IPv6 Default Allocation - What size allocation are you giving out

Thu Oct 9 21:13:16 UTC 2014

Hi Bill

Thanks for you response. About customer routers:

For IPv6 that answer is simple. The customer is using us as default gateway
and that always uses the IPv6 link local address. He has no need to know
the public IPv6 address of the uplink router, so we don't tell him. The
link local address is learned automatically from the RA packets.

The customer router needs an IP address. I do that by allocating a small
prefix, typically a /120, which covers all the users on the same access
switch. The IP, a /128, is assigned by DHCPv6 and he gets his /48 by prefix
delegation. There is no way to avoid a route for that /48.

This works great with asymmetric bridges (isolated vlans etc).

For IPv4 I do have an IP address on the customer facing interface.
Typically a /24 for users on the same access switch using MFF (MAC Forced
Forwarding). I do wonder if I could get away with using a /32 and push out
a host route through DHCP, but I am unsure if clients generally support
that.

But all this are customer facing interfaces, which do not really qualify
for "point to point" links. I might consider adding interface addressing
for IPv6, but for me IPv4 was the primary design parameter. Having IPv6
mirror the IPv4 setup means I have to think less about the setup. And we
are really constrained to use as few IPv4 addresses as possible. We only
got 1024 from RIPE and have to buy any additional at great expense.

My colleges wanted to completely drop using public IP addressing in the
infrastructure. I am wondering if all the nay sayers would not agree that
is it better to have a single public loopback address shared between all my
interfaces, than to go with private addressing completely? Because frankly,
that is the alternative.

Regards,

Baldur

On 9 October 2014 22:49, William Herrin <bill at herrin.us> wrote:

> On Thu, Oct 9, 2014 at 4:32 PM, Roland Dobbins <rdobbins at arbor.net> wrote:
> >
> > On Oct 10, 2014, at 3:25 AM, Baldur Norddahl <baldur.norddahl at gmail.com>
> wrote:
> >
> >> I am sure there are. Tell me about them.
> >
> > This issue has been discussed on all the various operational lists many,
> many times over the years.
> >
> > <http://tools.ietf.org/html/rfc6752>
>
>
> Hi Roland,
>
> 6752 isn't germane; it has to do with using private IP addresses on
> routers, which borks things up when the router has to generate an ICMP
> type 3. Baldur want's to know: why not just use one public IP address
> per router and use it on all interfaces?
>
> Baldur, one IP per router can work just as well as one subnet per
> interface. But there are some gotchas:
>
> Your router has one IP. Your customer has a subnet. Do you add an
> extra deaggregated single IP to your routing table for his router?
> There are more routers than links, so if you assign subnets to routers
> instead of links you'll have to carry more routes.
>
> If you borrow the customer LAN-side IP for the WAN side you'll get
> grief when his equipment is one of those that doesn't respond if the
> LAN-side interface is down (e.g. Cisco). That gets kind of nasty when
> troubleshooting and remediating problems.
>
> And of course the more knowledge you can gather from diagnostic tools
> like traceroute, the more quickly you can identify the problem when
> something doesn't work right..
>
>
> In my own networks... I want to keep as many IPv4 addresses as I can,
> so my router interfaces borrow their ip from loop0. In IPv6 where I
> can have a functionally infinite number of /124's I want to put one on
> each interface and gain the mild extra benefit.
>
> Regards,
> Bill Herrin
>
>
> --
> William Herrin ................ herrin at dirtside.com  bill at herrin.us
> Owner, Dirtside Systems ......... Web: <http://www.dirtside.com/>
> May I solve your unusual networking challenges?
>