Re: Unreachable aroot

[Masataka Ohta <mohta@necom830.hpcl.titech.ac.jp>]

Randy;

> >> Watching it seems to me to require a "reverse skitter"
> >> (http://www.caida.org/tools/measurement/skitter/), measuring the path
> >> to the same destination (the anycast address) from many points in the
> >> network topology.
> > I was thinking more of "do we mail this list everytime something goes 
> > boink?" as a first approximation.
> > My mailfilters wouldn't notice the increase much. But if others don't want 
> > the line noise, I won't mail about it.
> 
> for the current experiment, i doubt any of us will mind.

To write a research paper on the experiment, the measurement is strongly
desired.

> for root service, the phrase which comes to mind is "does not scale."

The correct phrase is "does not and does not have to scale".

For the authentic content of root zone, we do need a centric control.

But, reachability to a local root server is a pure local issue.

A cautious local ISP may periodically monitor reachability to external
local root servers they and their customers are relying on and modify
AS-path selection algorithm, all of which is local issues.

Of course, such an ISP is stupid, because a simpler, more local
approach for the ISP is to run their own servers.

							Masataka Ohta