RE: load-balancing [RE: zeroconf draft]

["Karen E. Nielsen (AH/TED)" <karen.e.nielsen@ericsson.com>]

> > 
> > I agree that that load balancing could be inforced
> > by the endpoint discovery mechanism using standard mechs, 
> e.g., round 
> > robin or advanced "active load measuring" mechanisms.
> > 
> > Personally I may still not be convinced that 
> > this is a strict requirement for zeroconf. 
> 
> I'm extremely surprised to see this view.  If a 3GPP operator has
> 1,000,000 customers (or even much more), do you think deploying just
> one server is sufficient, if the IPv6 tunneling code gets added to the
> phones?
> 
> Or what's your assumption in this kind of scenario?
> 

If 1,000,000 customers are using ipv6 the operators 
will implement native ipv6, I beleive.

But to the real question. 

I am not ruling out load balancing, but I want to think about
what the precise requirements are. 

Not having load balancing is not the same as to say that
you can only have one server, or if it is, then I agree that 
we should have some form of load balancing.

Karen