On Thu, Oct 25, 2012 at 2:42 AM, Steve Litt <slitt@xxxxxxxxxxxxxxxxxxx> wrote: > Also, with the organization they're using, one can make new "columns" > on the fly. ... Anyway, the keypuncher is punching > data, comes across a brand new type of data (let's say "artist"), so > for this row the keypuncher puts in a key-value pair of "artist=Lady > Gaga". From a practical point of view, data structure could be change > at key entry time, and needn't have been anticipated by the programmer > nor recompiled or reorganized when a new type of data element entered > the requirements. That's wonderfully flexible, but it forfeits the protection that a well-designed schema gives. A system like that is likely to end up with different records storing the same data under slightly different names, and you'll have a massive proliferation of "columns" that have only a single row's value in them. That's fine if that's what you want, but from a data entry standpoint, I think it's _too_ flexible for most purposes. ChrisA -- Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx) To make changes to your subscription: http://www.postgresql.org/mailpref/pgsql-general
