Re: Does Codd's view of a relational database differ from that ofDate&Darwin?[M.Gittens]

Jon Heggland schrieb:
> Perhaps I don't understand the problem the UR is meant to solve.
>
> Is it that the query with the joins is difficult/complicated to
> formulate? That is mostly a matter of syntax/UI. The good thing about
> the join query is that it explicitly states how the relationship between
> employees and managers is inferred---this has to be stated
> somewhere/sometime, either by the querier or by the database designer.

Technically the problem is that some joins are moved from queries to the database engine. Instead of transient joins with the scope of one query we make them persistent with the scope of the whole database. The queries are simpler and the joins can be executed more efficiently.

However, the main advantage is that our database now knows more about data and its relationships so now it manages these relationships (joins) rather that rows in tables.

When we follow this way we obviously loose some freedom of manipulating data but the potential advantage is that data management is more reliable, more consistent and more efficient.

Actually, the same high level goal is formulated for MS WinFS where they want all data items to be related and the system knows about these relationships and can retrieve correct data given simple queries. The main problem is that in this case we need to dramatically simplify our view of data and item relationships.

-- 
http://conceptoriented.com