[ppml] HD Ratio Applied to IPv4

[Charles Scott]

Paul:
  OK, I've read the RFC, but don't agree with how this affects allocation
from ARIN. I said before that I think each level of allocation/assignment
needs to stand on it's own justification. The current policies set up
parameters for this that I am hard-pressed to see why they are anything
but reasonable.
  RFC 3194 seems to incorrectly suggest that IP utilization is strictly
hierarchical. In fact, it's a segmented flat space where the utilization
of a specific segment is only minimally affected by how any other part is
utilized. The RCF uses phone numbers as an example, however, unlike
telephone numbers which are strictly hierarchical, network numbers are not
totally hierarchically dependent on each other unless you are truly anal
about routing aggregation at each and every level. While I would agree
that route aggregation is a concern, in most cases a provider assigns
space to downstream customers and as such routing for that space is
aggregated at their level anyway.
  Essentially, there are two types of levels in the hierarchy of address
usage. One is an allocation of an address block by ARIN or a provider, and
the other is utilization by an end user. Any level of assignment simply
concerns the current available space at that level and the assignments to
the next lower level. The 80% criteria seemingly provides adequate
available space for assignment at a particular level. Once an assignment
has been made, there is no other concern for lower levels of the
hierarchy other than to ensure that the recipient can justify the
assignment. The 80% criteria then similarly applies to that assignment if
the recipient is another provider, otherwise the end-user utilization
criteria apply.
  Since a provider does not need to show composite utilization for all
address space assigned to a lower level, only a certain level of
assignment (80%) and justification for those assignments, the overall
utilization is not the critical number. So where's the beef in this 
proposal? Keep in mind that if a particular end-user assignment meets the 
25%-50% criteria (clearly generous), the assignment is supportable and the 
provider ticks off the assigned block from their total. Also, if a 
provider assigns space to themselves for use, each such assignment that 
meets the 25%-50% criteria is considered consumed with respect to 
assignment and there is no need by the provider to show any particular 
level of composite utilization. 
  I also don't buy the argument that renumbering is required when an
end-user needs more space. This would be the case only if IP addressing
had to follow a strict hierarchical structure as suggested in the RFC.
  Perhaps I'm still missing something, but I still can't see any 
justification for jeopardizing additional IP space with this proposal.

Chuck




On Thu, 14 Aug 2003, Paul Wilson wrote:

> For background on the HD ratio, including the justification of this approach
> in preference to percentage-based utilisation measures, please see RFC3194.