[ntp:questions] Re: Cryptography

Dale Worley worley at dragon.ariadne.com
Thu Dec 11 12:57:30 UTC 2003


wolfgang+gnus20031210T193408 at dailyplanet.dontspam.wsrcc.com  (Wolfgang S. Rupprecht) writes:
> Well, I meant that mostly tongue in cheek.  After all if a client is
> going to talk to the stratum 1's at all, it might as well ask "by the
> way what time is it?"

That's what I'm trying to figure out how to avoid -- can the clients
get trustworthy information from the stratum 1 without talking to it?

> On the other hand, spot-checking on the stratum-2's by asking the
> stratum-1's to confirm the time only once in a great while could be
> an order of magnitude lower load (or more if the information that
> the stratum-2 appears truthful can be shared with a number of local
> machines that all trust each other to be trustworthy.)

OTOH, the scheme I outlined, while providing much less good
information to the client, can give the client an indication every,
say, minute.  And if we have 10^10 computers with the tree branching
by a factor of 100 at every level, there are 5 levels of branching,
and the stratum 1's see 10^8 less load.

(Not to mention that if the tree is laid out based on network
transmission cost, a tree can cut the network load significantly.)

Dale



More information about the questions mailing list