Noel,
In the interest of
completeness I would note that at the time the size of the global Internet
routing table was also a very high concern and core to at least one session at
each IETF meeting at the time. Pre-cidr we were at risk of running out of
class B addresses and thus causing a blowout in allocation of class
Cs. In short, the IP layer was in need of some "firming up", on
multiple fronts. (too bad we didn't go farther ...)
As a co-chair of the
TUBA (tcp/udp with Big Addresses) wg that recommended the use of CLNP I
would note most people in the TUBA wg supported the use of CLNP
because:
0) CLNP satisfied
the requirement for a globally unique huge address space and CLNP
and IPv4 were fairly comparable from an architecture perspective (thanks to
several people ( Dave Oran, Ross Callon, Lyman Chapin, Dino Farinacci, Tony
Hain, Yakov Rekhter, Radia Perlman, Tony Lauck, Brian Carpenter,
Richard Colella, etc.) who spanned the IETF and ISO/OSI
worlds )
1) there were ample existence proofs of
CLNP in router and host software that were shipping at the time
1a) several independent
implementors successfully moved TCP/UDP on top of CLNP along
with several apps (email, telnet, ftp) on several platforms in a
super short period of time (BSD, DOS, etc.)
2) there were existing, working, and
deployed routing protocols on a global basis (ES-IS, IS-IS) - e.g. no
tunneling to prove "it works"
3) CLNP supported autoconfiguration of host
addresses (ES-IS) (best feature of IPv6!)
4) it was felt that at the time it would be
the "quickest" transition, since time was of the essence!
5) CLNP also allowed for short addresses
via variable length addressing (e.g. use of IPng in private supercomputer
clusters, etc.)
6) the address space had a natural
delegation architecture which could suit the needs of a variety of parties at
the time.
Many of these same reasons were held to be
true by people who worked prior to TUBA in the pre-IPNG
discussions/debates.
Christian is correct in that several people
felt the answer of CLNP was "THE answer" and thus were over-zealous in pushing
in that direction and aligned the discussions of IPng with other discussions re:
CLNP. (keen time observers will note that the TUBA effort and the
Kobe "stuff" were unaligned. Several people who supported CLNP did
not believe they needed help from the IAB, especially in light of the IETF's
political climate at the time regarding the IAB, the NSF, the funding of the
IETF and NIC, etc., etc.)
In most ways, IPv6 has replicated all of
the properties enumerated above and thus, it is now clear that we do not need to
transition to CLNP.
regards, peterf