The cost of nines

[Eric Brunner-Williams <ebw () abenaki wabanaki net>]

Hi all,

On the 7th of next month I'll be participating in an ICANN 
consultation on the proposed draft registry agreement, and the number 
of "nines" that have crept into it, relative to what was expected of 
new registry operators a decade ago, is one of the hidden cost 
increases I will discuss with ICANN's lawyers, who are responsible for 
the extra nines.

I'm looking for sources of cost-per-nine, network provisioning, and 
host provisioning, where "host" is usually a bunch of boxen, not just 
a pizza box.

The way the requirements are now, a startup of another .museum, say 
for libraries or archives, or a new .coop, or a new linguistic and 
cultural say a .scot, has to provide a higher level of performance 
than Verisign currently does for com/net/name, which is slightly 
absurd, if not worse.

I can cite sources, or not, as preferred, and while CORE is 
comfortable at any number ICANN's lawyers can come up with under the 
theory that "more nines is what security and stability mean", my goal 
is to allow real startups, like .museum and .coop were in 2001, not be 
forced to outsource registry operations to an already highly 
capitalized registry service provider, for competition policy reasons.

I'm also "in the market" for recent failure data, such as Ultra's 
yesterday, and Verisign's v6, not for competitive reasons, but to show 
that the SLA expectation of ICANN's lawyers may need modification if 
placed proximal to actual operational failure data.

Off-list or on, and thanks in advance, from my Yule tree to your own.

Cheers,
Eric