see also - http://www.sobco.com/ipng/big_ten/big_ten_packet_format.txt
and
http://www.sobco.com/ipng/big_ten/big_ten_address_format.txt
variable length addresses were considered for IPng (now IPv6) - for various
reasons,
including the same issue Int raised of perceived programming complexity in the
hosts, the idea was not adopted
Scott
On Nov 12, 2016, at 2:44 PM, Bob Braden <braden(_at_)ISI(_dot_)EDU> wrote:
Brian et al,
I have not been paying attention to this thread, but Brian's answer got my
attention. I was in the room in 1977 when the decision was made to use
fixed-length 32 bit IP addresses. I did not have a strong opinion at the
time, but at least two members of the Internet research group, Jon Postel and
Danny Cohen, strongly urged variable length IP addresses. As manager of
ARPA's Internet Research program, Vint Cerf made the call for fixed length
addresses. His argument was that if we were to have any hope that the DoD
might accept the experimental protocols, TCP must not be too complex to
program. He thought that variable-length addresses would scare off our
possible customers.
Incidentally, the OSI folks did adopt variable -length network addresses;
this was one of the advantages of OSI that later led the IAB to suggest that
the Internet shou ld consider adapting the OSI protocols. Which caused the
famous Kobe lynching of the (original) IAB.
Good heavens, 1977 was 40 years ago!
Bob Braden
for a problem, that is IMHO not very clearly described.ally,
how about ipv4 and ipv6 are incompatible on the wire and this
has created a multi-decade ipv6 charlie foxtrot?
Yes, I suggest mentioning that to Vint, Bob and a few others in 1977,
so that they can design IPv4 with extensible addresses. People in
2016 will be grateful.
Brian