[ppml] Policy Proposal 2007-6 - Abandoned

[Jason Schiller]

---------- Forwarded message ----------
On Fri, 4 May 2007, David Williamson wrote:

> Quick question - if we could somehow fix the routing problem, would you
> be in favor of the now abandoned proposal?  I agree that the routing
> problem is serious and a major concern for any proposal that allows,
> frankly, any space to be allocated.

First off I don't think I would be in favor of the now abandoned
proposal.  As you point out there are a few issues (such as the spammers
using and black-lising lots of space) that need to be addressed.  I
suspect these have been addressed (or shortly will be).  However I still
have some concerns about their adequacy, but maybe that is just details.

> I think the concerns you've raised besides the routing table issues
> have been addressed by Owen, so I won't echo those thoughts.

I may have missed one of Owen's posts, but allowing ARIN to reclaim
addresses used illegitimately isn't enough.  I do admint, that
taking back resources from orgizations that stop being multi-homed would
probably be enough of a deterrent to prevent them for being multi-home for
just a year to get a PI block.  However, we need to prevent spammers from
getting lots of address space black-listted in the first place.  Once they
are black listed, it will be difficult to clean, and likely make the
reclaimed address space unusable.

This is why I suggested maybe if an orgization is multi-homed for 6
months, they should qualify.  This would likely deter spammers as they
would need to keep their accounts in order for 180 days.  This would also
deter many orgization who simply want PI addresses and have no intention
of actually multi-homing.

Ideally, I would like the policy to say anyone who is multi-homed prior to
Aug 2007 can qualify for PI /24s, but no new multi-homers will qualify.  
This is the only way the policy could be crafted to not create more routes
in the table, but of course that is not a balanced and fair policy.

As for the routing table size challenge, well, that depends on what
exactly you mean by "fix", but assuming the following:

1. I have no worries about my routers running out of RIB/FIB memory.
2. I have no worries about slow path selection
3. I have no worries about slow RIB to FIB population
4. I have no worries about slow boot time convergence
5. I have no worries about CPU utilization in handeling updates
6. I have no worries about doing FIB lookup at line rate w/ small packets
7. I have no worries that I will need to upgrade my routers just to keep
up with routing table growth (routers should last at least 5 years)
8. I have no worries that routers will be cost prohibitive
9. I have no worries that routers will be too large, or be too heavy
10. I have no worries that routers will run too hot or require too much
cooling or electricity

(did I miss anything?)

Then yes, if the consuming extra address space, and the consuming extra
routing slots issues are adquately addressed then I suppose I would vote
in favor of this policy.

Personally I suspect any "fix" that addresses the routing table size
challenge will be a new architecture where multi-homing, TE, and PI are
all accomplished without the need to route more specifics, or carry lots
of state.  Thus it is likely such an architectural approach would also
eliminate the need for PI addresses.

__Jason