Re: What does this NULL mean?

Frank Hamersley wrote:
> mountain man wrote:
> [..]
> > Theorists like Date who think the null should be rationalised
> > out of existence have no demonstrated understanding of the
> > database change management environment, and specifically
> > the management of schema evolution and its consequences
> > in regard to the generation of nulls.
>
> Just picking up on a minor point that has been front of mind for some
> while - who in this place considers Date to be a credible theorist and
> especially to be accorded stature commensurate with a luminary?

If I don't stretch it too far, I can make the analogy between Date and Roger Ebert. I like to read Ebert's opinion about a film because I can almost always decide if I want to go see a film based on his reviews. There are many times when my opinion of the film would be different, but that is a different issue altogether. If I propose a film and someone says to me "but Ebert only gave it two stars" I do not take that as highly relevant for whether I will like the film.

Similarly, if I read Date, it is often clear to me where it is I disagree. If someone says "Date says ..." that does not mean I will agree with it (you might have noticed) but if I read what he wrote, I can typically figure out where my disagreements come in. An exception would be when he starts out by defining a Relation the way he does in his Intro to Database Systems, 8e book. It is so hard to find and follow the def that it was hard to care about anything based on it.

>
> Personally my introduction to material produced by him was as an
> undergrad when Intro to DB Systems (2nd Edition) was the course text.
> It, I presume, is still considered a significant book but having picked
> it up recently I came to the conclusion it is basically a catalogue of
> preexisting material.

It is a textbook, afterall, so if he could adequately capture the existing material there, that would be a good start, right?

> Right place, right time if you like, in terms of
> capturing Codd's Damascus event and a good presentation. However the
> genesis of his career was as an instructor not an architect or even an
> engineer.
>
> Having read some of the free to air material attributed to him and his
> freely associated with colleagues I have my doubts that he should be
> accorded quite as much stature as perhaps is common today.

I don't have problems granting him stature as one who influences the landscape, even if his works are not all things to all people.

> To
> illustrate is quote from page 9 of the "Missing Info without Nulls"
> lecture notes presented by HD ...
>
> "And we have reduced the salary part of the database to the simplest
> possible terms.

One of my trigger words for Date's work (as well as others discussing the RM) is "simplest" -- that is a term about which there could certainly be a difference of opinion, even when working with mathematical theory. What is simpler -- a square or 1 as the identity in multiplication?

> Yes, some of the complicated queries get more difficult
> now, because we might have to combine these tables back together again,
> but the simple queries, such as 'How much salary does each person (who
> has a known salary) earn?' and 'Who earns no salary?' become trivial."
>
> So trivial remains trivial and complicated becomes near impossible.
> This type of progress I think conveys your point comprehensively Pete!

I agree.

>
> > People who construct database systems but do not maintain
> > them, or have not maintained them for more than a few years,
> > will never understand that the greatest entry point of nulls into
> > the database is during change.

But do not confuse the introduction of NULLS into the data with the need to use a 3VL.

>
> Ne'er a truer word wrote! Personally I try to manage backfilling (to
> simplify new queries dealing with old data) where possible without
> polluting history, but of course that is not always going to be possible.
>
> > Change management or schema evolution is not adequately
> > addressed by Date et al, however by the time they set forth
> > the processes covered under this subject, it will become very
> > apparent that the NULL will never be rationalised away, and
> > it is better to therefore appropriately manage its identification,
> > its existence and its resolution interactively.

I agree that Date does not address change management adequately. He doesn't seem to "feel it in his bones" as those who have lived there do. There will always be attributes where a lack of a value is a valid "value" -- the question is whether to treat it as a value or not within the logic system. Treating it as a value, with valid operations where useful, has many advantages over treating it as a value to the internal system that reflects itself as a lack of a value to the logic system, which needs to resolve to values back to the user.

> Not adequately, or not at all? I haven't yet (and given my views above
> am unlikely to ever) shell out for their dissertations on Temporal
> affairs, but I wonder how they might approach the issue when the schema
> itself has temporal variability. I suspect to apply this to SQL-D if it
> ever materialises would lead to an exponential number of relvars with
> untold amounts of information buried in the metadata. Of course for an
> inaugural novelty (tautology intended) I could be wrong :-).

Perhaps we can let SQL fade off into the sunset before then (while, of course, having to maintain and incredible amount of SQL code). Cheers!  --dawn

>
> Cheers, Frank.
Received on Thu Dec 15 2005 - 18:07:35 CET