Re: Idempotence and "Replication Insensitivity" are equivalent ?

From: Bob Badour <bbadour_at_pei.sympatico.ca>
Date: Wed, 20 Sep 2006 12:29:01 GMT
Message-ID: <huaQg.30086$9u.284234@ursa-nb00s0.nbnet.nb.ca>

William Hughes wrote:

> Chris Smith wrote:
>

>>Reordering to avoid repetition...
>>
>>William Hughes <wpihughes_at_hotmail.com> wrote:

[snip]

>>

> Here we have no restriction whatsoever on the output of an aggregate
> function. This seems much more reasonable than your very
> restrictive definition in which an aggregate function could only map
> into A. Still there is a second more fundamental issue. What could
> the binary form of a function that find the largest five elements be?

A special case of restrict that doesn't aggregate anything.

[snip]

>>Can you clarify what you mean by computed efficiently?  Perhaps I'm
>>wrong again.

>
> Just that it is practical to compute the median routinely. I did
> not have anything more formal in mind. My point is that
> the median function is clearly of practical importance,
> so defining aggregate function in such a way as to exclude
> the median needs justification.

As are all quota queries, which are special cases of restrict that don't aggregate anything. Received on Wed Sep 20 2006 - 07:29:01 CDT