Re: Is relational theory irrelevant?

"Serge Rielau" <srielau_at_ca.eye-bee-m.com> wrote in message news:bpfnq8$csd$1_at_hanover.torolab.ibm.com...
> Very good points.
> In development we call this "syntactic sugar". As a developer I'm
> responsible to implement orthogonality (i.e. supply RANK() OVER()),
> howvere, said education gap ,plus the complexity coming out of the
> optimizer strongly encourages "shorthands" for the most common
constraucts.
>
> Let me defend my position :-)
> In the latest version of DB2 we increased orthogonality by allowing
> INSERT, UPDATE and DELETE in the FROM clause. I.e. we declare these
> operators to return sets.

Indeed, and I highly applaud you guys for doing so. I've not used them in anger as yet, but from what I've seen they do look to be nicely orthogonal (or at least as nicely orthogonal as SQL lets you get to). Interestingly, such select from update is not mentioned in the relational language I know most about (i.e. Tutorial D), and is one (not insignificant) example of where lack of expressability (I might rather say *scope*) of the 'academic' languages, is also one of the problems we face. I'll note here for the record, that getting rid of logical transactions requires us to have select from update.

> One of our competitors took a different approach and allows
> "multi-table-insert" to address a specific customer problem.
> Multi-table insert is shorter to type, but less powerful.
> It will be interesting to see when customers force us to pull even with
> that competitor despite us having equivalent functionality.

Agreed. An interesting question would be to show a graph of customer satisfaction over time in such cases. I.e. the competitor might have pleased a few (probably important) customers by quickly solving specific problems, but (hopefully) we will please more customers over a longer period (our solution being more general, if later), while the competitors customers become less happy as they find other, similar situations are not covered by the quick fix.
So, then add in some factor that discounts the future (to a degree the depends on stuff like vendor and customer solvency, first mover advantage, etc, etc), add up the area under the resulting graphs and see which vendor comes out on top. Hopefully the answer will be the vendor that takes the longer term, orthogonal, view. :-)

> Cheers
> Serge
> --
> Serge Rielau
> DB2 SQL Compiler Development
> IBM Toronto Lab
>
Received on Wed Nov 19 2003 - 13:35:33 CST