Subject: Re: ipv6 over PPP
To: Robert Elz <kre@munnari.OZ.AU>
From: None <>
List: tech-net
Date: 09/25/2003 11:58:18
On Thu, Sep 25, 2003 at 04:37:52PM +0700, Robert Elz wrote:
>     Date:        Thu, 25 Sep 2003 11:16:18 +0200
>     From:
>     Message-ID:  <>
>   | I use IPv6 over PPP here (great ISP).
> Is your basic model of operations the same as the one I described?

Mostly. My internet link is dual IPv6/IPv4, and my home network grows
from there.

> That is, how do you configure your PPP client?
>   | Shouldn't the router use a static configuration?
> How?   From where does it determine its IPv6 address?   It will only
> discover that once it gets the PPP connection.   Sure, while I am only
> ever connecting to one destination (as people who use PPP over hard
> wired links do) I'm always going to get the same address - or in my
> situation I can force that, but I don't want to be constrained to
> only dialing one provider - I want to be able to dial anyone, and
> if they support IPv6, I want them to tell me what addresses I should
> use.

You have a point, my configuration is purely static. I have parts of
my /48 block assigned all over my network, thus I'm tied to my ISP.

>   | I really think this is the point here. You seem to want to push the
>   | auto-configuration to a very high level.
> Only as high as it works (using totally different mechanisms admittedly)
> using IPv4.
> With IPv4, all this works using NAT - the ISP tells me what IPv4 address
> to use, then my ppp client (router) uses that address, and arranges for
> all the other systems to look like they're also using the same address.
> Of course, I could do NAT for IPv6, and do everything the same way, but
> much of the point of IPv6 is supposed to be to avoid that.   Hence, I
> want something for IPv6 that gives equivalent functionality.

Yes, the problem is quite different in IPv4, because of NAT. For the ISP,
NAT means "do whatever you want, I don't care". With IPv6, it can't do

What you're lacking indeed is a way to obtain this information through
LCP, and I think implementing anything beyond this will be a hack and
will be obsoleted once anything gets standardized. It's true that in
the meantime, we're pretty much tied to static setups.

Quentin Garnier.