Re: All hail Neo!

It is worth noting that this is a *design* issue and not a theoretical one per se. The semantics of SQL's null are well-defined, if rather clunky. Likewise, my empty-set semantics are quite well established and theoretically sound. The question is, which one leads to the most useful software? I propose that a piece of software that refuses to run calculation on data you do have because of data you don't have is less useful that one that is not so pedantic. Since this is a design issue, the only way to validate that assertion is through HCI testing, which I don't expect either side to perform.

I also observe that argument-by-naive-user is not one I consider very strong, unless the software is designed specifically for naive users. (Which is clearly no the case for the DBMS category.) Nonetheless, it is interesting that excel, which is probably the math-oriented software with the largest penetration into the broad market, works the way I propose.

Marshall Received on Thu Apr 27 2006 - 00:30:06 CEST