social process that involved enlarging the networks through which metadata could travel. The rapid spread of metadata as opposed to other geographic data standards put forward during the same period was a fortuitous alliance between the formal languages in which the metadata was expressed and emerging networking technologies. The standard allowed for both a local narrative and the ability to link to a larger community. This was, and still remains, the source of its power. These issues are examined using ideas from the literature of science and technology studies.
 
 

Keywords: GIS. Metadata. Standards. Data Quality. History of GIS. Diffusion
 
 
 
 

AIn light of a frenzied drive for GIS acquisition in business and government, a drive to set a standard definition--a metanarrative of metadata--becomes paramount and GIS appears necessary to capture meaning over one=s geography

A concern for producer specification of geographic data quality, handled as part of a standardized package, goes back more than ten years (Chrisman 1984) and the idea of a structured approach to the description of geospatial data has been progressively codified in such federal standards as the Spatial Data Transfer Standard (SDTS) and the Content Standard for Digital Geographic Metadata. (National Institute of Standards 1992; Federal Geographic Data Committee 1994). Intended at first to contain producer-generated "truth in labeling" information, metadata requirements have come increasingly to tightly specify the formal presentation structure of geospatial metadata. At the same time, the emphasis has shifted from the producer to the consumer of data, and metadata have been allied with other technologies and configured to bring imagined users into a place of responsibility for evaluating the quality of digital geographic data (Chrisman 1986; Federal Geographic Data Committee 1995). This paper will discuss the construction of metadata as a social process embodying an ambiguous, figurative spatiality that allows metadata to become the carrier of geographic knowledge that is at once global and local.
 
 

It will be argued that it was in the interest of the federally sponsored committees that wrote the federal metadata standard to ensure its adoption. That to disseminate the standard it was necessary to express metadata using scientific terminology, to subject it to the witnessing of a professional elite, to enroll allies along the way, and to bundle it with technologies of dissemination. That the witnesses, the allies, and the technologies, as practiced in local situations, influenced in turn the course of metadata adoption and changed the meaning of metadata and the way metadata conveys local knowledge. This scenario follows the general pattern sketched by Bruno Latter to describe what he calls "science in the making" in which "[the object--in this case, metadata] is not only collectively transmitted from one actor to the next, it is collectively composed by actors." (1989; 104) Drawing on the writings of Fujimura it will also be argued that the standard, along with its allied technologies, acts as an interface between multiple social worlds. (1992; 176)
 
 

Metadata, black boxes, and GIS
 
 

Metadata makes a fine case study of science in the making since it is an example of a technology that has not yet been "black boxed." This term, used by Latter, describes the result of a hybrid process sometimes referred to as technoscience through which science by forming alliances with technologies, succeeds in becoming irreversible and unproblematic to the user. As black boxes, technologies and the sciences that recruit them move smoothly through society, enrolling convert after convert. Making a local telephone call is an example of a practice that has become a black box. There was the initial science of Bell, and the technologies that were recruited to support this practice, including the telephone itself, the wire into the house, the switching station down the street, and the network of wires and switches across the city, but these remain unproblematic and taken for granted. The science and the standardized technologies of telephony are settled and remain in the background. They act in the present as a unified whole although at one time they may not have been unified. (Latour 1989; 131)
 
 

The question of unity has a particular relevance to the study of GIS issues since proponents of GIS have claimed integration as its defining characteristic. (Dobson 1983 and 1993; Openshaw 1991) The synoptic vision, or the ability to simultaneously see representations of many different geographic elements together in the simulated place of a glowing screen is paradigmatic for GIS and a strong legacy of its origins in cartography. Although on the surface GIS appears to be integrated, metadata most definitely are not. Many have argued that metadata are spatial data and should be handled within a GIS system (Nebert in Federal Geographic Data Committee 1992, 7; Chrisman 1995), but the black box of metadata has not closed. In the CSDGM data quality information was split off from information about features and data structure that were part of the earlier standards. Paradoxically, this separation may have paved the way for metadata to travel through society in a way that standardized versions digital cartographic features and data structures have not.
 
