Minutes of the CIDR Deployment Working Group (CIDRD)
Address Space Growth Report
The usual report on the usage of the IPv4 address space was presented.
A revised estimate of 2018 +/- 8 years was given. The cause for the
recent decrease in the slope of the curve was discussed, but no firm
conclusion was reached.
Routing Table Size
As of Danvers, 25452 routes are present in the backbone routing tables.
CIDR routes have increased dramatically in the last year (from 113 in
Seattle to 2215 in Danvers). ASs doing CIDR have also increased
dramatically (35 to 289). More than half (56%) of the ASs are
announcing CIDRized routes, but there is also a high number (134 of
2215) announcing only one route. Tony Bates has a report he used to
produce that would show who has not CIDRized; this report should
continue to be posted. It appears that the total number of routes is
flattening out.
It was noted that flaps are an ongoing, serious problem and that the
flap rate of a prefix is proportional to the length of the prefix.
CIDR has had a significant impact on the number of theoretical
hosts/route. The effect also is different among the various blocks.
There is even an effect in 192.
Class A Allocation
The IANA has not yet released any Class A networks to be allocated. The
working group would like to survey the Class As to see which ones are
actually in use and are globally routed and which are unused. There is
also a need to provide a guideline document to IANA for the allocation
of As. Scott said that the IESG has discussed this, but is looking for
a document that helps define the situation and the potential impact of
using Class As. There has been a suggestion that a Class A be leased to
someone to do evaluations. Geoff Huston has an extensive note on the
topic in the IEPG archives and Bill Manning will work on helping to
document the IANA guideline. Why do this at all? Class A is the last
25% of the address space available. There is not a specific desire to
allocate this space, but there is a desire to be sure it is possible
before it is actually done. Matt Mattis suggests that each provider get
a subnet out of a particular A and see if they can communicate with each
other. There are also reverse lookup DNS issues. As of this writing,
an experiment has already been undertaken to test routing of subnets of
a Class A network.
Address Ownership
Yakov Rekhter gave a presentation on the effects of address ownership on
the scalability of Internet routing. This talk concluded that address
ownership is impractical if we intend to continue to use hierarchical
routing within the Internet. This talk is to become an RFC under the
CIDRD Working Group. There is consensus that renumbering is something
that should be promoted.
A proposal has been made to create a document that will advocate that
providers filter out prefixes of a certain size. What does that mean
for multi-homed networks? Yakov will create this document. There is
also an idea that settlements should be based on number the of routes or
unaggregated routes.
Progressing RFC 1597 to Proposed Standard
There has been a suggestion that RFC 1597 be moved from Informational to
Proposed Standard. Revisions are necessary to incorporate the comments
from RFC 1627 and to make the text classless. The authors of the
respective RFCs have agreed to collaborate to produce the synthesized
document.
