Delivered-To: dfarber+@xxxxxxxxxxxxxxxxxx
Date: Thu, 18 Sep 2003 04:02:11 -0500
From: "Stewart, William C (Bill), RTSLS" <billstewart@xxxxxxx>
Subject: RE: [IP] VeriSign update -- some numbers
To: dave@xxxxxxxxxx, karl@xxxxxxxxxxxx
Dave - Assuming that Karl's numbers are right, his conclusions are wrong.
(What Verisign is doing is still bogus, of course... he's not wrong about
that.)
20 Million messages/day is about 1000/sec during the busy hour,
and if they're 17KB of bloated Javascript, that's about 135 megabits/sec.
That means that Verisign has to _buy_ about an extra OC3 of capacity from
its ISPs,
and probably 1/3 of that traffic goes to them, 1/3 to other Tier 1 ISP's
that they have
big peering pipes with, and 1/3 goes to Tier 2 and smaller ISPs.
So the ~4500 small ISPs are carrying that extra 45 Mbps of traffic, about
10kbps each.
It's not zero cost, but it's pretty close, and if your margin's that thin,
null route Verisign's server.
Verisign's probably paying $15-30K/month to their ISPs for it.
The real costs are the bandwidth and support for broken email,
mostly extra bouncegrams that result from SMTP servers contacting Verisign
and getting rejected deliveries instead of being able to reject
misdirected mail earlier
based on DNS queries. There'll also be extra costs from spam,
because this breaks some of the tools used by spam blockers,
and Verisign will also get hit by spam to mistaken domain names,
especially if some spammer tries to dictionary-spam yet-another-typo.com.
Some people have spammer-bait web pages with lots of bogus email addresses -
in the past this was considered a bit rude because of the extra load it
placed on the root DNS servers,
but Verisign's increased that load now.
Bill Stewart bill.stewart@xxxxxxxxx
Disclaimer: This is my personal opinion, not my employer's.