Re: draft-wbeebee-ipv6-cpe-router-04 comments

[Antonio Querubin <tony@lava.net>]

On Thu, 26 Mar 2009, Stark, Barbara wrote:

> That's an interesting use case, with definite application, but I wonder 
> if it's really compelling in the case of the simple CPE Router.
>
> When CPE Routers are cascaded, the cascaded router is generally directly 
> connected to the upstream router, so that there would be no added 
> efficiencies from this.

I don't think it's that uncommon.  I've seen households, small businesses, 
churches, etc that have multiple routers/switches/WAP devices scattered 
throughout the home or their buildings/offices.  While these types of 
deployments do not represent a very large percentage of how these devices 
are being used, I think we'll see more and more of these devices being 
cascaded together just for the addition of ports or access points.  The 
cost difference between a switch-only device vs a router is so minimal 
that many consumers opt for the latter even though all they really need is 
another switch in most cases.  Few of these entities have IT staff to 
advise them otherwise, nevermind efficiency.  Last time I browsed the 
network section of a computer store, I saw as many wireless devices with 
built-in routers vs just standalone WAPs/switches.  Same for the 
firewall/NAT devices.

Changing the topic slightly, it seems to me that one of the big challenges 
for the CPER is determining how the prefix sub-delegation behaviour should 
be when you do have cascaded devices.  Current IPv4 devices just avoid the 
issue by doing NAT upon NAT.  If IPv6 frees users from limits on addresses 
and subnets then the CPER behaviour should be designed to automatically 
handle the partitioning of the SP-delegated /56 or /48 among 2 or more 
downstream CPERs.  Should we try to spell out a recommended behaviour in 
more detail or are we gonna leave that up to vendor implementation?

Antonio Querubin
whois:  AQ7-ARIN