Geoffrey Dutton
Some remarks about geospatial interoperability, written in response to questions posed by Guy Maynard to members of a discussion panel at the 1995 URISA conference.
Any application has strong and weak points, and most users only work with one or two mapping/GIS packages. This may require them to access data created in a "foreign" GIS; even when standard interchange formats are available, they tend to lose information, create multiple representations of spatial data, and are just too inefficient and cumbersome to use on a regular basis. The tired model of export/import is showing its age nowadays.
Both closer and farther than most practitioners might believe. The ideal is getting closer because major vendors are putting energy into achieving some form of interoperability, a recent trend. But the goal is elusive, primarily because it depends on a robust array of services that only operating systems can effectively deliver. The OGIS toolbox is therefore dependent on vendors and consortia that develop Unix, Windows, VMS, Mac and other OS software, not merely on the GIS vendor/research/standards community. The vector/raster and object/ relation dichotomies further aggravate achieving true consensus.
I have observed two principal aspects of OGIS interoperability. The first is transparent access to spatial data, similar to how RDBMS data is now handled across applications and platforms. The second involves asking GIS software to perform services for other (GIS or non-GIS) applications. The former is most often handled using API calls, the latter by scripting. Both techniques lean upon layers of OS architecture, and cannot be achieved simply be cooperation among GIS vendors. To some extent, OOP and OODBMS models incorporate both approaches, but we are quite far from seeing a full embrace of CORBA or any similar standard by all major vendors of GIS and mapping software.
Users of OGIS will expect to have a common language that works across many platforms, applications and datatypes. The language may be packaged in the form of a GUI, but a regular syntax and grammar must live at its heart. The quest for a language of spatial relations and operations is an intense and continuing one, but little research in this area has born edible fruit for actual users. To my mind, the most successful is Dana Tomlin's efforts to characterize spatial transformations in a canon of operations on fields. No analogous or equally accessible approach to manipulating features or objects has yet come to my attention or held it very long. I feel that while system-level tools are needed and must be held up for inspection, they cannot be a substitute for agreement on how to characterize spatial objects, operators and operands that users expect or need to manipulate. Achieving this kind of consensus is a necessary, delicate, higher-order task.
All of the above and more; one technology that deserves more credit is that which enables the internet. Without the immediacy of cyberspace, we would still be tempted to think that SDTS is the solution. But now that so many spatial data archives are going online, both publishers and subscribers need good tools to evaluate how generally useful such datasets really are, despite their seductive accessibility.
Another resource that will make a big difference is spatial metadata; When data quality isn't documented, data users may have foggy notions about how well someone else's data can serve their own purposes. Now that metadata is available in canonical forms, potential users can begin to evaluate fitness for use. For too many years, this was left to chance and good will. Even though spatial metadata standards have emerged, users still lack tools to filter and assess metadata content for practical purposes. Its implementation still has a long way to go.
They help because they indicate a consensus that OGIS can and should be achieved. They don't help because there are too many of them. Who can keep track of all these developments, or understand how they fit together? The quest for OGIS is unfortunately convolved with the battle for OS hegemony, along with other systemic aspects of cross-platform computing. In the end, OGIS will be achieved no faster than other kinds of digital data integration, and probably slower, due to the bulk and complexity of spatial data. We should recognize this and not grow impatient.
It depends what you mean when you say "open GIS." It can mean as little as accessing Arc/Info data from an Intergraph platform, or as much as processing environmental data from all over the internet on distributed CPUs and publishing hypermedia documents that include discussion, data and animated renderings of inputs, outputs and metadata on a web server. In any case, it demands adoption of standards in a number of areas, many of which already have a superabundance of protocols. Until computing platforms are truly interoperable, and until there is more agreement on canonical data representation schemes, OGIS will remain a chimera.
I've never come clean before, but I have psychic powers that I've used in secret for years to shape my career path to achieve maximum returns. Here, for the first time in public, are my perceptions of the future of GIS:
Six Months: Spatial metadata in FGDC format will start to go online. Windows 95 will (probably) be released, setting back PC users almost two years. More GIS/Mapping packages support SDTS and additional exchange formats.
One Year: Jack Dangermond will be born again; the Lord will tell him to put Arc/Info coverage format into the public domain. Intergraph will merge with Genasys to form Intersys; the joint venture will be dedicated to implementing OGIS.
Two Years: A public domain GIS from Europe will sweep the field; it will integrate many data types into a unified representation that allows 0-5D analysis and display. OGC will issue a spec for OGIS compliance, which will be generally ignored. GIS will seep into policy-making at an increasing rate.
