Re: "class E" (was: Consensus Call: draft-weil-shared-transition-space-r

Subject: Re: "class E" (was: Consensus Call: 
draft-weil-shared-transition-space-request) Date: Wed, Dec 07, 2011 at 
12:19:49AM +1100 Quoting Mark Andrews (marka(_at_)isc(_dot_)org):
> In message <20111206055756(_dot_)GD20780(_at_)besserwisser(_dot_)org>, 
> =?utf-8?B?TcOlbnM=?= Nils
> son writes:
> > Subject: Re: "class E" (was: Consensus Call: draft-weil-shared-transition-s=
> > pace-request) Date: Tue, Dec 06, 2011 at 08:38:56AM +1100 Quoting Mark Andr=
> > ews (marka(_at_)isc(_dot_)org):
> > =20
> > > Ask everyone everywhere that is using this block, in good faith,
> > > for some purpose other than supporting addresses behind a CGN to
> > > renumber out of this block of RFC 1918 addresss which is now being
> > > re-purposed 16 years after it was allocated.
> > I do not understand why it is so hard to come to terms with the fact
> > that IF you have -- in whatever faith -- chosen to use non-unique address
> > space, you have been taking your chanches that sometime, in the future,
> > you WILL have to renumber to keep the illusion of quasi-uniqueness. This
> > goes for everybody. Customer, operator, middlebox or CPE vendor,
> > and my mother. This is inherent in all non-unique space. A new shared
> > allocation from the RIR pools or Class E will not change this fundamental
> > characteristic.  Therefore, 1918 space, being the prime example of
> > non-uniqueness, should be quite suited to populate the inside of a CGN.
> Actually it isn't inherent.  It's only if two of the parties involed
> are forced to use the same address pools that renumbering is inherent.
Oh, sorry, I meant the _risk_ (taking your chanches) of conflict as being
inherent. Unique addresses are meant to be collision-free, whereas any
allocation that might be made as a consequence of this draft can only
result in potentially conflicting allocations.

Thus, RFC 1918 is as [bad|good] as any other allocation. I furthermore
think that any recommendation about address space for CGN insides should
try to point out likely candidates (172.28/15, perhaps) but not do more
than that. Then broadband router vendors could stay away from that prefix
for their defaults (which they do; I've never seen, and anecdotal data[0]
confirms it, anything but 192.168/16 and 10/8.) and things would "work"
until the infrastructure and services are upgraded.

Further, I think, with personal experience from several 1918-using
enterprises[1], that the situation with "large organisation with all 1918
space used internally gets put behind CGN" is a hoax. Those people nearly
always have a guarded stash of old class C swampspace that they use for
things like DNS servers, firewall outsides, MX records, etc. Ususally,
they can come up with a /27 from that to number a bunch of p2p links.

-- 
Måns Nilsson     primary/secondary/besserwisser/machina
MN-1334-RIPE                             +46 705 989668
Xerox your lunch and file it under "sex offenders"!


[0] 
http://www.answersthatwork.com/Download_Area/ATW_Library/Networking/Network__4-List_of_default_Router_Admin_Passwords_and_IP_addresses.pdf

[1] here defined as largish organisation with dedicated IT staff and some 
cashflow.

signature.asc
Description: Digital signature

_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf