RE: [Fwd: I-D Action: draft-carpenter-6man-why64-00.txt]

[Christian Huitema <[e-mail]>]

> But maybe someone who has Christian's book could send a paraphrase
> of his explanation.

I have of course a few copies on my shelves. Mark ZZZ Smith is probably referring to the Chapter 4 of "IPv6, the new Internet Protocol," which was published in 1996. The chapter deals with "plug and play," and the first paragraph reads:

"By the end of June 1994, the IPng saga was coming to a close. The IPng selection committee had almost made up its' mind. The choice had to be based on the SIPP proposal, but SIPP could not be bought lock, stock, and barrel. A substantial number of directorate members thought that the 64-bit addresses were too narrow, that they would not provide enough flexibility to implement proper routing protocols. There was a compromise in sight, if only the SIPP proponents would agree to increase the address size to 128 bits. The IETF decision process is based on building consensus, part of which implies polling opinion leaders. I was one of those leaders by this time, or so they thought, so they polled me. Would I accept inflating the address to 128 bits? I had publicly stated many times that 64 bits was more than enough and I could back my opinion with substantial mathematical analysis. However, I did not hesitate too long. Going to 128 bits was a little price to pay to reach consensus and save the Internet. Moreover, 128-bit addresses have a definite advantage. They have twice the width that routing procedures require, hence they make a lot of room avail-able for proper implementation of auto configuration procedures. "

I then go on to explain how auto configuration works by concatenation of a network prefix and a unique identifier of the host, normally a MAC address. The 1996 version of the book was written before the ND specifications was finalized, and does not actually specify the 64 bit prefix size. Many examples show an 80 bit prefix and a 48 bit MAC address.

-- Christian Huitema