Re: Why all the max length constraints?

Dawn wrote :

> If I know it, I don't know that I know it.

Eventually, the penny will drop. The question is, will you wait for gravity to exert its pull, or do a little extra reading up to help it on its way ?

> It is possible, but until I know I know it, I don't know it.

A "point of seeing" is a process.

> I don't know how you took it, but I know that I do not have the same
> background as most of the cdp readers and writers, given that I do not
> have a degree in CS and have never been a DBA.

The one good thing about formal computing science is learning to look at the theory first, then consider various implementations separately. Once in this habit or world view, it becomes not quite instinctual, but reasonably straightforward. Without this habit, however acquired, it becomes difficult to maintain a proper separation between issues with theory and issues with implementations. Sadly, this seems to have become expendable in degree curricula over the last few years.

> It might be the case that whatever framework (combination of terms, experience,
> and perspective) I am using that leads me to such a question is not one
> that is common nor would it be easy for someone without a similar
> framework to understand.

Or perhaps the framework is broken, and needs repair ? It's by absolutely no means impossible, but it's a bit more work without a formal background. (God alone knows how, but I managed to miss databases completely in my degree, so I have to pedal quite a bit harder to keep up sometimes than I really should have to.)

The relevance of Euler and Algol-68 was to point out the gap between what was designed or proposed, and what eventually gets implemented - although, especially in the case of Algol-68, it was a formal design that more than a few implementations simply didn't follow. You have to remember that Codd explicitly didn't say anything about physical representation, and very little about types (apart from an unfortunate use of the word "atomic"), so the relational model was pretty much at the mercy of whoever implemented it - or as happened, on some crucial issues grabbed the wrong end of the stick and shook it hard. It appears that both IBM & Ingres chose a disk image method of representing tables. (There may also have been a touch of the Fortran syndrome - "we didn't know we were designing a language - we were just writing a compiler to prove you could".)

The sad part is that no-one seems to have gone much beyond the disk image method of representing tables (unless you count TRM, which is a subject kicked to death on another thread recently).

• Tony