[cap-talk] Impossibility vs infeasibility (was: Confused Deputy & Paradigm Regained homework at MIT)

[David Hopwood]

Ka-Ping Yee wrote:
> (Hmm, i wonder if the systems courses here illustrate the contrasts
> between event-loop and shared-state concurrency.  Perhaps someone ought
> to lobby for a guest lecture there, too -- Mark's illustration of the
> impossibility of a safe observer pattern in Java is straightforward and
> devastating.)

Careful about calling things impossible. If you say that X is possible in
system Y but impossible in system Z, some people will think, "Of course
it's possible on Z; simulate Y using Z", and dismiss the main point
you're trying to make. Technically they would be correct, even though
such a simulation, e.g. of event loop concurrency in terms of shared-state
concurrency in this case, would be difficult and complicated enough that
no-one would do it just to solve problem X.

"Infeasible" is a useful alternative to "impossible" in most cases like
this.

-- 
David Hopwood <david.nospam.hopwood at blueyonder.co.uk>