Re: [OFFTOPIC] Mailing Lists

Jared Mauch (jared@puck.nether.net)
Fri, 23 May 1997 15:54:58 -0400 (EDT)


Christoph Lameter graced my mailbox with this long sought knowledge:
> vger is frequently down and there is quite a bit of e-mail going around
>
> (last Monday 120.000 on .com-exploder and 50.000 on .edu exploder!)
>
> Could we come up with a scheme to decentralize the mailing lists so that
> one host cannot result in everything failing and some means of automatic
> load determination? Sounds complicated... Maybe someone here has an idea
> to do it in a simple and easy way.
>
> I have a 400MB/day limit by the provider on the .com exploder and I was
> definitely over it on Monday.

I run the .net exploder here, and peak about a t1's worth
of bandwidth on my port, with an average of about half a t1's worth
of bandwidth that I use here.. This machine is a dual p133 and
my desktop machine, and I work at home some of the time, so I can determine
what's related and unrelated traffic based on the graphs
of my ethernet port...

(I can supply pointers to my utilization for those interested)

The biggest part of the problem with why exploders are required
is that so many folks have broken DNS.

Example:

7BIT (host map: lookup (gaianet.net): deferred)
(host map: lookup (mary-world.com): deferred)
7BIT (Host unknown (Name server: auralis.airmail.net: host not fou)
7BIT (host map: lookup (teclink.net): deferred)

This is just a little bit, I scan this list daily and get these folks
removed asap to reduce the load on my system.

This is due to a few things:

1) people who have a t1 or something like that from MCI or Sprint
or someone else, and have both their nameservers on the same subnet, so
when they lose connectivity to the rest of the world, both their nameservers
are unreachable (This is the reason for secondaries in geographical
different locations).

2) folks do stuff like this:

7BIT (Deferred: Connection timed out with jacques.ebicom.net.)
; <<>> DiG 8.1 <<>> any jacques.ebicom.net
;; res options: init recurs defnam dnsrch
;; got answer:
;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 6
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 3, ADDITIONAL: 3
;; QUERY SECTION:
;; jacques.ebicom.net, type = ANY, class = IN

;; ANSWER SECTION:
jacques.ebicom.net. 11h18m33s IN A 205.218.114.247

With no MX records, or anything, when there other hosts on their
network which could be taking the mail and holding it locally within their
control if their server is down for some extended period of time. (Ex:
the ebicom.net mail host is up and running just fine, this other host is
down though, so I have to block waiting for the connection to time out to
deliver the mail to this person).

I'm willing to do the *.edu exploding again, those domains tend
to be well maintained, but the *.com domains are the ones that really suck,
because every tom, dick, and harry have one of those, and it's really a
pain in the ass to do mail delivery to them, because every little podunk
provider that thinks they know what they're doing puts all their
nameservers in one basket (or ethernet segment).

- Jared

-- 
jared@CIC.Net - CICNET --------- jared@Nether.Net - Nether Network
"I've got a question"  "What is it?" "An interrogative expression often used 
to test knowledge, but that's not important right now."