[ppml] Effects of explosive routing table growth on ISP behavior
sleibrand at internap.com
Wed Oct 31 18:25:05 EDT 2007
Ted Mittelstaedt wrote:
> My guess is if this kind of thing ever came to pass you would see some
> kind of "dynamic filtering" come into place. In other words, the ISP's
> BGP peer router would look at ALL inbound and outbound traffic that
> passes through it, and automatically modify the filter to allow the
> route into the table for any source or destination of any packet going
> through the router. You would set the expiration of the "hole" in the
> filter to be proportional to the number and size of the packets and
> how long they were coming through the hole.
> Oh dear, I've probably done gone and screwed up someone's patent
Agreed. I've asked my router vendors for such features from time to
time. Yet more prior art. :-)
> But I still think that a lot of this is chicken little stuff. Think
> of how many credit card numbers there are in the world, and how any
> of them could suddenly appear at some card swipe machine anywhere in
> the world, yet the global banking community seems to be able to
> handle this just fine. The hardware exists to do that, it can be built
> to do this.
Agreed. However, doesn't dynamic filtering rely on having a
less-specific covering route, so you can at least route whatever packets
come in until you get enough of them to start accepting more-specifics?
I'm not worried that the routing table will explode unexpectedly, but I
am worried about implementing policies (such as allowing transfers of
/24s at the RIR level) that are incompatible with the kind of filtering
that would be needed to deal with a rapid routing table expansion.
More information about the ARIN-PPML