 

The prehistory of metadata
 
 

In 1982 the U.S, Geolgoical Survey convened an independent body, the National Committee for Digital Cartographic Data Standards (NCDCDS) under the auspices of the American Congress on Surveying and Mapping. The committee, composed mainly of representatives from academia, industry and the federal government, saw their task from the first as a scientific process: "In order to develop effective digital cartographic standards, the most efficient approach is to follow the same strategy one uses to solve scientific problems. Therefore one begins with the general considerations and progressively works down to the specific detailed problems and then back up to the general problem."(NCDCDS, 1987, 2)
 
 

So the cartographic data standards are seen as a form scientific speech, and the process of developing standards was to provide "a professional forum" where those in the profession could witness and themselves testify about the proposed speech. The explicit goal of the standard was to expand the use of geographic data by encouraging the sharing of data among different organizations. These three characteristics--appeal to scientific method and presentation, gathering the consensus of experts in a forum, and expanding use or enrolling others are common to the three standards under discussion--the draft NCDCDS, the SDTS, and the final FGDC metadata content standard. The tension between scientific speech generated by a group of experts on the one hand and the desire to enroll new users on the other affected the development of these standards. The forums are a place, not unlike the laboratories of the physical scientist, where witnesses testify and establish the necessary trust in the process. (Shapin 1998)
 
 

The draft NCDCDS standard was developed through several "cycles" of writing, circulation, comment, and further writing, It was finally published in 1987 and formally presented to the public at a professional conference, Auto-Carto 8, in Baltimore that year. Further comments were solicited and brought back within the committee. The SDTS developed directly out of this draft standard. The work of consolidating comments to the draft standard, reaching consensus, and shepherding the standard through its adoption as a Federal Information Processing Standard, which occurred in 1992, was undertaken by a special task force, chaired by the U.S. Geological Survey. This task force was again drawn from a professional domain, and had a slightly overlapping membership with the earlier committee.
 
 

The three topics that had formed the core of NCDCDS deliberations--data organization, data set quality, and cartographic features, were incorporated into the SDTS. As a move to gain allies, the committee also incorporated ideas from a proposed standard for data exchange which had been developed by the Standards Working Group of the Federal Interagency Coordinating Committee on Digital Cartography (FICCDC). Data exchange was meant to enable the transfer of data between "noncommunicating parties using dissimilar computer systems, preserving the meaning of the information." (NCDCDS 1987). Both the NCDCDS draft and the SDTS contained data exchange modules--groups of functionally related data fields such as spatial reference, security, or data quality--that transferred as a package. Even though parts of modules could in theory be linked to other modules, modules were self-contained and could be transmitted in any order, rather the way bundles of information were separately wrapped and transmitted through the emerging Internet.
 
 

The modularization of information was a concrete expression of the desire to transfer the responsibility for assessment of data quality from producer to user. Chrisman had argued that data quality was related to the issue of fitness for use, as opposed to quality concepts of "meeting an expectation" or "conformance to a standard." These latter definitions applied to older, industrially managed systems of cartographic production and control where the data producer in effect used quality measures to discipline its workforce. (Chrisman 1986, 351; Beniger 1986) Fitness for use implies that there is no one way to use data and that the user is the ultimate judge. To evaluate fitness for use one would need to have the quality information in advance of the actual data transfer. The creation of modules makes that more of a reality. The impetus behind the NCDCDS draft was to facilitate the transfer of data among federal agencies--at that time the major data producers. However, as the standards developed so did an emerging sentiment in favor of universal access to geographic data as a public good? This movement saw a much wider stage for the eventual deployment of data quality information, extending a national conversation about geographic data to data producers and users outside the federal government. (National Research Council 1993)
 
 

From producer to user--the metanarrative of metadata
 
 

The deliberations that occurred over the exchange of spatial data between dissimilar systems were at first largely confined to the federal data producing agencies and to academics. These conversations followed a pattern of convergence/divergence/convergence where a small group, broadcast to larger groups of professionals, brought back into a committee, and so forth through several iterations generated ideas. This process is similar to that described by Fujimura in her discussion of collective work in cancer research--the aim is to understand science as collective action from the viewpoints of all actors. "All actors are simultaneously attempting to interest others in their concerns and objectives. The final (or temporary) outcomes of these efforts are constructed through the process of negotiation, articulation, translation, triangulation, debating, and sometimes even coercion." (Fujimura 1992, 171)
 
 

