[ppml] Longer prefixes burden the FIBs of DFZ routers
stephen at sprunk.org
Mon Aug 20 12:39:08 EDT 2007
Thus spake "Leo Bicknell" <bicknell at ufp.org>
> I'm quite afraid this sounds like early IPv4 routers. They did
> Class A's, B's, and C's quite well, and some of them when
> you put in other length prefixes fell over. I suppose it looked
> like a good trade off for the software and hardware of the day;
> and in fact it might have been the right choice to get us where
> we are today. Those designs didn't last well though, and got
> fork lifted out when a problem occurred.
I don't remember it occurring the way you describe. In the early 90s we
needed software upgrades to support CIDR-aware protocols due to address
space and routing table issues, which was painful but didn't require a
forklift, and in the late 90s we needed hardware upgrades needed to switch
from route-caching to FIBs due to the inversion of the 80/20 rule, which did
generally require a forklift.
> I'm afraid it would have to be a global policy to mean
> something to the vendors, but perhaps a "under no
> circumstances will RIR's require prefixes longer than /XX
> before 2015" might be a way to reduce deployment costs.
That wouldn't help, AFAIK, since internal routes are already up to /128 and
if forwarding packets to those destinations puts more strain on routers than
forwarding to a /32 or even /64, it'll become a (perhaps unintentional) DDoS
attack vector. The IETF has told vendors to not optimize for any particular
route length, and so far it appears they're heeding that advice.
Stephen Sprunk "Those people who think they know everything
CCIE #3723 are a great annoyance to those of us who do."
K5SSS --Isaac Asimov
More information about the ARIN-PPML