Re: Does Codd's view of a relational database differ from that ofDate&Darwin?[M.Gittens]

Jon Heggland wrote:
>
> Would that it were that simple. The same could be said about the "XML
> data model", yet well-known researchers are happily reinventing the
> square wheels of IMS and similar systems, apparently blissfully unaware
> of the history and fate of hierarchical databases. I recently reviewed a
> paper for VLDB which was quite scary in that regard---especially since
> another reviewer, a really *big* name, said it was the strongest he had
> reviewed.

That's quite possible, but since I don't know the details I couldn't possibly comment. On the other hand, I know from personal experience that even some of the big names that are right at the core of the hype, are very clear about what they think XML is useful for what it is not useful for, and who would judge any claims that it would somehow replace the relational model as, and I quote, ridiculous.

In that respect I really like the work by Torsten Grust et al that shows that you can do XML on top of an RDBMS, and that even though you are pushing the square XML peg in the round RM hole, you will still get something that works in many respects better than XML-native solutions. So even if you are basically horribly abusing the RDBMS and shoving all that vile XML stuff down her throat, which it was not really built for in the first place, she will still happily and with a smile on her face provide you the scalability that we have become so accustomed to. Almost brings tears to my eyes. *snif* :-)

> I might agree that the model presented as a "pure object-oriented
> database model" in the paper you mentioned is slightly different from
> the (most common interpretations of the) network data model. (Note,
> however, that the paper talks of modelS, not THE pure OODB model.)

Oh, yes, and I never denied that there is more than one. But they are essentially the same, and even so much the same that you can do theory on them that will apply to all of them.

> The
> question is whether such models are different *enough* from the network
> model(s) to make it worthwhile to distinguish between them. The concept
> exists, but that in itself does not mean very much.

I don't completely agree that this is the right question. What has changed is not so much the data model, but the whole cloud of knowledge and silent assumptions that surrounds it. For the network model it was usually assumed that you wouldn't need query optimization. That data-independence was not possible or just a nice-to-have. In the beginning some also thought that about OODBs. That, in my experience, has changed.

> (Another point is the question of what "OO" in that context really
> means---the paper does not mention encapsulation, polymorphism or
> inheritance. What is an object?)

Well, the main reason for that is that these aspects are not relevant and somewhat orthogonal to the problem that it studies. I think I have a pretty good idea of their view on that, though.

• Jan Hidders