Re: Idempotence and "Replication Insensitivity" are equivalent ?

Chris Smith ha scritto:

> The only remaining open question that I'm aware of is how to
> characterize the set of binary functions that yield well-defined
> aggregate functions under the second definition. I don't know the
> answer to that.
>

What so far I am not really clear is why should we restricting so much the aggregate function set, while makes a lot sense to consider things like median or CountDistinct or Range, etc

And, actually, the problem it is not the restriction, but its very nature.

In fact we are not even guaranteed with respect to computational complexity, as probably it is possible to provide functions with any cc which satisfy those restriction, including factorial. Would I call that "well-defined" ?

Hmmm. I know nothing about dbms theory, except than practical stuff, but if this has been proposed by someone, perhaps has been done as a theorical model, just to study perhaps its formal properties.

I am missing to see how all that could, for instance, change they way I write my programs for aggregative functions.

I am certainly missing some important point ...

-P

> Chris Smith
Received on Thu Sep 21 2006 - 12:49:05 CDT