A more extensive case study of the creation of the metadata standard would expose these multiple viewpoints, but the important issue for this discussion is that even though these data conversations may have been as varied as those Fujimura enumerates, the groups that were negotiating were part of a fairly closed society. Makeup of these committees was restricted to groups of professionals who were known to others in the group. Many of the members were data producers or academics who had helped define the discipline. Larger audiences were asked to comment on the draft at meetings of professional societies and yet attendance at these meetings is generally restricted to those with the same interests and the means to attend. This notion of a rather closed society to a large extent determined the languages within which the standards were expressed. The gentlemen (and a few ladies) who attended meetings were accustomed to the science speak of standards and the production languages in which the standards were expressed. If they were not personally able to interpret these languages, they would have staff or students who could. Both the draft NCDCDS and the SDTS specified production rules that would guide implementers in using the standard. Two of those languages, the Backus-Naur Form which was used to express structure, relationships, and layout of the standards constructs, and ISO 8211, a media-independent encapsulation standard for physical data exchange have bearing on the development and representation of the Content Standard for Digital Geospatial Metadata.
 
 

The Backus-Naur Form (BNF) is a metasyntax in computer science; that is, a language used to describe other computer languages. It was first used in the 1960's to describe ALGOL 60, a portable language for scientific computing. It has been used widely in Internet work, though not in a standardized form. (Howe, 1997) BNF takes the form of a mathematical equation as in:
 
 

ISO 8211 is a data transfer standard. It is a media-independent data encapsulation standard that acts as a wrapper for the actual data being transferred. ISO 8211 specifies the mapping of data elements to fields and subfields, the record structures, the allocation of records to files, and the allocation of files to volumes. In this sense, then, ISO 8211 is also a metalanguage. ISO 8211 is a very powerful data exchange mechanism for sequential media such as tapes and diskettes because it incorporates the data structure itself onto the physical medium thus allowing the receiving software to interpret the data without accompanying information (or metadata). (Podczasy and Carlson 1994) A sample of the encapsulation language for the data quality module is rendered as:
 
 

...and so forth
 
 

In choosing to express the standard in these scientific languages, the committees that created NCDCDS and SDTS were not unique. Other standards use BNF notations or other types of metalanguages to express constructs. The appropriateness of the metalanguages is also not in question, and in fact, is beyond this author’s ability to judge. What is of interest to this paper is that these metalanguages, while they may serve a logical and appropriate function in the writing of the standards, also function as a devices to signal that allies from other, more established communities have been recruited. The languages point to established, authoritative ways of speaking--those found in the computer science community and in an endorsed international standard. By this means, the committees creating the standard are in effect saying: "there, there is your black box. We’ve linked it to universal scientific languages--take it and run with it."
 
 

But it is not enough to simply build a technoscience artifact and link it to more established artifacts. One of the roles of a standard is to bring order to emerging disciplines and technologies. In 1992 too many voices had spoken and the situation in the geographic information community was chaotic. Different applications used different systems that were incompatible, different organizations were collecting the same data sets over the same areas at all scales, systems--sometimes within the same organization--could not talk to one another. (Office of Management and Budget 1990, National Research Council 1993, 25; Clinton 1994) The standards under discussion were all intended to prescribe a way that systems could communicate. Standards, and particularly the metadata standard were touted as the ultimate solution to the data Tower of Babel. In this sense, the metadata standard acted as a metanarrative as defined by French critic Lyotard, that is "any large idea or presence that exists as an uncontested phenomenon outside the realm of human social action." (Lyotard 1984) But, to become effective, these standard descriptions of data quality, of cartographic features, of data organization had to circulate widely outside the places in which they were created and be embodied in implementations. A way had to be found for the standards to act at a distance. (Latour 1989)
 
 

Acting at a distance
 
 

Latour describes the encounter of French explorer Lapérouse in 1787 with Chinese living on the island of "Sakhalin" in the East Pacific. Lapérouse and his shipmates are interested in solving a navigational issue--whether Sakhalin is an island or a peninsula. The Chinese have a great deal of local knowledge, but they do not travel beyond their island and have no interest in what the French want to know. Their knowledge, therefore is looked on by the French as crude and inexact, but the French persuade them to draw a sketch map anyway. This map and the information about the island and on the strait that separates it from the mainland is carried back to the French king. This transmitted knowledge, embodied or crystallized in an artifact, will in turn allow future French voyagers to become stronger than the Chinese in certain respects When they return to that area they have already in effect seen where they are going. Latour defines the knowledge that is carried back in a pragmatic way as familiarity with things that are distant so that "whatever happens is only one instance of other events already mastered, one member of the same family." (Latour 1989, 219) To act at a distance, one must render events, places, and people mobile so they can be brought back; stable so they will not lose their meaning as they are brought back; and combinable--so they can be cumulated and aggregated. The metadata standard allowed all these things to happen through the manipulation and structure of languages or text.
 
