[arin-ppml] Rationale for /22
bicknell at ufp.org
Thu Jul 30 13:56:43 EDT 2009
In a message written on Thu, Jul 30, 2009 at 12:19:19PM -0500, George, Wes E [NTK] wrote:
> To the question of "existing filters/inertia aside, is there a
> justification for *any* minimum allocation size?"
While it is a bit of a reductio ad absurdum argument, consider what
(could) happen if the minimum were a /32.
Everyone with a computer can justify a single address. There's no
way global backbone routers can support a prefix per-user, even for
a modest amount of users. Many of these users would be single
homed, taking the main advantage of a portable /32 as being a
"vanity" IP address. Presumably the price for a single address
would be so low (see cost of domain names) a substantial number of
people would not see the price as a barrier. Even if direct providers
wouldn't route these (e.g. Cable/DSL providers), it's likely tunnel
brokers could offer services to end users for modest costs.
I think many folks miss the most important item though when considering
the effects of minimum allocation size, or other polices that are
likely to affect the size of the global table. The single most
important thing is predictability. ISP's are buying equipment now
based on growth assumptions. They want that equipment to have a
reasonable lifespan (3, 5, 7 years?). If our policies and actions
allow them to predict with reasonable accuracy that the table will
be 500k routes in 3 years, and 700k routes in 5 years, and they can
buy a 1M route box, well then all is good.
But if they bought that box yesterday, and today we change the
policy such that in 3 years there are 1M routes then that ISP is
going to be in a huge bind. It's either going to have to spend a
lot of capital early, or it's going to have to drop routes.
We've had a trend of lowering the minimum allocation nice and slowly
over the past 5-7 years, and we've been carefully watching the
results. I think this is generally the prudent course of action,
there is a huge penality for overshooting in terms of disruption
to the global network.
Leo Bicknell - bicknell at ufp.org - CCIE 3440
PGP keys at http://www.ufp.org/~bicknell/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 825 bytes
Desc: not available
More information about the ARIN-PPML