> From: Bernd Kreimeier [mailto:[EMAIL PROTECTED]]
>
> John Keiser writes:
> > > There was a list debate about initialization in general.
> > > I voted for "initialize on load", which is not in
> > > accordance with the specs.
> >
> > After having written the bootstrap classes, I *must* disagree.
> > They have good reasons for delaying initialization, it makes
> > it much easier to handle codependent classes.
>
> I wrote: "not in accordance with the specs". That settled
> the issue already - some Java code out there will rely
> on that, good reason or not. Now, if you were to say,
> "don't init on load, but init as soon as possible (all
> codependents loaded)", there should be disagreement - the
> consensus seemed to be that the specs delay initialization
> of classes as long as by any means possible.
>
> Apps relying on classes (e.g. accessing only finals) that
> can be loaded but not initialized are supposed to work.
> Anything but "just in time init" will break such apps.
> That seemed to be the conclusion a month ago, right?
>
>
> b.
Yes, I was just unclear what point I was responding to. I was saying
initialize on load would not be a good idea for the reason I specified
above. Lazy initialization has a lot to do with handling circular
dependencies.
--John Keiser
