Re: A new proof of the superiority of set oriented approaches: numerical/time serie linear interpolation

Brian Selzer wrote:
> "Jon Heggland" <jon.heggland_at_idi.ntnu.no> wrote in message
> news:f1c09b$8bk$1_at_orkan.itea.ntnu.no...

>> Straw man. If transition constraints were tuple-based, and assignment
>> would allow you to bypass them, then they would be bypassed. But I don't
>> think anyone is advocating that.

For instance, by not expressing transition constraints in terms of tuples. TTM does this. Or, by not letting assignment bypass tuple-based transition constraints. Dataphor does this. (---admittedly in a clunky way: they treat assignment as delete followed by insert. They also don't literally translate insert, update and delete to assignment, but I'm not sure anyone is advocating that either.)

I must point out, though, that I'm not all that keen on the idea of actually using assignment for database updates in practise (where would that leave cascaded and triggered actions?)---I consider it mainly relevant for theory (as in "an insert should have the equivalent effect of assigning to the relvar the union of its current value and the given relation"). But transition constraints isn't its biggest problem.

-- 
Jon