Five Years: Intersys deploys its long-awaited second release of its GIS; the first was limited and flawed, but gained converts to its modular, extensible, object-oriented, distributed architecture. EuroGIS counters with a distributed GIS/RS platform featuring internet servers based on SIMPLEX spatial data stores. ESRI relies more heavily on VAR solutions and stimulates record sales by bundling workstations and peripherals at no charge.
Ten Years: Nobody will care about GIS any more than they care about "computer graphics" or "multimedia;" all such stuff will be taken for granted in the digital world community. Much GIS activity will be devoted to integrating local, regional and global environmental, demographic, economic and infrastructure modeling efforts across the net. Stresses and strains on planetary resources will focus attention on this enterprise. There will still be no easy answers.
Eventually: The planet suffers a catastrophic environmental collapse due to the unmitigated success of market economics. The few people who remain alive dedicate themselves to the care and feeding of the global digital network, and form religions based around the revelations of geodata modeling software. All surviving corporate entities merge into the Atman, but AT&T and IBM still fight.
There is a larger context that concerns me as much as the technical OGIS issues; this concerns who gets to determine what GIS options users have, and why that process -- as well as much of what is going on in the western world right now -- is deeply flawed. We live in very scary times, and what we do as GIS professionals makes a difference in our collective future.
I speak as both a developer of GIS and other software applications, and as a user of products in those categories. In neither role have I been able to achieve all I wanted, but somehow I don't believe my limited acquaintance with the state of any art is entirely responsible for my frustrations. I am also aware that I work within a set of global intellectual enterprises that have no central control, but perhaps enough feedback loops to remind themselves of their priorities. These are not market mechanisms, and without email and related connective technologies they would wither, if not die. Direct access to information, services and other users is now crucial for science and a great many other enterprises. It is up to us users to safeguard the internet services we rely upon from incursions of infomercials and other forms of commercial speech.
Sadly, "user" is usually tantamount to "consumer;" a quasi-black body that receives energy, goods, services and information, but emits only market signals in the form of buying and bitching. While producers may listen to consumers, and do modify their production in response to their reactions, users aren't often pro-active regarding vendors (but may be in other respects), and the best ideas that users have often fail to motivate producers. This is partly due to proprietary concerns: most vendors aren't willing to develop products with core technologies they do not own or control, most bureaucrats are uninterested in priorities other than their own, and most users are just too busy to offer more than sporadic feedback.
Massively passive users and consumers is exactly what our corporatist culture cultivates worldwide. Government and commercial organizations both fish in a sea of customers that is most bountiful when it is most tranquil. It makes sense: keeping consumers preoccupied with cheap thrills and their shortcomings engenders dependency and drives sales. But it still doesn't necessarily give us what we need, regardless of what we have been led to want.
Most of us are aware that the stakes are very high. Not only are fortunes made on lost based on whose GIS best optimizes or predicts land uses, our fortunes are subject to availability of far-flung resources that may be used up before their significance is understood. GIS is a tool that can empower, document and even anticipate the perpetration of greed on the fabric of society and the biotic resources which sustain it. Until now, it has been used to discover knowledge and foster stewardship; increasingly it will serve to empower exploitative activities that are likely to be increasingly deregulated. The genii is out of the bottle, and the three wishes that it grants better be global in space, geologic in time and generous in spirit.
There are big differences between achieving consensus on technical standards for such things as plugs and sockets, or even network protocols, versus the essential characteristics of spatial data and operations upon it. One of the biggest ones is that spatial datasets come from all over the planet, and, once in users' hands, tend to mutate, miscegenate and propagate unchecked. This indicates that spatial metadata must itself be spatial data. The provisional standard the U.S. now has for documenting spatial data files does not recognize this, and much of the community involved in its development appears not to understand the need for it.
Finally, whatever their quality, spatial databases only model how things we name are laid out in the world, abstracting its diversity and complexity. Most spatial data standards are based on conventions used in drafting maps, and ignore many aspects of function, role and process. In my view, most spatial data handling paradigms are pale shadows of immanent, personal and collective perceptions of what it is like to inhabit our corner of the space-time-foobar we like to call the real world, and that we should not be satisfied with classifying cartoons into categories and calling the result a standard. Call me unpragmatic, but if I am to accept more GIS in my reality, I'd like a bit more reality in my GIS.
I challenge GIS researchers, vendors, standard-makers, data publishers and users to meld their minds in a quest to make spatial data so rich that it doesn't need any metadata it doesn't already carry. Together, I think we can make it happen.