What is best suited for GIS and data warehousing?
Both RDB and OODB are useful technologies
GIS Industry standards are all driving toward object-oriented designs (SQL3, OGC)
Middleware will change nearly everything!
Notes:
It is not clear that any one approach best serves all applications.
Each has distinct advantages and disadvantages, and both relational and OO approaches are evolving rapidly now. It is important to recognize and understand the trade-offs in each approach, and fit those trade-offs to your application requirements.
The greatest immediate GIS issue in 1998 is not the OO v.R issue, but the provision of secure, distributed GIS data repositories and WWW access capabilities. As it happens, these are both areas where the OO approach has very significant potential advantages, but we will have to see whether the OO GIS vendors can turn this potential into practical products faster than much bigger companies who are throwing money and resources at staying ahead despite their innate disadvantages.
ISO TC211 and OGC are developing standards, but we know that a 'full set' to cover a commercially sensible amount of a product such as Gothic cannot be done within 10 years. Also, this 'full set' covers several quite different levels of abstraction and overlaps significantly with non-spatial data warehousing and with generic WWW proposals, e.g. unique identity, metadata, catalogues etc.
Thus we can confidently expect that the entire software landscape will be radically different before and if the full OGC 'componentisation' architecture is available. One possibility is 'Java everywhere' coupled with generic middleware and replication which could entirely hide the current distinction between fat and thin servers.