At 04:41 PM 10/24/01 +0100, Ivan Prenosil wrote:

>> Netfrastructure, incidentally, while supporting (and using, heavily)
>> primary and foreign keys, does not enforce referential integrity
>> for a couple of reasons. First, good programs (and Netfrastructure
>> programs are good or I won't let them at it) check that stuff as
>> a matter of good person hygiene, so having the system repeat the
>> check is a waste of time (though admittedly cache hits).
>
>How can good programs be responsible for reliable checking of referential
>integrity (without possibility to see over transaction boundaries) ?
>Or does Netfrastructure support dirty reads ? :-)
>

First, my typist meant to "programmers" not "programs". It is
hard to get good help these days.

Second, referential integrity can be enforced within the
transaction rules in a multi-generational system. Enforcement
of unique keys (which Netfrastructure, like Firebird, does
enforce), cannot, requiring special engine handling.

Third, the concept of "dirty reads" does not belong in any
discussion of database semantics, and never, ever, in the
same sentence with the word "integrity" without the word
"not". Among the things inconsistent with the concept of
"dirty reads" are:

1. Data integrity
2. Referential integrity
3. Intellectual integrity

Jim Starkey