The spatial databases of tomorrow will be shaped by two major influences - an improved ability to model reality and users demanding increased database access. Other potential influences such as price, platform, speed, and storage capacity are considered here to be less important. This paper first outlines the limitations of current spatial databases and database design and then develops a conceptual framework for future design and access.
In the last decade spatial databases have continued to be designed and developed at a frenetic pace. This pace, in relative terms, is slowing - the quantum leap has yet to occur. Current spatial databases have a short history - not much more than 15 years. For the majority of that time they have evolved out of the proprietary pursuits of the leading vendors. Initially, Intergraph and ESRI (and others of less endurance) dominated the field. Both companies responded to market forces and developed systems with simple database structures. The databases may have been flat files but mostly of network and hierarchical structures. Advances have moved to relational and commercial databases. This has facilitated development and increased accessibility to data.
Spatial data remains separated from attribute data and there remains two distinct structures for representing spatial objects - vector or raster. Topological associations are now incorporated into the data structures but a true object paradigm has not become entrenched. This apparent reluctance to embrace the object oriented structures is intriguing given the early appearance for object oriented GIS's (e.g. System 9). The acceptance of full topological and object structures were initially constrained by hardware and secondary storage. This is no longer the case but still there is no common use of OODMS's.
There are considerable demand for improved spatial systems - where will they develop? This can partly be answered by understanding that current systems emerged from geography, geomatics and mapping disciplines. These disciplines, it is contended, are responsible, rightly or wrongly, for the current spatial modelling paradigms. Will pressures from users and business imperatives direct development into areas such as, visualisation, spatial analysis, data structures and databases or general functionality?
This paper proposes that development will occur where commerce responds to pressure but that theoretical research should occur in the areas of spatial modelling concepts. It will be important to discard the historical mapping shackles (with due acknowledgement to the valuable but constraining contribution). A (r)evolutionary spatial modelling paradigm awaits discovery.
Where will it come from - what will its concepts be founded on - how will it be revolutionary? If it is possible to visually perceive the world as it surrounds us, then it logically follows that it is possible to mentally store and recall these images. The human mind is capable of observing and understanding and even analysing space and spatial objects.
Spatial databases will evolve to store and retrieve representations of space where these representations are more accurately (closer to the truth) aligned with the phenomena we perceive. It is tempting to delve into the workings of the mind, though this is resisted here and now. What is more important to realise is that current databases are constrained by their own pedigrees. In future, space and objects within it, collectively need to be represented differently in databases. Saying `space and objects' is already pre-emptive. It is a break down of convenience; it may not be the best form. It is used here simply as a convenient way to explain an emerging concept.
It is held and developed here that spatial databases of the future will store space and objects as a whole - just as we perceive the world around us. Raster and vector will give way to space, object and knowledge. Representations will be born out of the concepts that objects exist and have knowledge embedded in them - that they exist and behave in an observable and representable way. This may seem to be much the object oriented paradigm. It is more than that - much more if we are to succeed.
Hand in glove with database developments is a concept of federated databases. This is, in simple terms, the connection of several databases in such a way that the union is transparent. Furthermore, any application, can access an index which contains information on data, data structures, data use rules, data knowledge and other applications. The federation makes this possible from any application domain to any other using any data for any reason. A grand concept that is close to reality.
As an example consider a client - a private electricity company which relies on hydro-electricity to meet peak load demands. Such a company will have to manage its resource, water at altitude and have appropriate models of supply and demand. These will be very variable in both time and location. Applications would include socio-demographics, climate and rainfall models, snow melts and load determination. It takes little further imagination to list the data sets required for these applications and that there will not be one single data source.
Future federated databases will provide an environment where execution of the models will be possible without detailed knowledge of where the data are or what structure they are stored in. It will even be possible to remotely bill your account - of course.