[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: radical solutions



At 4:33 PM +0100 2/14/01, Sean Doran wrote:
>v6's addressing format initially implied routing that -was- radically
>different from v4.   

Sean,

Are you talking about the early proposals for metro-based addressing
and routing?  If not (i.e., if you are just talking about the
"aggregatable global unicast" address format), then I think this is
another one of those persistent misunderstandings.  It was always just
the same old routing protocols, building up FIBs of variable-length
prefixes on which routers perform longest-prefix match to decide which
way to send a packet.  Indeed, the fact that IPv6 stuck with the same
old routing approach, with all of its known problems, is an often-heard
criticism of IPv6.

The field boundaries in the IPv6 Address Architecture spec (with
the exception of the /64 boundary) are/were allocation-policy boundaries,
not boundaries that ought to be known to, or constraining on, routers,
routing protocols, or the inter-connectivity of ISPs.  Yes, one can well
question what allocation policy is doing in an "architecture" spec (it
is a legacy of the GSE proposal), and I believe we've decided to update
it to remove the field width values.  Regardless, subsequent WG documents
(e.g., the 6Bone address format and the recommendation to the RIRs)
should have made it clearer that the structure of the left-hand-side of
IPv6 addresses, and IPv6 routing, is not and was not intended to be
radically different than that of IPv4.

Anyway, I can see how you leapt to the conclusions you did, but I believe
they were incorrect conclusions (but maybe I misunderstand what you
concluded).  In any case, this is all water under the bridge at this
point, and not worth belaboring further on this list, since we are
supposed to be looking forward to how to make things work well,
regardless of what has been proposed, or (mis)assumed to have been
proposed, in the past.

Steve