 

The metalanguages in which the earlier standards are expressed are at once a distinct barrier and a tool for acting at a distance. At first glance these languages are the languages of machines, not people. How could these language be understood outside the technoscientific places where they were created? How could they speak to the tax assessor, the student, the forester, the surveyor? There was a growing perception among the very people who created these two standards that many such ordinary users were clamoring for access to geographic data sets. One solution to enlarging the base of users for geographic information was wide dissemination of data quality information--opening the lines of communication between producer and user. Acting on this perception, the Federal Geographic Data Committee called a specialist meeting in 1992, to discuss the data quality issue (Federal Geographic Data Committee 1992). This meeting was pivotal in linking ideas about data quality that had been developed in the two previous standards with the idea that data quality information, or metadata could and should be managed in a controlled way. It was also at this meeting and in the crafting of the federal metadata standard that followed, that the metalanguages were used to create cycles of accumulation that strengthened the growing networks of people interested in geographic information.
 
 

The crystallization of the metadata standard
 
 

The 1992 forum on metadata brought together people from a number of organizations that either had an interest in metadata or had experimented with metadata. The size of the community at this point was about 150. The vast majority of those who attended were from various branches of the federal government, with a smattering of academics, and representatives of private companies and local government. Most of the participants had a background in geographic information systems. There were few representatives from the computer science community or the library community who might have brought expertise on metadata from different domains. Dr. Michael Goodchild in his opening speech, equated compiling metadata to "extending the membership of the club by expanding the set of people who can use the information." Metadata were a process of communication between provider and user, a social process that extended the reach of spatial data. (Goodchild in Federal Geographic Data Committee 1992) So the power of metadata to broaden the networks interested in geographic data were clearly forseen.
 
 

At this forum, another ally was recruited--a metadata standard that had been in development since 1990 by the American Society for Testing Materials (ASTM). Although this group is recognized by the American National Standards Institute, and thus has standing in the official U. S. standards process, ASTM is a forum for testing standards for industrial products rather than for GIS or information technology standards. The geospatial metadata standard was thus in the company of standards for cementitious and concrete materials and soaps and other detergents. (ASTM 1998) This insured that the quality control and legitimation mindset of the industrialized producer was embodied in this developing standard at its origins. The recruitment of this standard and its use by the Federal Geographic Data Committee as the foundation for the CSDGM is a clear demonstration of bricolage--constructing something out of the chance elements that come to hand. This type of activity is at the heart of the technoscience endeavor. (Latour 1993, 1)
 
 

Despite the quality control bias of the standard, the author of the ASTM cited user-centric needs for metadata: to build potential clearinghouses of geographic data, to provide "extended context and quality information to potential customers," and to provide "data automation techniques to local and remote users of the data for evaluation and implementation." (Nebert in FGDC 1992, 70). The ASTM standard incorporated data quality information from the SDTS, specifically information related to lineage, positional accuracy, attribute accuracy, logical consistency, and completeness. Lineage information was moved into a separate category of its own and no longer defined as data quality. The requirements for an extended narrative on data quality in SDTS was sharply abbreviated. For example the faithful reporting of a number of tests in the SDTS section on attribute accuracy, were summed up in a simple sentence in ASTM. (FIPS 1992, 22; FGDC 1992, 80). This information was even further truncated in the CSDGM to the point where persons seeking information on tests for data quality are referred back to SDTS; the tests are neither described nor named . (FGDC 1994).
 
