More on (not "moron") my first 'baby steps' into the Toolbox

I've gotten as far as a program that reads through all the records in a
file (using VIEWCUS, the venerable demonstration file from our QuestView
product) after getting its record format dynamically.

Then I tried making my test class a "FileListener," and having it listen
for FileEvents. It correctly caught the even from its own open() operation
on the file, but failed to notice when another job opened and modified the
file. Is it supposed to happen that way??

The object here is (in connection with our Business Integration project)
to eventually monitor an Event File, and it seems like if such monitoring
could be driven by the arrival of new records in the event file (rather
than by polling the event file every second or so, or at least in addition
to periodic polling), it would operate more efficiently.

And speaking of that project, I'm looking for advice on the question of
"big-event-model" vs. "small-event-model," i.e., on whether initial event
records should actually contain record images, or just record keys. The
former would presumably entail huge records (not especially good), while
the latter might create this scenario:

                 Suppose we are interested in an event involving changes
to a field. Postulate that we have two events coming in,
one of which (we'll call it Event 23) represents a change
of that field from FOO to BAR, with a later event (we'll
call it 42) involving a change from BAR to BAZ. (For now,
we'll disregard QUX, BAT, and XYZZY.) Now suppose that
the business integration engine has been flooded with
events, and it doesn't get to Event 23 until after Event
42 has already occurred. With "big-event-model," the
business integration engine correctly processes both
events, but with "small-event-model," it appears to me,
at least, that it ends up processing Event 42 twice.

Am I straining at gnats here?

--
JHHL