Re: TclOO: Renaming "new" and "create" methods?



tom.rmadilo wrote:
The interesting thing here is that the DBMS could easily determine
that the view is not updateable and if an object system was built,
update operations would not be produced.

There's an interesting grey area: view update operations in which the
affected row set and the effect that the update would have upon it
could in theory be determined, but the techniques that a particular
database manager uses are insufficiently powerful to determine them.
I'm reasonably certain that the general question -- in actual SQL,
not Codd's relational calculus -- of determining
all effects of a view update is undecidable, or at the very least
computationally intractable. So the SHOULD in the statement,
"all views SHOULD be updatable" is necessarily a weak constraint.

For that reason, beyond a fairly limited set of constraints, the
standard treats the question of what view update operations are
permitted as a "quality of implementation" issue.

Codd's relational calculus is much more tightly constructed
than SQL. It may well be possible either to update all views
that it can present, or to determine (with reasonable effort) unambiguously whether a view expressed in it may be updated.
SQL, for weal or woe, has departed from Codd's vision in a number
of significant ways and is considerably thornier to deal with.

--
73 de ke9tv/2, Kevin
.