 

Juxtaposing the STDS and the ASTM standards in this way makes the difference clear--both in content and in structure, the ASTM standard is a radically stripped-down version of the data quality section of the SDTS with identifying information pulled from other sections. The major feature of the ASTM is the abandonment of long narrative requirements for data quality and the establishment of a format that emphasizes small units of information built, through logical parent-child relationships, into a larger structure. Thus, instead of dividing the domain of geographic information into the three separate topics, data quality, cartographic features, and data organization that are more or less walled off as in SDTS, the ASTM and the FGDC standards have an essentially atomic structure. Small bits of information are linked into a larger whole. This may relate to the difference in world views of the creators of the standards. The academicians and government officials who participated in the development of the NCDCDS and the SDTS were largely from the generation that evolved digital cartography out of making maps. To them, digital cartography was very much linked to the notion of topology and the necessity for complete partitioning of planar space. The ASTM and FGDC standards reflect a different way of thinking about space and the relationships of objects--that space is built up of objects and their logical connections or patterns of interrelationships. (Curry 1996)
 
 

The atomic structure of the standard was explicitly addressed as a positive issue in the review comments of the FGDC’s CSDGM which grew out of the ASTM standard.(Standards Working Group 1993) The structure permitted elements to be linked into more complex statements, and also allowed for the crossreferencing of individual elements. Thus names and addresses of data producers could be used several different places in the standard. This type of thinking was, of course, common to data base structures of many kinds, but beyond just sound data management (write once, use many times), metadata was envisioned as belonging to a new category of communication--hypertext. Hyperlinking allows a radically new method of enrolling converts. For the ASTM and FGDC standards are envisioned as working in conjunction with emerging Internet technologies. In another instance of technoalliance, the creators of the federal metadata standard conceived of a distributed, Internet-based clearinghouse of geospatial data by which users throughout the country could search for metadata on specific data sets residing on many servers. (Nebert 1995)
 
 

Widening the networks--metadata and the clearinghouse
 
 

The clearinghouse would leverage the atomic structure of the metadata elements to allow complex searching for particular fields in indexed metadata records. To achieve this, support for bounding rectangles and other field unique to the geosptial metadata standard were built into a piece of software that supported clients using Z39.50, a search and retrieval protocol from the library community. The search could be initiated from any web browser through a gateway service which ran the Z39.50 server. The gateways indexed the holdings of many other machines. (Nebert 1995)The strength of the clearinghouse concept is that it allows one to search many servers at once for specific fields in metadata records. Hits are returned as hypertext headlines that when opened, reveal a full metadata file. Often links are embedded that take users to the producing agency, to a browse graphic, or to the downloadable data set itself.
 
 

The process is not unlike searching through a well established Internet search engine such as Lycos or Alta Vista. The difference is that because the metadata is structured and organized into a hierarchy of containers, very specific results can, in theory, be achieved. A data user can say, "Give me all the data sets produced since 1995 in the counties surrounding Houston where the theme is water, and the data producer is the Texas Natural Resources Conservation Commission." In theory, such a mechanism would avoid the overload of irrelevant information that is common when using Internet search engines. The downside of the clearinghouse concept is that it requires discipline on the part of its members and knowledge on the part of its users. It is not enough that one throw any old document on a website for a geowebcrawler to pick up. One must create a very precisely structured metadata record that can be read and indexed by a machine. One must also store the index on a specifically configured server so that it can be targeted by a clearinghouse search. In order to make full use of the complexities of the metadata, the user needs to know quite a bit about the what servers are active, what geographies they cover, what data sets are indexed, what kinds of data they might contain, as well as the complex field structure of the metadata itself.
 
 

The tension between producer and user is clearly visible in the evolution of the Content Standard for Digital Geospatial Metadata. When the standard was written in 1994, the purpose of metadata is said to be to "organize and maintain an organization’s investment in data, provide information to data catalogs and clearinghouses, and provide information to aid data transfer." (FGDC 1995) A mechanism for data users to access a clearinghouse was only one use for metadata. The standard was written as a large umbrella and meant to be infinitely customizable. It contained all possible terms that any data producer might use to describe his or her data set and specified logical relationships among these terms. For example, different and mutually exclusive terminologies for raster and vector data sets are specified. But implementations of the standard were aimed toward clearinghouse use. The prospectus for the 1998 revision of the Content Standard makes this use clear: "the standard was developed from the perspective of defining the information required by a prospective user." (FGDC 1998) In order for the standard to be available through a clearinghouse, the formal arrangements of the elements had to be tightly specified. The files had to be parseable. The narrative of a data producer, which might consist of a simple description of the contents of the data set, the sources of the data and the procedures that were performed on the source data had to be made into a machine narrative--in essence a program that would run. To achieve machine narration, the standard could not remain purely a content or semantic standard--syntax or structure had to be specified. The Content Standard for Digital Geospatial Metadata, therefore continued the formal specification conventions of previous standards. The Bakus-Naur form, used for part of SDTS, describes the hierarchy of container elements, including the repeatable, mandatory and optional elements. While not as confusing as the ISO 8211 the metadata production rules are daunting:
 
