[43827] in North American Network Operators' Group
Re: Digital Island sponsors DoS attempt
daemon@ATHENA.MIT.EDU (Ian Cooper)
Fri Oct 26 13:40:41 2001
Mime-Version: 1.0
Message-Id: <p05101007b7ff4bf0f270@[64.174.134.222]>
In-Reply-To:
<EF4A9841BCC9D5119E28009027923DF0137061@yosemite.icn.state.ia.us>
Date: Fri, 26 Oct 2001 10:26:45 -0700
To: "Quibell, Marc" <mquibell@icn.state.ia.us>, nanog@merit.edu
From: Ian Cooper <ian@the-coopers.org>
Content-Type: text/plain; charset="us-ascii" ; format="flowed"
Errors-To: owner-nanog-outgoing@merit.edu
At 09:45 -0500 2001-10-26, Quibell, Marc wrote:
>Curious Chris,
>What IP's are they probing? LAN IPs? Or just the DNS server as they claim?
Er, what if the DNS "server" happens to also be the host generating the query?
>And what then do these pings have to do with [quote from D.I.] "finding the
>nearest servers from our customers (Microsoft) to you"? What good does it do
>to ping to find distances,
>since the timing on the internet is not a constant, and neither are the
>pops, routes, hops...etc. This is why we use routing tables that are not
>constant...
It's one additional measure that probably wasn't available
beforehand. One could also potentially correlate with what the
routing tables are suggesting to see if there's a "better" way to get
content out of a different surrogate to the requesting host(s).
Routing tables define reachability but aren't so hot at determining
the actual performance of an end-to-end connection. If you're only
interested in the packets getting there "eventually" then that's OK.
If you need the packets to get there in the most optimal manner (and
yes, "optimal" can mean many things) then you probably need to look
at some other way to do things.
--
