Relational database failing internet: Simon Williams Lazy Software. (Database & Network News and Products).

`Web-enabled applications are so different from those of its predecessors - host based or client/server - that most applications will need to be entirely redeveloped As long as we continue to rely on the Relational Model of Data, application software will be far more costly than it needs to be and the goal of a wholly web-enabled IT environment will remain distant. The logical fulfilment of the web enabled world is the ability to access databases via the web, extract the relevant data onto a variety of personal computing devices, work disconnected for as long as we like, and link up again via the web whenever and wherever it suits us, to re-synch and refresh our personal databases.

To deliver this, a database architecture needs four things:,

* A small footprint, so the facilities we have on our personal devices are not vastly inferior to those of the parent database. The established relational databases are many years old, and have bloated to the point where the goal of a small footprint is unattainable.

* A good conceptual model for distributed data so we don't need a DBA on hand to help us extract the data we want to carry with us. The relational model is a very poor model for distributed data - with the Internet epitomising decentralisation, why else would Oracle advise its customers to base their web strategies around a central database server?

* A simple, dynamic schema, so we can add features to our mobile databases as we go, without needing a systems analyst to design them, or a programmer to integrate them.

* A common language platform across mobile and permanently connected mobile devices, so we don't need to understand two different technologies. Whatever languages this is based on - Java looks favourite today - they will not be C, C++ or SQL.

The fundamental problem, lies in the design of the relational database which requires a new set of programs to be developed from scratch, because a program written to use one set of tables can't be reused with a different set. OK one might say, the market leaders need only rewrite their flagship products in internet-friendly languages. But this will not happen. Of the dominant players, only Oracle, IBM and Microsoft have the resources to attempt such an undertaking and probably only Oracle would have the focus and the will to attempt it. But then they would be confronted by one of the key economic realities of the software business: it is never economically viable to re-engineer an established software product unless the process adds major new features that benefit customers and enhance competitive edge.

Rewriting say Oracle around XML using Java or C# would not qualify. It is also worth observing that SQL and XML are fundamentally incompatible. If, as seems likely, XML becomes the de facto standard for Internet data exchange, its incompatibility with SQL will be a thorn in the side of application developers for as long as the relational model persists.

www.lazysoft.com