 

Data_Quality_Information=

0{Attribute_Accuracy}1 + Logical_Consistency_Report+Completeness_Report+0{Positional_Accuracy}1+Lineage+(Cloud_Cover)
 
 

and so forth.
 
 
 
 

From this one can deduce that a data quality information report consists of an attribute accuracy statement, if applicable, plus a logical consistency report, a completeness report, a positional accuracy report if applicable, a lineage report, and an optional report on cloud cover. This language is highly complex and requires careful reading with a cheat sheet in hand. The use of a scientific production rule gives the standard the aura of alliance with the world of computer technology, and it is useful for programmers who wish to translate the standard into software (Schweitzer 1998).
 
 

For ordinary users, even those who are fairly computer literate however, these production rules needlessly complicate the comprehension of the standard. Almost immediately on its approval, different representations of the standard began appearing. There was a workbook version that clarified the meaning of data elements and gave examples of usage. This workbook represented the standard as a series of nested boxes that graphically portrayed the parent-child relationships among elements (FGDC 1995). Those who wished to represent the standard numerically used the numeric equivalents (1, 1.1, 1.2 and so forth) which accompanied the element names in the standard. Some implementations used element names with indentations to capture the nested relationships. Many of these variations turned up on the websites of data producers. (Hart and Phillips 1998)
 
 

In order to be used in the clearinghouse, a metadata file had to conform to a certain format. The element names had to be correctly spelled, presented in the right order, and indented a certain way. Peter Schweizer wrote a metadata parser that would read such a text file and validate it, then output the file in several formats including a version in hypertext markup language that could be used on the Internet and a version in the tagged standard generalized markup language that could be indexed by Z39.50-compliant software and used in a clearinghouse.
 
 

So the representational tension embodied in the standard was obvious. The ambiguity of the metadata standard--that it was at once flexibly constructed and tightly specified--gave it enormous power. It could both recruit new converts by acting as a boundary object that could negotiate many different worlds and it could act at a distance by standardizing the delivery of spatial information remotely. On the one hand, data producers both in the federal government and outside could use the very generality of the standard’s existence as a collection of data elements, some of which pertained to their local situation, some of which did not. They were encouraged to use these elements in a mix and match fashion to describe their local geographies. Their narratives of data quality could be served over the Internet and read by many users. On the other hand, without the tight specification of metadata structure and all that entailed, the full spatiality of their narratives could not be exploited. Without a parseable metadata file, the metadata cannot be indexed, the clearinghouse node cannot be constructed, and the spatial search cannot be enabled. The data producers could describe their local geography in as elaborately as they wished, but without the narrative of the machine, their local narratives could not join that collective that can travel everywhere. Without structure, the local narratives of space could not act at a distance. Even though non-conformant metadata were posted to a website, users from afar would have to pursue the knowledge without knowing it was there. Once they found the knowledge, it could be in a format that they were unfamiliar with, and therefore they might have trouble understanding it. Encountering metadata divorced from its structure would be like Lapérouse encountering the local geographic knowledge of the Chinese natives. Metadata that is structured, parsed, indexed, served is metadata that acts at a distance. It is mobile, stable, and combinable. "Give me that data set from that county in Texas," we say, and the machine answers "yes."
 
 
 
 

References
 
 

Beniger, J. R. 1986. The control revolution:Technological and economic origins of the information society. Cambridge, MA: Harvard University Press.
 
 

Chrisman, N.R. 1984. The role of quality information in the long-term functioning of a geographic information system. Cartographica. 21 (3 and 4):79-87.
 
 

---. 1986. Obtaining information on the quality of digital data. In Proceedings Auto Carto London, ed. M. Blakemore, 350-58. Vol. 1. London: Auto Carto London, Ltd. N.
 
