Re: draft-bourbaki-6man-classless-ipv6-00

[Mark Andrews <[e-mail]>]

In message <[e-mail]>, "Manfredi, Albert E" writes:
> From: ipv6 mailto:[e-mail] On Behalf Of james woodyatt
>
> > This category of problem is about connecting *routers* (not hosts)
> > to the Internet via RFC 7278.
>
> Or hosts which operate like routers?
>
> > If all youre connecting at the hotspot are hosts, then why would you
> > need more than one subnet?
>
> My smartphone is assigned a single /64. It becomes a WiFi hotspot. So, de
> minimis, it creates a /65, for its WiFi interface, which any number of
> local WiFi users can use. Why not allow SLAAC for these WiFi users? And,
> why should we not allow, trivially, more flexibility? We now have these
> "whole home WiFi" systems. Why should that smartphone's /64 not be
> capable of supporting two or three local WiFi hotspots?
>
> Point being, all of this can be done with absolutely no consequences to
> the service provider, other than potentially some extra traffic volume on
> that one /64 facing the operator. But only if that 64-bit boundary is not
> fixed.

Operators that only give you a single /64 have already STOLEN 65535
subnets from you.  The architecture document say to give each site
a /48.  Addresses are handed to ISP's in the basis of /48's being
handed out to customers (you don't need to justify the space beyond
saying you are giving out /48s).  It doesn't cost more to route /48
than it does a /64.  These is less than a cent difference per
customer to get a /48 vs a /64 from the RIRs.  A cell phone is a
site border router the moment it turns on tethering.

> > And if you need more than one subnet for some premium mobile hotspot
> > thingie I guess why not use more than one UE?
>
> Of course you can use multiple phones with /64s, if that's what you're
> asking?  But not without going hat in hand to the operator.
>
> With IoT, example in cars, or in homes, no one can tell how many
> different subnets will make sense. If a home router is assigned a /64, it
> should be able to use that one /64 effectively, without constantly having
> to ask for more? And SLAAC should be available, for inside the home. Why
> not? Either that, or for sure, you will see NAT used with IPv6.
>
> Summary: I keep seeing people saying that they dont want to see NAT used
> with IPv6. In the next sentence, they say they want a fixed boundary at
> /64. The only way the two ideas can be compatible is if we have a prefix
> police that mandates some arbitrary max prefix length to be handed out,
> by operators, like /48. But operators keep saying no way, and they hand
> out /64s. Rock, hard place. (And, not to be ignored, then there's the
> hype of this limitless address space offered by 128-bit addresses, as if
> it's actually usable.)

More it is the MBA that thinks saving a couple of dollars on RIR
fees by only requesting just enough space to support a /64 per
customer is a good idea without looking at the entire picture.  Yes,
it adds up when there are millions of customers.  Charging 5 cents
extra per month would cover the additional costs many times over.

> Bert
>
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [e-mail]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: [phone]                 INTERNET: [e-mail]