One example I often use is launch window computation for rockets. One way to compute them is brute force simulation: you simply try all launch times and for each one you perform the simulation of the early orbits phase. If at any time something breaks (a polynomial that never crosses zero, an optimizer that fails to converge, a square root from a negative number ...), you simply know you have exceeded the launch window boundary and you try another date to identify the limit. The computations are far too complex to be able to say before hand using preconditions: this launch time will not work (it is an example of the Turing halting problem). Also the fact a computation totally fails in this context is not an indication something is wrong in what the user asked for, the user was especially looking for such failures! This example is rather specific, I admit it, but the point is that at low level, you cannot decide this is a problem that should stop the application so I can trigger a runtime exception and this other case is a problem the user can handle and I should trigger a regular exception. >> To be hopefully clear (referring to your remark about >> "MathIllegalArgumentException"): Wrong arguments are the cause of the >> failure and it is the role of the _caller_ to prevent this from happening. > It's not always possible and it was proved by Turing in 1936. It is an > undecidable problem.
Great example of RE vs recursive above :) Phil --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
