Re: more on delete from join

From: Joe Thurbon <usenet_at_thurbon.com>
Date: Wed, 02 Sep 2009 09:44:35 GMT
Message-ID: <op.uzmiokv5q7k8pw_at_imac.local>



On Wed, 02 Sep 2009 15:54:21 +1000, Marshall <marshall.spight_at_gmail.com> wrote:

> On Sep 1, 8:03 am, Kevin Kirkpatrick <kvnkrkpt..._at_gmail.com> wrote:
>> On Sep 1, 3:17 am, "Joe Thurbon" <use..._at_thurbon.com> wrote:
>>
>> EXACTLY!  It's the point I've been trying to make all along - view
>> updates amount to abductive reasoning, which, as Joe put so
>> eloquently, is "guessing".
>
> I see no reason to think this, and I don't see that you've supplied
> any reason to think this. View updates are just a way of
> writing updates; nothing about their nature is different than
> regular updates. Are regular updates abductive? I would
> say certainly not.

I agree that updating a base relvar is not abductive.

>
> I also disagree with the notion that the database state before
> an update is a "premise" and the state after is a "conclusion."

I thought that Kevin was saying that a base relvar is like a premise and a view is like a conclusion ...

> Any premises are table values; any conclusions are query
> results;

... which you seem to agree with, since views are 'populated' with query results (if I'm not mistaken).

So, to update a view, rather than update the conclusion directly, one must update (one of) the base relvars that are used to derive the conclusion/view. That is abductive.

Just to be clear, I'm not using abductive perjoritively. I think that the idea of doing view updates by constraint-satisfaction/equation-solving is a terrific idea. When there is a unique solution, the approach is obviously not guessing.

There has also been a lot of work done in forming orderings over possible revisions, such that some 'guesses' are considered preferable to others. Educated guessing, if you like.

[snip lots, I agree with it all]

Cheers,
Joe Received on Wed Sep 02 2009 - 04:44:35 CDT

Original text of this message