Re: database integrity

x wrote:
> Suppose at the core of your business is a huge database maintained by
your
> organisation and each false piece of data from the database can cost
you
> millions of dollars if you rely on it.
> What can you do if you are not 100% sure of the data ?
> The database become useless because you need to manually verify each
piece
> of data you retrieve.

Although this is an absolutely critical issue, it has nothing to do with relational integrity. Conflating this definition of "integrity" with that used in relational models doesn't buy you anything, though of course both are important to an organization.

> Now suppose that each false piece of data cost you 0.01 dollars and
you have
> only a few hundreds of those pieces.
> Then you risk a few dollars. In this case maybe you can talk about
levels.
>
> Integrity is not the only issue. The database may contain false data
(by
> aging for example).
> Integrity measures are the only way for a DBMS to protect the data
from
> coruption.

A DBMS can't protect the data unless it knows what constraints to enforce. Falsehood isn't enforceable in general, although specific variants might be - for example, aging could be, although it raises the issue of whether the fact that something was ONCE true, but no longer is, matters - and when it became invalid, of course.

To me, aging falls into the category of integrity constraints over the system catalog, which is implemented... well, nowhere I know of.

• Eric