[ih] When did "32" bits for IP register as "not enough"?

[Vint Cerf]

removing all the previous emails and responding to the question about when
we realized
that address space would be a problem.

I seem to recall that around 1977 the splitting off of IP from TCP led to
discussions about
address size. There were three camps: 32 bit, 128 bit and variable address
lengths.
The latter was Danny Cohen's theme but software folks worried about all the
extra
cycles needed to find fields in the header if the address field were
variable length.
At the time, I thought 128 bit addresses seemed way over the top - even
though there
was a belief in some circles that we would be swimming in ethernet LANS and
need
the additional address space. I finally concluded, as program manager of
the Internet
project at the time that I did not want another cycle of implementation and
testing of
a new packet format and that there was plenty of room in 32 bits to test
the design
at reasonable scale. Remember this was only 4 years into the development
program.
We had a lot of other questions (some of which Jack Haverty and Craig
Partridge and
others have mentioned) that needed attention and changing the packet format
yet again would divert effort from dealing with more complex issues (some
of which
remain unresolved today). So I decided to stick with 32 bit address space
with
reinterpretation from 8 bits of network to A,B.C.D,E. Eventually CIDR
arrived
which development came after my departure from DARPA in late 1982.

Vint
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://elists.isoc.org/pipermail/internet-history/attachments/20190214/537b414e/attachment.htm>