Re: Experiments in multi-placed root servers

[bmanning@ISI.EDU]

> Equinix-provided switch.  We are not an ISP or in the bandwidth
> business, but we will have our own AS and a corporate WAN which we can
> use to synchronize delivery of services. We will also be able to
> provide the single point of contact discussed at the BOF.
> 	Our aim would be to use a single name and a single IP,
> announcing the availability of that network at every center and
> delivering the traffic to the server physically located at that
> center.  There are some interesting synchronization and routing
> problems details, but we don't see any current gotchas.  We'd
> be interested in working with the community on this, especially
> to make sure that the system works in the face of real traffic.

	There is such technology deployed already in a number of 
	places, GTEi has some IP patents on a thing called "hopsotch"
	and Centergate is doing similar work with a product called
	ultraDNS.  

	Without much more comment, I'm not sure that this is a sound
	idea for a DNS root.  This tactic has been promoted over the
	last few years and has never been shown to be a viable 
	stratagy since the failure modes are so baroque as to make
	debugging/troublshooting such events lifetime employment
	for certain classes of engineers.  While things are improving,
	I remain unconvinced that something as critical as a DNS
	root be constructed in such a fashion.  There are better 
	ways to ensure redundency.
	
> 	Comments, questions, interest in helping out?
> 			regards,
> 				Ted Hardie
> 				Equinix
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 


-- 
--bill