> On Thu, 04 Oct 2007 19:40:50 -0400 > Brian Dickson <[EMAIL PROTECTED]> wrote: > >> (Please excuse minor formatting issues and the occasional spelling >> mistake - this is version 00 of the draft.) >> >> This I-D is the one you've all been waiting for, after the "16 bits >> between friends" thread. >> > > Couple of quick things from a first glance. I understand Rick Adams is > considered the founder of UUNET, not Mike O'Dell.
Thanks - my memory has apparently failed me in that regard. :-) It'll get fixed in the next draft. I'll save up a bunch of changes before modifying it, though, so there's not too much confusion over versions of the draft. > You might also want to consider whether you want to quote Mike, > as from what I understand his GSE proposal was the one that suggested > that all node IDs should be the same length universally, separate from > variable length prefixes in the routing part of the address. > > A internet search has dug up what appears to be a version of that proposal > - > > http://arneill-py.sacramento.ca.us/ipv6mh/draft-ipng-gseaddr-00.txt Actually, I think that's more reason to quote him. The quote is good, and not everything anyone comes up with is necessarily going to seem as prophetic after 10 year's time. The interesting thing in the GSE proposal, was that at the time, IEEE-1394 was seen as a potential take-off point for new devices and addresses. The large uptake in those devices is, however, largely host-connected peripherals, rather than network devices - similar to USB in that regard. I think that re-reading the GSE proposal, with the added perspective that we have of 10 years' more experience, combined with more recent experiences and context within which the requirements can be viewed, lends credence to moving the boundary for the autoconf (host) part of the V6 addresses. Since we still see the majority of the non-mobile network access devices as being connected with Ethernet of various flavours, having the ability to take advantage of that in the numbering, addressing, and allocation & aggregation, is merely prudent. I don't actually disagree with the notion of fixed length II's, if the length chosen is 48 bits. But given the deployed code and connected networks and hosts, backward compatibility for those exiting sites and hosts (for 64 bits) is a compromise that I think is reasonable. If folks would rather only see one II size, then we need to give serious consideration to changing it universally to 48 bits. Brian > Regards, > Mark. > > >> Enjoy, >> >> Brian Dickson >> > -------------------------------------------------------------------- IETF IPv6 working group mailing list [email protected] Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------
