<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
<title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
I would question why a small ISP would need to buy bigger hardware.
I've setup small ISPs and multi-homed businesses and know that most
don't really need full routes. <br>
A small ISP can always default. Chances are one of their upstreams is
either better or cheaper than the other one, and their backup link is
more costly or slower. It's always been costly to run full BGP, and if
small business or ISP doesn't want to keep pace, then that's their
choice. They won't be out of business if they start filtering or
defaulting. <br>
<br>
Anyway, I think when we're talking about the fundamental scalability
and future of the Internet, and basing policy that affects everyone -
the argument has always been that the core would collapse if we did or
didn't do X. ("the sky is falling" argument). These vendors and
presentations seem to show the complete opposite. <br>
<br>
<br>
<br>
David Williamson wrote:
<blockquote cite="mid:20070619171118.GJ24827@shell01.corp.tellme.com"
type="cite">
<pre wrap="">On Tue, Jun 19, 2007 at 09:54:07AM -0700, John Paul Morrison wrote:
</pre>
<blockquote type="cite">
<pre wrap="">- 5 to 10 years of growth, scalability not yet limited by FIB/RIB size -
(speed and forwarding path more the issue)
- No Need for Panic
</pre>
</blockquote>
<pre wrap=""><!---->
So the existing small ISP at the edge of the DFZ behind a couple of
DS-3s is going to be fine? Even if this poor sap buys bigger hardware
to accomodate a larger FIB/RIB, how's he going to get a full table
update (after maintenance or a crash) in a timely fashion?
I agree there's no need for panic, but there's need for some serious
work. The biggest players are going to be pushing the edge of the
RIB/FIB envelope. The next set down in scale (but still near the
"core", whatever that is) will be pretty much okay. The little guys
near the edge are going to be pretty screwed unless they put out some
serious dollars.
I don't think the scaling issues are solved for everyone. Not even
close. There may be solutions for some, but it's not general. The
vendors can stand up and pat themselves on the back, but that doesn't
mean this will all really work.
Put me down in the class of cynics. It's not entirely obvious to me
that the network as we know it today will continue to operate nearly as
smoothly as it does now (and I'm surprised that it does). Without a
forcing cuntion of some kind, there's no real possibility for change.
We can all hope that the transition to IPv6 does that. I suspect it
won't.
For ppml purposes, I'm very tempted to write a followup to 2007-6
that's even less restrictive. Let's remove all minimum allocation
sizes from ARIN policies. You get what you can justify. That would be
good stewardship of the number resources, even if it would be horribly
bad for the routing system. ARIN doesn't guarantee routability,
though, so I don't see a problem there.
If I do propose such a thing, I fully expect to get shouted down, even
though I think it might be in the best interests of the network as a
whole. If it forces a change in the routing system (a loc/id split
solution, or a way to monetize routing slots, or...something more
creative) - that's good!
Stoking the fires,
-David
</pre>
</blockquote>
<br>
</body>
</html>