[ppml] Free Market

[David Conrad]

Paul,

On Aug 25, 2007, at 9:34 AM, Paul Vixie wrote:
> the limitation isn't on address space, it's on routing table size.

My understanding is that the limitation is actually routing table  
flux, not size.

> how many /16's can be carved up into /19's before the global  
> routing table
> exceeds the size of routers that other networks can afford?

All of them?  Perhaps you meant a prefix longer than /19 since there  
are only 524,288 of them theoretically possible (much less in  
practice) and we're told we can handle "2M routes today, 10M with no  
change in technology" (if you believe your friendly router vendor).

However, if what our friendly router vendors are telling us is true,  
the fact that you have a lot of prefixes in the routing system  
doesn't mean a whole lot.  What matters is how often those prefixes  
bounce.  The (perfectly reasonable) assumption is that the more  
prefixes you have, the more routing flux you'll get.  Of course,  
using limitations on trading address space as a means to try to  
control this flux is a bit like trying to limit carbon emissions by  
blocking the sale of oil fields, but whatever.  I guess if all you  
have is a sledgehammer, everything looks like a watermelon.

> so, provider-assigned (PA) has a provably bad short term economic  
> model (it
> functions as a price lock due to its renumbering penalty), whereas  
> provider-
> independent (PI) has a provably bad long term economic model (it  
> blows out
> the routing table).

PI has a _presumed_ bad long term economic model if you assume there  
are no additional constraints on the flux in the routing system.  As  
you are aware, in the past some ISPs have imposed additional  
constraints to protect their customers routability within their  
networks (namely prefix length filters and flap dampening).  These  
measures had both positive and negative impacts but it isn't like  
"OMG End of The Internet WE'RE DOOMED" is a foregone conclusion.

>> So, one way or another when IPv4 becomes scares, restricted or
>> unavailable, the market will handle it.
> by moving the problem from "us" to "our grandchildren".  just like  
> with oil.

With oil, or rather carbon dioxide (and perhaps less controversially,  
with sulphur dioxide) emissions, governments have implemented  
'tradable emission allowances', a reasonable short summary of one  
done in the US is at http://www.house.gov/jec/cost-gov/regs/cost/ 
emission.htm. One could imagine an environment where routing slots  
are managed the same was as SO2 emissions, perhaps with a periodic  
review of the maximum allowable number of slots to track improvements  
in technology.  Of course, to imagine this, you also have to imagine  
either government (which?) intervention or a level of cooperation in  
self-regulation amongst ISPs that stretches credulity.  Sort of like  
getting global agreement on addressing climate change.  Oh well.

Regards,
-drc