[libvirt] [RFC] making (newly public) EventImpl interface moreconsistent

[David Lively]

On Fri, 2008-11-14 at 14:11 +0000, Daniel P. Berrange wrote:
> On Fri, Nov 14, 2008 at 07:36:27AM -0500, Ben Guthro wrote:
> > I'll answer for Dave, while I'm looking at this.
> > 
> > As far as I know, Dave is of the opinion that we are just "getting lucky" 
> > using the APIs as we are, and remains convinced that his suggested change 
> > is necessary here.
> > 
> > He (and I) remain worried that release of the EventImpl API without this 
> > API change could get us into trouble in the future, as we would have to 
> > support the released API that has different semantics than DBus, which 
> > we were supposed to be modeled closely to.

Yep.

> Basically, there is no downside to implementing your suggestion of allowing
> the same FD to be registered, and a clear potential downside to our current
> impl. So I'll re-write the Add/RemoveHandle API as you suggested to eliminate
> the risk

Thank you!  If you'd rather, I'd be happy to make those changes (today)
and submit a patch.  (But I haven't implemented them yet, other than
changing the decls and documentation.)

Dave