Maybe I was remembering the period from 1985 to 1992, when I was young (¿¡when did I start calling my 40s young!?) and naïve and thought that ANSI would want the technically best standard instread of the best profits for American companies. But then came 1986 and 1992. I remember people I worked with back then (some from universities, some from Groupe Bull, some from INRIA, and various others) agreeing that it wouldn't possible to get anything into SQL2 (what we called the standard being worked on at that time, which eventually necame sql1992) unless it had already been implemented by IBM or by Oracle. I didn't keep my illusions. Really I should have lost them years before, because I had been bitten by the updowngrade from a sensible version of Fortran to ASA (or was it already USASI? anyway, what is now called ANSI) Fortran66, which was basically Fortran IV (and was often incorrectly called Fortran IV) with the bits that hadn't been implemented by all of the big boys left out (Fortran IV was of course bad enough, but ASA can't be blamed for that, it wasn't involved in that "standard") when Fortran66 hit me in 1967.

But SQL standardisation wasn't all bad - it did have NULL in it, despite not all American RDBMS implementations of the time time having NULL. Just think what life would have been like if it had followed what RTI had implemented under the direction of Stonebraker and Wong; we would have had no NULLs, just the sort of insane default value system so loudly advocated by Date. RTI's INGRES/QUEL system at that time used default values; and the default value for all numeric types was zero, so one couldn't know whether a zero meant "zero" or "value not known". This prompted Codd to write "I consider the number zero to be far too valuable in its normal role in all kinds of business activities", which amused me greatly when first I read it.