>From tha archives, "J" was only supposed to be at NSI for a temporary
period before moving to a different location (and organization), much like
"L" and "M" moved to LINX and WIDE after a brief period at ISI and NSI.
The real question isn't why "J" has moved a few miles to a different
Verisign building, but where in the world should "J" move?
>From my limited understanding of the data, Hong Kong appears to be the
most technically sound location for a new root server. Asia-Pacific rim
is heavly dependant on "M" now. Yes, a lot of A-P traffic is exchanged on
the west coast of the US. But HK is probably the second most central
telcomm location for the regiona. South America, Africa, Russia, India
have lots of people, but aren't very central network-wise. Root servers
need to be able to serve the world, not just a local region or country.
patience grasshopper.
pushing "J" to a distinctly different broadcast domain
is the first step to pushing that instance elsewhere.
pre-ICANN, things moved fairly quickly as compared to
post-ICANN.
Perhaps we shouldn't be saying "post-ICANN" as we're in "ICANN age" now,
if they were to be gone, it then be "post-ICANN" and things might even
move faster (or not at all ...
But seriously are there any volunteers there to run root name servers in
Europe and Asia or are people now expecting to get paid to it through
ICANN contract. We do need root name server for every continent and
perhaps something like "official mirror" should be considered where
somebody would run nameserver with complete mirror of all zones that root
name server would have but it would not be considered "official root" name
server (but ISPs in its region would know about and us it). Are other
regions ever considered something like this to ease load on current root
servers or perhaps as a first step to having root server there?
Given the numbers presented on http://www.nanog.org/mtg-0210/wessels.html,
I suspect that the actual location won't matter. As both Wessels and
Vixie have demonstrated, if we wanted to cut the load 30% real fast,
we'd implement 1918 filtering. We wanted to cut the load 98%, we'd fix
all the OTHER stupidity.
