You said “you can’t drop the index if it supports a key — you must lose the key too.”
I understand you can disable to index if you do not wish to use this index.
Anyways, if you want Oracle to be RDBMS by removing this constraints, then you go 90 to 0 in 100 scale as you have data inconsistency all around which falls into the category of non DBMS further?
There are some other views on relational model also – Like:
Uses a multi-set model rather than a set-based model: ??
Breaks the Information Rule in multiple ways (duplicate rows, duplicate columns, nulls) : Why you want to have duplicate columns in a table? nulls? I thought it addresses to RDBMS concept
Lacks true projection and relational closure: ??
Does not deliver guaranteed access: how
Lacks relational comparison: how
Lacks relational assignment: how
Lacks key derivation: wow ! how?
Does not support keys on views or derived tables: Materialized views
Uses three-value logic: what is this now?
Has poor support for data independence: I like “Poor” and pls define what is independence
I do not know how David could prove these and if he proves this then he is pointing towards very disorganized database systems for which oracle already has provisions.
My point is let’s not talk theory and be practical. What Oracle as an RDBMS has achieved has not been achieved by any other players and we cannot ride on a bus towards San Antonio if we want to go to Virginia simply because that bus is well equipped!
The mismatch of incompatible OLAP tools, associated storage models and the expense of integrating them with operational systems are a direct legacy of SQL’s failure to implement the relational model.
Another example is the pernicious paradox of denormalization. Again, SQL doesn’t implement the model therefore it fails to achieve better data independence and makes necessary some messy and expensive compromises.
A final example, what is the cost of dealing with duplicate data in SQL operations? Date and others have demonstrated convincingly how inferior SQL’s multi-set model can be from an optimization perspective.]]>
NO ONE! well suppose Oracle is the best relational or not]]>
“Most people seem to think that SQL is synonymous with relational and that SQL’s problems are therefore problems with the Relational Model.
“Because of this, some have even attempted to devise alternative ‘models’ of data. [However,] the problems are with SQL and not with the Relational Model so nobody has been able to present another actual model. Such a task is roughly like trying to find a new, ‘simpler’ way of doing arithmetic. The Relational Model is the ‘mathematics of data’ and SQL is simply a poor calculator.
“[We are suggesting that] if the Relational Model is implemented correctly and completely, then most of today’s software tools, systems, and frameworks can be drastically simplified and many eliminated.”
Therefore it cannot be relational
PS The above is not intended to be a complete list by any means
That’s not necessarily my contention — it’s something that has been written about by, for example, Fabian Pascal:
“The implementation of keys via indexes in SQL products [eg, Oracle] is a well known violation of physical data independence. A key is a logical feature to uniquely identify a tuple. An index is a physical feature to improve performance. They have nothing to do with one another, but SQL products makes keys dependent on this specific physical implementation feature. This means that you can’t drop the index if it supports a key — you must lose the key too.”
He also doesn’t like Oracle’s treatment of NULLs:
Hope that helps,
The discussions obviously do not throw much light because your observations does not clearly specify why you think Oracle right now is not an RDBMS. Let’s start from you at this point – then we can discuss further on the subject.]]>