 

---. 1995. Re: Metadata exchange format ideas. electronic mail message to members of list NSDI-L, March 31, 1995; archived at www.fgdc.gov.
 
 

Clinton, W. J. 1994. Executive Order 12906: Coordinating geographic data and access: The National Spatial Data Infrastructure. Federal Register, 59:71, April 13, 1994,17671-17674.
 
 

Curry, M. R. 1996. Space and place in geographic decision making. http://www.sscnet.ucla.edu/currym/IEEE.htm.
 
 

Dobson, J. 1983. Automated geography. The Professional Geographer 35 (2):135-43.
 
 

---. 1993. The geographic revolution: A retrospective on the age of automated geography. The Professional Geographer 45 (4):431-39.
 
 

Federal Geographic Data Committee. 1992a. Information Exchange on Spatial Metadata. Reston, VA: Federal Geographic Data Committee.
 
 

---. 1995a. Content Standard for digital geospatial metadata workbook. Washington, D.C.: Federal Geographic Data Committee.
 
 

---. 1995b. Factsheet on Metadata. Washington, D.C.;.
 
 

Fujimura, J. H. 1992. Crafting Science: Standardized packages, boundary objects, and "translation". In Science as Practice and Culture, ed. A. Pickering, 168-211. Chicago: University of Chicago Press.
 
 

Hart, D., and H. Phillips. 1998. Production Rules of CSDGM. http://badger.state.wi.us/agencies/wlib/sco/metex/prodrule.htm; accessed February 1, 1999.
 
 

Howe, D. 1997. Free on-line dictionary of computing. http://wombat.docic.ac.uk.
 
 

Latour, B. 1983. Give me a laboratory and I will raise the world. In Science observed: Perspectives on the social study of science, ed. K. D. Knorr-Cetina and M. Mulkay, 141-70. London: Sage Publications.
 
 

---. 1987. Science in action: How to follow engineers and scientists through society. Cambridge, MA: Harvard University Press.
 
 

---. 1993. We Have Never Been Modern. Cambridge, MA: Harvard University Press.
 
 

Lyotard, Jean-F. 1984. The postmodern condition: A report on knowledge; translated by G. Bennington and B. Missumi. Minneapolis: University of Minnesota Press.
 
 

National Committee for Digital Cartographic Data Standards. 1987. Issues in digital cartographic data standards, Report #8: A draft proposed standard for digital cartographic data, Harold Moellering, editor. Columbus, OH: National Committee for Digital Cartographic Data Standards.
 
 

National Institute of Standards. 1992. Federal Information Processing Standards Publication 173: Spatial Data Transfer Standard. Washington, D.C.: U.S. Government Printing Office.
 
 

National Research Council. 1993. Toward a Coordinated Spatial Data Infrastructure for the Nation. Washington, D.C.: National Academy Press.
 
 

Nebert, D. D. 1995. Use of the Isite Z39.50 software to search and retrieve spatially-referenced data. http://csdl.tamu.edu/DL95/papers/nebert/nebert.html. N.
 
 

Office of Management and Budget. 1990. Circular No. A-16: Coordination of surveying, mapping, and related spatial data activities. Washington, D.C. Executive Office of the President.
 
 

Openshaw, S. 1991. A view on the crisis in geography, or using GIS to put Humpty-Dumpty back together again. Environment and Planning A 23:621-28.
 
 

Podczasy, K., and R. D. Carlson. 1994. Spatial data exchange standards harmonization for the Defense Mapping Agency. In GIS/LIS Proceedings. Bethesda, MD: American Society for Photogrammetry and Remote Sensing.
 
 

Schweitzer, Peter. 1998. Personal communication.
 
 

Shapin, S. 1998. Placing the view from nowhere: historical and sociological problems in the location of science. Transactions, Institute of British Geographers 23:5-12.
 
 

Sieber, Renee. 1998. Computers in the Grassroots: Environmentalists, geographic information systems, and public policy. Unpublished Ph. D. dissertation, Rutgers University, New Brunswick, N.J.
 
 

Standards Working Group. 1993. Adjudication of the comments to the draft "Content Standards for Spatial Metadata.". Federal Geographic Data Committee, Washington, D.C.