Re: Is mysql a RDBMS ?

Leandro Guimarães Faria Corsetti Dutra <lgcdutra_at_terra.com.br> wrote in message news:<pan.2003.08.30.11.17.56.505239_at_terra.com.br>...
> On Mon, 25 Aug 2003 11:57:09 -0700, Mikito Harakiri wrote:
>
> > if relational theory would
> > comprise all the math, not predicate calculus only.
>
> But why would one want that?
>
> BTW, the relational model is based on set theory (hence
> relational) and predicate *logic*. It can make use of either
> relational algebra, or relational calculus, or both.

Once again, predicate calculus and set theory are just 2 tiny subjects in math. You are not implying that the all the rest of the math is irrelevant, aren't you?  

> > For example, how would
> > you represent multivariate polynomial
> >
> > 3*x*y^2+5*x^3
> >
> > relationally?
>
> Again, why would one want that?'

I want to answer some queries about a system of multivariate polynomial equations. For example I would like to know if there is a vector that satisfies all equations.  

> > The examples of
> > (still) open research topics include view updates
>
> Nice of you to ignore the nicest view updateability model
> proposed by D&D... or even worse, about thinking the difference
> between their model and the angelic space is more important than the
> one between their model and SQL!

I'm ignoring it because view updates are as hard as equation solving, and D&D model suggests no insight into it.

> The thing is, we're stuck with SQL
> for now while there is a much better model available, and only one or
> two small-time vendors striving for it. That it is not perfect is the
> irrelevant issue here.

Once again, we disagree how much better. Bob's party may be right, the industry is stupid, but not *that* stupid. If there were obvious benefits "the big 3" would be busy implementing tutorial D language already.

> > transitive closure
>
> What's missing?

Suppose your system can answer transitive closure query. Can it also output all the paths?

Research problems like this usually have very little to do with model purity.  

> > optimization
>
> D&D and others have again and again shewn that the relational
> model is much more optimisable than SQL... we don't even need them
> telling us so, as everyone who've used SQL is familiar with its
> arbitrary restrictions, lack of logic etc that have a negative impact
> on optmisation.

Yes, you would certanly have more nice identities in the pure model. But this not necessarily mean expression transformation is the most important thing in the query optimization! For example, one of the fundamental difficulties is cardinality and selectivity estimation. Being able to handle corelated predicates correctly. Or making the cost function right. It is difficult no matter what model you use.

> > multidimensional methods
>
> I'm not sure what do you mean, but it seems this is irrelevant
> to RDBMSs and the model, being more on the application level.

Spatial database: "How many roads intersect this area".  

> > constraint databases
>
> What's the problem here? Again not sure of your meaning, but
> the relational models have a complete declarative integrity
> constraints system.

Those are generalisation of spatial. BTW, what D&D take on spatial databases? Received on Sun Aug 31 2003 - 06:01:00 CEST