Folks,

The scope discussion is flawed, held in the wrong forum and should cease.

The charter -- according to the web page -- says: "The primary focus
of the IPv6 w.g. is to complete the standardization of the IPv6
protocols." Consequently, the wg is an Internet Area working group;
*not* an Ops WG.

We have reached (albeit rough) consensus that the scoped addresses
are to be limited to link-local only.

But, as far as I can tell, the scope discussion has not terminated,
but instead ended up in a swamp where implementors and protocol
architects are trying to teach operators how to run their networks,
by inventing useless management complications that neither will
contribute to the simple, smooth operation of an IPv6 Internet, nor
assist in securing hosts against evil-minded attacks.

Let's look at the some of the core arguments:

* "Non-routable prefixes are inherently safe". Perhaps. One could rewrite
   this to: "Non-routed prefixes are inherently safe". If I route a part of
   my /48 only in my IGP, and blackhole it in my border routers, I have
   created a non-routed prefix, as long as my border devices can throw
   packets. (There is overwhelming operational experience that says
   "routers can drop packets".) In terms of "scope" this so treated
   global prefix chunk walks like a site-local, talks like a site-local,
   is safe like a site-local[0] but lacks the need for extra scope-
   checking code.

* "Renumbering is hard". Well, stop whining and help work on the
  renumbering drafts instead. I have not renumbered any v6 networks yet,
  but have done a fair bit of v4 network restruction. To me, the solution
  lies in abandoning the identifier overloading that takes place when
  people configure applications to use IP addresses directly, and
  instead apply suitable layers of abstraction. Allowing people to
  preserve the overloading by making it "convenient" to keep the address
  for long times is a step in the wrong direction.

To me, this looks like material for an operational discussion, that
should result in two BCP documents, "Practices for controlled
limitation of node  reachability in IP networks" and "IPv6 Network
address plan design with renumbering in mind.", none of which look
like Internet area documents but instead like Ops stuff.

May I humbly suggest that the people so greatly concerned with how
networks are to be operated go and write these BCPen in an Ops
group, and leave the crippling featurism out of the IP protocol.

Best regards,
--
M�ns Nilsson            Systems Specialist
+46 70 681 7204         KTHNOC  MN1334-RIPE

We're sysadmins. To us, data is a protocol-overhead.

[0] Maybe. I help run a pretty large multi-AS network, edge and
core mixed, the uses more and more IPv6. We have tried, but can't
come up with any reasons for inherently crippled prefixes (we are
at times way too good at crippling the useful ones ourselves...),
so we can't really tell whether they are useful, they just do not
seem so.

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to