Re: The Fact of relational algebra (was Re: Clean Object Class Design -- What is it?)

In article <9nh0g5$f46$07$1_at_news.t-online.com>, Carl Rosenberger says...
>
>Somehow all the large object database companies seem to have lost the fun at
>the development of new concepts some years ago and they drifted off to
>develop other products (like web server suites) instead of perfectioning
>object database technology. Why, I wonder?
>
In the mean time, major relational vendors didn't stay still. User defined statistics, user programmed indexes, query rewrite based upon materialised views -- just to name the few latest developments. In short, Relational databases today are extremely powerful -- I wouldn't agree with Bob calling them SQL-databases.

Of course, there are [still huge] drawbacks. The fundamantal one is that schema design is hard and most often results in the models unnatural for object proponents.

The storage model is awful. They invented 5 levels: blocks, files, extents, segments and tablespaces. Only blocks and files have physical OS counterparts, of course. When they describe implementation details I want to scream. Users have to pay big bucks to DBAs, which have to deal with all that mess.

The process model -- virtual OS -- is not pretty either. Nothing compared to Java Virtual machine.

Finally, they entered 21st century with a limit on string size!

Overall, however, my prescription is different: 1. Add three java types -- int, java.lang.String, java.util.Date to the primary SQL types and deprecate redundant SQL types. 2. Add ability to create new SQL domains as java datatypes. 3. Clean implementation. Received on Mon Sep 10 2001 - 05:30:14 CEST