Re: History question -- Tarski and Codd

In article <231020021148525703%johnfl_at_cs.uoregon.edu>, John Fiskio-Lasseter <johnfl_at_cs.uoregon.edu> wrote:
>In article <3dae743a$1_at_news.uia.ac.be>, Jan.Hidders
><hidders_at_hcoss.uia.ac.be> wrote:
>>
>> I've been wondering about that myself. There is for example an
>> interesting article by Jan Van den Bussche on the influence of Tarski's
>> work on database theory:
>>
>> http://citeseer.nj.nec.com/443292.html
>
>This was extremely helpful, although there are parts of it that for I'm
>stuck in stupidity. "Cylindrification along a dimension i", for
>example, makes no sense to me, nor is the equivalence between the RA
>presented by Van den Bussche and that actually given by Tarski in the
>JSL paper completely obvious (that one I can probably work out myself,
>though).

Yes, cylindrifaction is a bit mysterious because it seems such a destructive operator. Basically what it does is fill up one column with all the values in your domain. So if you have a relation:

  { ( a, b, c ),
    ( b, c, a ) }

and you cylindrify it on column 2 then you get

  { ( a, a, c ),
    ( a, b, c ),
    ( a, c, c ),
    ( b, a, a ),
    ( b, b, a ),
    ( b, c, a ) }

I don't exactly remember the simulation of the relational algebra. The union is given. The intersection is given by A INTERSECT B = COMPL ( COMPL(A) UNION COMPL(B) ). The difference is then given by A MINUS B = A INTERSECT COMPL(B). The selection is simulated by intersecting with the corresponding diagonal relation. The projection and cartesion product have slipped my mind at the moment.

>On a side note, JAN HIDDERS: I've tried to send an email "thank you"
>for the additional message you sent, relaying Van den Bussche's
>thoughts (and the translation from Flemish for this monolingually
>challenged American :-) It's bounced twice, though. Anyway, thanks!

Yes, thanks, and sorry for not configuring my trn properly. :-) I hope that's better now.

• Jan Hidders