Configuration reloading via Tomcat 7 parallel deployment ?

[Rod Widdowson]

> I think that's simply because Java abhors destructors.

Funnily enough I have a case (in C) where a subsystem is doing the
destructor thing for you and you have the same issue and end up needing to
implement a lot of stuff yourself which you thought you got for free.  My C#
is too rusty to know if it suffers from the same issue...
 
> I guess the question is, do we need the destroy half or not.

My understanding is that we absolutely have to for heavyweight, non java
things like JDBC and LDAP connections - not least because they consume real
resources on other peoples connections.  For the rest it's just a nicety.
 
That’s my take.  As you said previously, the two approaches are really the
same when you come to it.
 
> Did the V2 code have a destroy semantic on the components we reload?

Yes - and in particular for the attribute resolver the databases/LDAP
connections are destroyed, and there is all the required locking in place
(using a toplevel read/write lock which I'd like to demote to the component
level)

/Rod