Richard S. Westmoreland postd this to the Forth-ML.

 In years past, we have implemented some extremely complicated databases in
 Forth.  The first was done in the mid 1970's for the company Cybek in NJ,
 and it was used in some extremely complex applications.  FORTH, Inc. also
 did some very complex databases for other companies, one of which was still
 in use last time I checked, at www.calmuni.com.  That one was a
 2-dimensional database, with a huge bit matrix in the center used to
 calculate overlapping bonded indebtedness.  A few years ago my contact  there
 told me that a state agency had just spent several million $$ trying to
 replicate it using modern database tools, but the result was too large and
 too slow to be usable.

 In the late 1980's we added class-based techniques to it, which many people
 liked (although I personally preferred the earlier, simpler version).

 It's hard to describe the whole approach in a newsgroup post, though.  It
 certainly didn't resemble SQL!