|
---------------------------------------------------------------------------
Rapporteur: Rob Blokzijl, RIPE Chair
---------------------------------------------------------------------------
1. Internet Routing - How it Works
Ray Plzak, ARIN
2. IPv4 & IPv6 Address and Assignment Policies and their Constraints
Rob Blokzijl, RIPE
3. Discussion: IPv6 and Routing
Pankaj Agrawala, Ministry of Communication and Information Technology,
India
4. IPv6 Address Distribution Mechanisms
Geoff Huston, Paul Wilson, APNIC
5. IPv6 and Innovation
Jordi Palet, Consulintel
6. Discussion: IPv6 and Routing
Milton Mueller, University of Syracuse, USA
7. Q&A and Discussion
This report highlights the points the speakers considered as the core
messages they wanted to convey. The agenda was rather packed which did not
leave much time for questions and discussions.
Ray Plzak gave a short, but extensive, overview of how the routing
system on the Internet works, both for IPv4 and IPv6. It was emphasised
that routing is based on IP addresses. Domain names do not play a role
in transporting data over the Internet. It was also made clear that
routing is performed by Internet Service Providers (ISPs). End Users are
generally not involved in routing.
Routing policies are defined by individual ISPs determining how they
interact with their nearest neighbours. Since there are tens of
thousands of ISPs that make up the global Internet, one can imagine
that the resulting routing system is not trivial, to say the least.
Rob Blokzijl explained that the two most important constraints for IP
address assignment policies are the conservation of address space and
the aggregation of routing information. Aggregation can be understood
as conservation of routing space. Routing equipment can only accomodate
so much routing information. This is a technical fact, not policy.
These two conservation constraints result in two opposing requirements
for the size of IP address block assignments. Address conservation is
best served by small blocks, whereas routing conservation is best
served by larger blocks. It is the responsibility of the ISPs to find a
compromise that best serves the operation of the Internet. This
compromise is what we call policy.
Rob Blokzijl also explained that the address policies of the Internet
have changed over time. Basically two policies are in place today: one
for the Internet up until the early '90s, and one for the Internet after that time.
These two different policies have to be supported by all ISPs. This is
a burden that has to be accepted. However, most ISPs are not
particularly pleased with current proposals to introduce numerous new
policies outside the current system of industry self-regulation.
Pankaj Agrawala posed two questions as input for a discussion that
unfortunately did not take place due to time constraints.
The first question dealt with the problem of whether under the current
IPv6 policies large, incumbent, ISPs are favoured in the allocation of
IPv6 address space.
The second question concerned the problem of local connections between
ISPs, where large players refuse to connect to smaller players. This
seriously hinders the development of a local Internet industry.
[Rapporteurs note: these issues certainly merit some time for
explanation and discussion. The first one is more a matter of
perception than an actual fact. The second one is usually related to a
near monopoly position of a large incumbent player. Here a national
government could play a regulatory role.]
Geoff Huston started by stating that the Internet, and IPv6, is not an
experiment. There is a huge expectation and a big vision: It
is an investment platform for trillions of dollars over the next
decades. It will be the communications infrastructure for many
generations to come.
Networking is not a game of chance, packets need to go where they are
meant to go. The routing system must work! Networking has to be simple,
cheap, stable and scalable. And routability is an essential part of that.
It is not magic, it is constrained by mathematical restrictions.
There are a number of risks that could make that system fail. There are
a number of alternative distribution models being proposed currently
e.g. national distribution channels. These alternative models are based
on either political grounds or on some notions of introducing competion
in policies.
What is not clear in all of these proposals is how they could be
implemented, or how they would affect the routing system. One should
remember that the current policies are the result of many years of
industry self-regulation.
- Discussion:
The question was raised as to whether a market would be developed in
address space.
The speaker explained that that was highly improbable, especially in
IPv6. New requests for address space are satisfied by inserting
new address blocks into the system.
Another question was raised concerning the possibility of retrieving
unused address space for future assignments.
It was explained that address retrieval is both very complicated
and costly. Within the current system, the problem is studied from
time to time, but so far no real incentive has been found.
Some substantial address retrieval has been done in the address
space that was handed out in the early days of the Internet. A
substantial amount of IPv4 address space has been returned to the
central IANA pool this way.
Jordi Palet gave an overview of new applications that were especially
suited to run over IPv6. Most came in the realm of end-to-end
applications. These applications are made easier by the fact that the
increased address space of IPv6 does away with the need to hide
end-user equipment behind address translation equipment.
As an illustration of new applications, Jordi Palet showed how he could control
the blinds in his house in Barcelona, scaring his cat by doing so.
Milton Mueller introduced a proposal to create an alternative system
for the assignment of IPv6 address space. He explained that the current
system of industry self-regulation as represented by the Regional
Internet Registries has worked quite well so far, and is inherently
robust.
However, he advocated that it might be worthwhile to create a number of
alternative mechanisms with the aim to see whether this would result in
additional address policies.
- Discussion:
Participants raised the question of how such alternative
policies would be implemented in the current routing system, and
what the potential risks to the system would be.
The speaker replied that he had no answer to both questions and that
the technological constraints would have to be investigated.
|
