InterNIC - We put the "I" in incompetent.
Michael Dillon
michael at memra.com
Sun Mar 14 20:34:56 UTC 1999
On Sun, 14 Mar 1999, I Am Not An Isp wrote:
> >Nothing except community consensus that it is a *BAD* thing to pollute the
> >global routing table with lots of long prefixes.
>
> I am capable of putting /32s in my network an announcing the aggregate. We
> did this at Priori, Michael. Justin programmed each loopback as a /32 out
> of the same /24, so we had x.x.x.1 on the west coast and x.x.x.2 on the
> east coast, but still only announced the /18.
Most of the people on this list do not operate a national backbone. This
aggregation technique is just fine if you really do have geographical
diversity of nameserver location inside your AS. I'm not sure why anyone
would go to the trouble of making it appear that their nameservers are in
the same room when they are not. Priori had nameservers at PAIX and at
Erols in Fairfax County, VA.
But if a network does not have geographical diversity inside their AS then
all of this /32 aggregation magic is for naught. If national backbones
look like this:
NS0.VERIO.NET 205.238.52.46
NS1.VERIO.NET 204.91.99.140
NS1.SPRINTLINK.NET 204.117.214.10
NS2.SPRINTLINK.NET 199.2.252.9
NS3.SPRINTLINK.NET 204.97.212.10
NS.CW.NET 204.70.128.1
NS2.CW.NET 204.70.57.242
NS3.CW.NET 204.70.25.234
NS4.CW.NET 204.70.49.234
then why wouldn't all ISPs look like this?
NS1.EXAMPLE.COM 192.0.2.17
NS2.EXAMPLE.COM 204.97.212.10
NS3.EXAMPLE.COM 205.91.99.140
instead of the minimalist slapdash technique
NS1.EXAMPLE.COM 192.0.2.44
NS2.EXAMPLE.COM 192.0.2.45
It's not hard to find an ISP in another state or another country to trade
secondary DNS. And when the backhoe cuts a major fiber link in your area
used by all three of your upstream providers, the world will know that you
still exist. If you use the minimalist slapdash technique the world will
think that you've gone out of business.
--
Michael Dillon - E-mail: michael at memra.com
Check the website for my Internet World articles - http://www.memra.com
More information about the NANOG
mailing list