why have parent for <wildcard> case if "top of your own tree"

From: George Michaelson <ggm@dont-contact.us>
Date: Fri, 16 Aug 1996 13:54:51 +1000 (EST)

I'm finding that now people in non-US locations are using .net and .org
a number of queries are going very strange paths. A typical odd case
(sorry, I suppose that can hardly be typical but you understand in context)
would be a downstream customer, whose domainname cannot be trapped by
a simple "anything .au via parent <x>" rule and instead is passed to NLANR.

IF we ran echo, we'd doubless find these bogons out. Because echo is causing
local security people grief, I've disabled it

[sidenote: why use a high-level port and not ICMP? ICMP would be MUCH more
 acceptable to firewall/security people, and code is out there (bsd ping src)
 and free...]

So instead, I reach an immediate child-connected resource via the USA.

I am tempted to remove wildcard parenting by NLANR, and only send to other
caches that which I can explicitly tell is likely to be optimal. That would
mean any 2-letter domain could be passed on, but .com/.org/.mil/.net/.int
might not go via other caches.

Another issue is that people are findin non-cached data is very very slow
to come back nowadays. If I did more DIRECT logged fetches, I think this
might improve a bit.

What do other people close to their own offshore pipes do? Do you mesh into
nlanr for a wildcard or for an explicit list of domains?

Received on Thu Aug 15 1996 - 20:56:42 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:32:48 MST