Where should the type information be 341

...

Ah! I was peripherally involved in that POSC effort too, and a good friend of mine was in the thick of things. Quite interesting stuff, and good people were involved.

As you mention later that was more about online (RDBMS) than offline (tape logs). Also, the tape logs whose format had become a bit of mystery were a bit older than POSC.

Well, that happens to any data model, not just the relational one. The relational one at least has the advantage that if properly understood handles ''complicated'' data structure more safely than others.

Ah yes, I remember that the POSC schemas were a bit too detailed; it was also, IIRC the outcome of two alternative schema proposals being shotgun-merged-by-committe :-).

But large, excessively detailed schemas are not uncommon: in another consulting buttignment in the same timeframe I found around 300 tables in the schema for a fashion garments company. Usually this is due to overzealous DBAs more than normalization requirements. Conversely my impression is that when when no DBA is involved, schemas end up as 2-3 tables with a few dozens or even hundreds of columns each :-).

My impression is that data architecture and design is an erroneously underestimated subject, and that as a rule it is rather more important than program architecture and design.

But then I use relational analysis to understand in-program data structures too, and that helps me a lot. Unfortunately such approaches are not buzzword-compliant, and run against an overwhelming deskilling trend.

...