[ppml] Draft 2 of proposal for ip assignment with sponsorship
Alec H. Peterson
ahp at hilander.com
Fri Feb 28 09:38:40 EST 2003
--On Friday, February 28, 2003 8:27 AM -0600 Forrest
<forrest at almighty.c64.org> wrote:
> If you get a /24 from one of your upstreams, your /24 is part of your
> ISP's larger aggregate. The people arguing routing table explosion
> caused by a micro allocation policy seem to take the stance that they
> can filter your /24 from your ISP's aggregate and you'll be fine because
> they'll still have a route to you via that aggregate. If you receive a
> micro allocation they will no longer be able to filter your /24.
That is precisely the point.
> If you filter my /24 from my provider's aggregate, what exactly is the
> point of me multihoming? If I lose connectivity with that upstream,
> you'll no longer be able to reach me because you won't hear my /24
> announcement from my other upstream. To me, it doesn't seem acceptable
> to filter out multihoming /24's at all. Creating a micro allocation
> policy would seem to address this issue. Create a block of addresses
> that /24's won't get filtered from, while still allowing everyone to
> filter out the garbage more specific /24's from elsewhere.
You are forgetting that each network provider is free to decide which
routes it will carry. Just because you connect to provider 'a' and
provider 'b' does not mean that provider 'c' has to listen to each and
every route that 'a' and 'b' carry.
Now I grant that filtering more specifics of an aggregate does reduce the
benefit of multi-homing, but my point is that keeping things the way they
are at least gives service providers a way to keep the network running in
the event that their routers can't handle the large routing table size.
Wouldn't you agree that in that situation it is better to have a reduced
benefit to your multi-homing than have an entire backbone shutdown?
I know I know, it will never happen because of Moore's law and we have such
powerful routers now. But it is precisely that type of attitude that got
us in the hole in the 90s. Given history I think it is remis of us to not
look very very closely at our past before making any decisions going
See, the major problem here is the method we are using for multi-homing.
IPv4 was not desgined with this issue in mind (well perhaps it was, but
certainly not with the scale issues that we have been faced with). This is
why the multi6 IETF working group exists, to find an elegant way to
multi-home using provider-assigned address space. Clearly we are not going
to solve this problem in in the ARIN public policy working group, but
people involved in this discussion may have valuable insight for the multi6
Alec H. Peterson -- ahp at hilander.com
Chief Technology Officer
Catbird Networks, http://www.catbird.com
More information about the ARIN-PPML