Hybrid Systems Approach

It seems like the hybrid systems approach described in the paper Ras

gave me would work.

Slogan: we are modeling or simulating a continuous system on a digital

computer.

The constructs in the fran.rkt file seem mostly OK. "while" asserts a

set of constraints that hold continuously. (Note that they hold as

long as the condition is true -- we might transition to a new state

and then back into the while.)

Probably: we can have multiple processes (frames of reference). One

important relation is that of an observer process O that references

but does not change a model process M. (This ignores Heisenberg and

such; we could perhaps add that later. Certainly eventually if we

have multiple processes, they should be able to mutually interact.)

Or we could have just one process, with lots of little interrupts to

show frames of the animation. This seems conceptually messier.

Handling animation: have a separate observer process.

Handling input events: have a separate 'system' process that we can

observe. It seems like we can mostly ignore the separate process part

by having local variables constrained to be equal to system ones.

(What about lags, missed observations, etc? Or should we ignore those

issues for now?)

Does every process have a clock? Do the clocks run at an arbitrary

rate unless related to the system clock? Or ...?

To reference old state:

(prev expr) is the same as (evaluate expr)

Implementation

system-frame-of-reference -- have something that updates the clock and

mouse position periodically (when??)

event detection: one approach would be to propose a time step. Let t0

be the current time and t1 be the proposed next time. Then ask

Rosette to find a time t such that t0<=t<=t1 and such that the event

occurs in that interval. If there is such an event, then need to

halve the time step. Otherwise can advance time.