Re: The Practical Benefits of the Relational Model

From: Peter Koch Larsen <pkl_at_mailme.dk>
Date: Tue, 1 Oct 2002 11:42:52 +0200
Message-ID: <3d996e20$0$52166$edfadb0f_at_dspool01.news.tele.dk>


"John Jacob" <jingleheimerschmitt_at_hotmail.com> skrev i en meddelelse news:72f08f6c.0209301959.41f55690_at_posting.google.com...
> We welcome criticism of the product and of the theory on which it is
> based as long as that criticism is well founded. This was the reason
> for my retort to Mikito. We actively search for solutions.

Well, this answer is far more precise than Nathans!

> I firmly believe we must maintain some degree of
> precision in these discussions. Sorry if I used D4 syntax, but I feel
> it is fairly clear, perhaps even more so than SQL because it
> explicitly outlines the operations taking place.

I will accept that answer. For my part, I do believe that I understand D4 syntax.

> While I must admit that the statement does sound like a bullet from a
> data sheet, it is in no way a vacuous remark. Perhaps we could
> rephrase it thus: all the operators of the relational algebra have
> well-defined semantics regarding updatability. These semantics work in
> concert with each other to effect updatability no matter what
> operators are involved in a given expression, just as the retrieval
> operations have a specific affect, so the update operations have a
> specific effect, meaning that there is a kind of updatability closure
> over the operators of the relational algebra. Therefore, for any
> combination of relational operators used to define a given view,
> including ones involving scalar expressions such as extend, we can
> determine the effect of an update against that view. One of those
> effects is obviously an error.

While I have no doubt that Alphora has a much better view-update mechanism than SQL, I do fail to see any difference from SQL in your above claim. SQL supports view updates to, how ad hoc and bad these choices may be. What is needed is a more formal specification describing - in more detail - the view update policy used in D4. This would then be the base for arguing if the update is ad hoc and complete. I do not believe that this forum is appropriate for such a discussion though. What we might agree on then probably is that any view update mechanism involving set union or set difference would be ad hoc in the sense that alternative mechanisms could both succeed, but choosing among the different options must be based upon intuition and what feels right in the given situation.

Kind regards
Peter Koch Larsen Received on Tue Oct 01 2002 - 11:42:52 CEST

Original text of this message