Address block problem

[Andy Dills]

On Mon, 31 Jul 2000, Bruce Robertson wrote:

> > I mean, 7 times out of 10 there will be two routes for the given CIDR
> > block...different prefix lengths for managing inbound traffic, multi-homed
> > customers, etc. So number of routes isn't a large consideration, at least
> > to me.
> 
> Hmmm, I must be one of the 3 out of 10... I would never use anything but
> a single advertisement for the entire block.

Here's an example of why one would do this: we have transit with UUnet and
Abovenet. We service some customers from our colo'ed pop at Abovenet.
Thus, I announce an additional /22 to get traffic to return directly to
our customers via Abovenet, instead of UUnet-us-Abovenet.

BTW, my 7 out of 10 figure was wildly OOMA. I just remember seeing the
number of needless routes leaked by UUnet (over 200), and at that point I
realized that as long as all of my routes have a purpose, things will be
ok. You need 128MB to reliably take 2 full views nowadays anyhow...

> All of my inbound traffic is considered equivalent, and I force my
> multihomed customers to get their own address space.

Why do you force multihomed customers to get their own address space? You
need to be using 8 /24's to get PI space. None of my multihomed customers
come anywhere near qualifying.

De-aggregate routes are not an implication of sloppy routing, but sloppy
routing will often utilize de-aggregate routes.

> > Is it just an annoyance thing?
> 
> Mostly, yes.

That's cool, I can feel for you. I certainly can imagine that finding out
this information would put a sour note onto the end of your day...

Andy

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Andy Dills                              301-682-9972
Xecunet, LLC                            www.xecu.net
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Dialup * Webhosting * E-Commerce * High-Speed Access