Haven't tried with one of those fancy low-volume models, but
regular toilets can be operated without use of the UI: a sufficient
volume of liquid in the bowl is enough to run the flush cycle.
Unfortunately, if the server in Singapore (a nation whose censors
have been known to leave their population in suspense by not
airing portions of multiple-episode Star Treks) has caused one's
entire water supply to go walkabout, then one may very literally
be SOL. (Even a very dissipated evening at Sid's is unlikely to
provide sufficient volume for operation, but YMMV)
A less scatalogical example of UI being syntactic sugar can be
found with telephones. If someone has inconveniently locked up
the keypad or, worse yet, failed to provide one, it may still be
possible to place a call by pulse dialing on the switch-hook. (One
will appreciate the amount of variance tolerated by the switching
system when dialing numbers with a 909 prefix!)
If one wishes to reduce variance, such as when operating a kanban
process, one can see where six-sigma figures come from: it takes
extreme reliability to confidently run a brittle system. (there's a
management consultancy: anyone care to run a seminar series on
ductile manufacturing?)
Related processes: traditional call-return subroutine linkage (with
immediate applications to unsuitability for WAN/mobile/etc. apps);
investment strategies (given that one is betting on the right side of a
biased coin, it still takes a high degree of predictability before one
should commit any large fraction of capital to each throw - even if
one follows the advice given by the fellow who said "If I wish to gamble,
I'll buy a casino", it's best to avoid high-rolling baccarat players);
multi-level marketing (or, to the more cynical, just about any chasm
crossing activity via branding - though in these cases, the payouts are
tempered by the double-or-nothing nature of the game)
Speaking of very simple models, are there any theories for optimal
Gini coefficients? Eyeballing the US Census curves for income/wealth
gives a very Boltzmannich distribution, which leads one to believe that
a simple random-interactions model might suffice.
-Dave