Re: WWW/Internet 2009: 2nd CFP until 21 September x

Sorry for the delay in responding. It's been a busy week.

"Walter Mitty" <wamitty_at_verizon.net> wrote in message news:CAKim.2731$nh2.2036_at_nwrddc02.gnilink.net...
>
> "Mr. Scott" <do_not_reply_at_noone.com> wrote in message
> news:i5WdnRK1WL_HaRfXnZ2dnUVZ_qudnZ2d_at_giganews.com...
>>
>>> solution to be a materialized outer join gives a way of conceptualizing
>>> the two tables, where no NULLS are necessary or permitted. It also cuts
>>> back on the proliferation of tables. I'm all for decomposing tables
>>> when it will really do some good, but there's no need to make a
>>> religious ritual of it.
>>
>> I'm now wondering if there are really three kinds of null. There is the
>> applicable null, which when submitted in an update indicates 'I know
>> there is supposed to be a value here but I don't know what it is,' there
>> is the inapplicable null, which when submitted in an update indicates 'I
>> know there is not supposed to be a value here,' and now there is the
>> I-don't-have-a-clue null, which when submitted in an update indicates 'I
>> don't know if there is supposed to be a value here.' In order to
>> simulate the I-don't-have-a-clue null that can be in your Employee table,
>> your Employee_Middle_Initial table has to have the open world
>> interpretation, so that whenever a row is missing it indicates that
>> either there should be a value here but it is at present unknown or that
>> there shouldn't be a value here.
>>
>
> The difference between these three kinds of nulls (and the other seventeen
> kinds of nulls) is not in what the null says about the real world. It''s
> all about the state of mind of the writer.
> Reread what you've written here... it's all about the narrator knows or
> doesn't know or doesn't have a clue about.

I think it's about what should be considered information and what shouldn't. The closed world assumption effectively denies every valid atomic formula that can't be derived from what has explicitly been stated, so it would be redundant to assert negative information. An inapplicable null states that there is not supposed to be a value in a specific role. An I-don't-have-a-clue null essentially states that it doesn't matter whether there is a value in a specific role. In neither instance does the fact being asserted depend on there being a value in that specific role, but where there is an applicable null, the fact being asserted does depend on there being a value, even if it isn't clear which one. This is why I think that only applicable nulls convey information that can't otherwise be derived from what has explicitly been stated.

> So what does a null say about the real world? Nothing.

An applicable null indicates that there is supposed to be a value in a specific role drawn from a particular domain. That's quite a bit more than nothing.

> I'm gl;ad you brought up the OWA/CWA question with regard to the middle
> initial table. The question about what a missing row means in the Middle
> Initial table is precisely the same question as to what a missing value
> means in the middle initial column in the plain old one table design. And
> you could come up with a OWA convention or a CWA convention with regard to
> this table (or any table), but that is a convention about the semantics of
> the data. More precisely, it's a convention about the semantics of the
> data that isn't there.

The point is that the semantics of the data is unclear when there are I-don't-have-a-clue nulls or when a table has the open world interpretation. It is not clear whether there is supposed to be a value.

> That's beyond the scope of the point I was trying to make. My point was
> just that if you think of a table containing nulls as a materialized outer
> join, it can help you to avoid getting all balled up in SQL's three
> valued logic and like that.
Received on Mon Aug 24 2009 - 07:58:24 CEST