Re: Idempotence and "Replication Insensitivity" are equivalent ?

From: Chris Smith <>
Date: Sat, 23 Sep 2006 07:44:54 -0600
Message-ID: <>

<> wrote:
> For some reason some of you guys want to stick to a definition that is
> proving to be too strict, not even coerent since treat AVG and MEDIAN
> are treated differently (I have shown that conceptually they are both
> systesis of a distribution). Not practical.


Please reflect on what you're doing. You are coming into a newsgroup on database theory (and then adding one on math; and yes, you added it); and then telling us that you don't know anything about database theory; and subsequently complaining that the standard formulation of aggregate functions is wrong. It's not wrong. It is widely known and accepted that the median calculation fits poorly, if at all, into the category of aggregate functions. You insist on acting like this result is an invention of the people you are speakig to on this newsgroup; in fact, it is probably older than I am.

If you think that query languages should have a median function, and that it should look just like an aggregate function, then fine. You can choose a DBMS that does so, or even write a DBMS that does so if you are so inclined. It remains true that your syntactic similarity has nothing to do with aggregate functions, and that your implementation of the function will look extremely different from a reasonable implementation of any aggregate function.

> Nor useful to the purpose of implementations.

It most certainly is. Your assertion to the contrary is not very convincing without evidence.

> There is no need to stick with old theory if a new
> one proves to be better ...

That's a little presumptuous for someone who couldn't be bothered to learn basic relational algebra operators until a couple days ago.

Chris Smith
Received on Sat Sep 23 2006 - 15:44:54 CEST

Original text of this message