Next: References

From GISystems to GIServices:
Spatial Computing on the Internet Marketplace

Oliver Günther
Institut für Wirtschaftsinformatik
Humboldt-Universität zu Berlin
guenther@wiwi.hu-berlin.de

Question: Why do people buy a GIS? Answer: Because their neighbor has one. Richard Newell of Smallworld Systems told this joke during his keynote speech at SSD'97 - and he did not only refer to Smallworld customers. The truth behind his joke is that GIS are often greatly underutilized. Many customers use only a small fraction of the functionalities offered by their GIS. Some of them are aware of that: they simply do not care about the remaining features. Others are not: they may thus miss functionalities that are actually there and use complicated ways to reimplement them with the features they know. Yet other users may not use their GIS at all: they bought it because they thought it may help them with their problems but then found out that it does not. Some customers may not even have bothered to look: they bought the GIS and left it in the package.

To be fair, this can be said not only about GIS but also about many other types of software. Microsoft, for example, estimates that 90% of Excel's functionalities are used by only 10% of its users worldwide. A large part of requests for new functionalities received by Microsoft each day can be answered simply by telling the customer that the requested functionality already exists. What makes the situation somewhat different for GIS, however, is the relatively high price of a GIS license. GIS come in relatively large packages: a single license often costs 2,000 US$ or more, it requires powerful hardware to run on, and it takes considerable training on the customer's part to use the software in a productive way. For most commercial GIS, potential customers face an all-or-nothing choice. Either they invest a relatively large amount to get the license, the required hardware, and some training - or they do not, in which case they get nothing.

We claim that a large number of potential customers in the second category could well become faithful users if they could do so at a smaller entry cost. Of course, the lower ticket price would not buy them the whole license indefinitely. But rather than putting a time limit on the license, as is typically done, vendors should try to tailor their offerings to the specific user requirements. This could mean in particular that the GIS vendor does not sell a classical system license but a ``service'' to perform a set of GIS-typical tasks. Typical services are, for example, a data conversion, a map overlay, some special-purpose spatial analysis, or simply the retrieval of a specific data set.

The service could be performed either at the site of the customer (client-side computing) or at some site run by the vendor or a third party (server-side computing). In the first case, the vendor software would have to be installed at the customer's site for the time of the computation. Hardware requirements remain basically unchanged compared to the traditional licensing process. Training requirements could possibly be reduced depending on the task in question. Nevertheless, the only substantial difference to traditional licensing is the duration and possibly the scope of the license. In the second case, however, customers would simply make their data available to the vendor software and pick up the results once the computation has been performed. No special hardware or training is required on the user's part.

Payment schemes would follow this service-oriented approach. Users just pay for a particular usage of the vendor's software. This would most likely result in a larger number of customers with a lower per capita revenue than in the case of the classical license business. Depending on the application, however, overall revenue could well increase considerably.

A direct consequence of such a shift from Geographic Information Systems to Geographic Information Services would be the rise of an Internet marketplace [6] for spatial data and services. Anybody with Internet access could act as both a provider and a consumer of related goods.

We recently presented our MMM (Method ManageMent) system, a distributed computing infrastructure that supports the business model and electronic marketplace described above [5]. MMM is a collection of middleware services that facilitate Web-based access to software modules. The idea is that it should be equally easy to post a software module on the Internet as it is to post a Web page. Similarly, it should be equally easy to use such a software module as it is to read a Web page. Some key features of MMM are

A prototype is available on the Internet at http://mmm.wiwi.hu-berlin.de. A CORBA-based reimplementation is currently in progress [7].

MMM is cooperating closely with two other projects that have related objectives. DecisionNet [4][3] is an organized electronic market for decision support technologies. The market infrastructure consists of agents that support consumers and providers in transactions. The decision technologies themselves reside on provider machines distributed across the Internet. DecisionNet is accessible via the World Wide Web, at http://dnet.sm.nps.navy.mil/. SMART [2][1] is another Internet marketplace model with an emphasis on spatial data and related algorithms. Like MMM and DecisionNet, SMART is based on the asynchronous communication between service providers and consumers. It offers query services to obtain data from a provider, function services to model computational tasks (such as conversion between representations), planning services to combine and coordinate different tasks, and execution services to execute a plan on behalf of a customer.

All of these approaches represent important steps toward an open marketplace for computational services. However, there are still several critical issues to resolve before similar schemes will become commonplace. First, the development of appropriate licensing and payment schemes is still work in progress. It is crucial for the success of Internet marketplaces that service providers can be sure to collect fees from all customers that use their services (directly or indirectly). Second, one needs sophisticated algorithms to encrypt the input data for an algorithm without compromising the results of the computation. I.e., the service provider should be able to perform the service without necessarily having access to the data in unencrypted form. This is not always possible but one should know whether it is the case in a given application. Third, there is the issue of data volume. Large data sets are hard to ship and encrypt; sometimes it may be easier to port the algorithm to the machine where the data resides. Fourth, the usage and configuration of services should not be too complicated. Many GIS vendors pride themselves on the turnkey nature of their systems: setup efforts are minimal, and one can start using the system shortly after purchase. This may not always be the case in a digital marketplace type of situation, where users have to select and combine the services they need.

In summary, many of the functions performed by GIS seem to be amenable to a business model that is fundamentally different from the one we see today. At present, GIS users typically own the hardware and software they use. They pay license and maintenance fees to various vendors and they have to train their staff in using the system. The alternative would be a service-oriented approach where users make their input data available to some GIS service center that performs the necessary computations remotely and sends the results back to the user. Customers pay only for that particular usage of the GIS technology - without having to own a GIS. Our MMM system is one example of a communication infrastructure to support this business model.
 



Katja Andresen

Fri Nov 14 15:18:00 MET 1997