Re: [Gnumed-devel] GnuMed/Archive // patient object

[Karsten Hilbert]

> If we would rely on a gobal variable to figure out which patient is the 
> current one, ... how would we prevent safely inconsitency across modules? Not 
> possible, IMHO.
I would have thought that upon being changed to another
patient the curr_patient objects notifies all registered
interested parties about its change of state via the
messenger/dispatcher.

If we don't use a "global object" at all for the patient we
loose most benefits of caching commonly accessed data. That
object may life in a broker of sorts, sure, but that still
makes it global albeit with more explicit synchronisation.

> - define a variable private to the module
> - define a callback function that registers interest in change of current 
> patient with the messenger
> - this callback function then modifies that variable
> 
> However, this approach has one flaw: if the current patient changes *while* 
> the routine that makes use of the current patient obtained via variable is 
> active, it would remain blissfully ignorant of that fact unless it polls for 
> changes of the variable.
Well, of course would it do that ! That's what that variable
would be for after all ! Also note, the patient isn't
"obtained via variable" but rather "obtained WHEN variable
indicates change".

Karsten
-- 
GPG key ID E4071346 @ wwwkeys.pgp.net
E167 67FD A291 2BEA 73BD  4537 78B9 A9F9 E407 1346