Re: IPv7 (CLNP) a mistake

[Steve Deering <[e-mail]>]

I share Craig's dismay with the IAB's decision.

In addition to the real and serious problems that Craig raises concerning
change control over the fundamental protocol of the Internet, I am concerned
about technical and procedural aspects of the IAB decision.

Technically, changing from IP to CLNP means trading addresses that are
too short for address that are too long (yeah, I know, memory and
bandwidth keep getting cheaper, but it's a shame to fill it up with
a bunch of bytes that just say who got to allocate the next bunch of
bytes, not to mention all the null padding), plus a checksum that's more
expensive to compute in software and a bunch of non-32-bit-aligned
fields that are more expensive to parse.  If I understand the transition
plan, it also means adding an entirely new set of router-to-router and
host-to-router protocols to all of the boxes in the Internet, in addition
to the IP-based protocols already present, which will have a significant
cost, not only in terms of computing and network resources, but more
importantly in terms of the manageability of the Internet and the training
of its users and maintainers.

If all we want is bigger addresses (and that's all that CLNP gives us),
it would be far better just to make the minimal changes to IPv4 and
its routing protocols to carry bigger addresses, in such a way that much
of the existing routing, management, and "cultural" infrastructure can be
retained.  The recent proposals by Hinden & Crocker and by Zheng Wang
appear to be viable examples of that approach.

On the other hand, if we are going to make a change of the magnitude
of that suggested, let's get something more out of it than just bigger
addresses.  As Craig pointed out, the Internet is just starting to
see an explosion of new applications and services -- it's not just
telnet, FTP, and email any more.  We need a more capable infrastructure,
not just a more scalable one.  This is an exciting period of creative
activity in the Internet; it would be a real shame to derail (and
possibly smother) that activity in a small-gain, big-pain transition
to CLNP.

Procedurally, I am dismayed at the undemocratic and closed nature of the
decision making process, and of the haste with which such a major decision
was made.  I understand the need for immediate action on the CIDR front;
however, given the reprieve of CIDR, I am not convinced that the IP address
space will run out so soon that there is no time for an open evaluation of
the longer-term alternatives, and possibly even a building of concensus in
the community.  I thought the IESG recommendation of a six-month evaluation
period was very sensible; instead, we got a decision handed down from a
closed IAB meeting, based on the work of the closed ROAD group.  (I was a
member of the ROAD group, invited to participate after I discovered its
existence by accident and asked what was going on.)  I also am not
convinced that the ROAD group had sufficient time to evaluate the
alternatives.  For example, Bob Hinden came up with his particular
encapsulation scheme part way through the ROAD activity, and it has
evolved considerably since the last ROAD meeting, addressing most of its
previously-identified shortcomings.  Another example is Paul Tsuchiya's
PIP proposal, which has emerged since ROAD.  I hope the IAB will explain
why it deemed it necessary to endorse CLNP in the midst of active
development of less disruptive alternatives (e.g., the Hinden scheme) and
more capable alternatives (e.g., PIP).

Steve Deering