On Mon, 29 Mar 2010 17:34:30 -0500
"Frank Bulk - iName.com" <[email protected]> wrote:

> Did you get assigned a /48 that you feel that you are address constrained?
> 

No.

> The whole point of IPv6 is plentiful addresses.  At this time I'm planning
> to assign a /64 per PPPoE link, and if they want it, a /56 via DHCP-PD.
> 

I'm happy with using /64s for PPPoE links. However, if the /127 draft is
accepted, then I'd want to be able to take advantage of them on
PPP/PPPoE sessions - if there is an approved mechanism available to
save address space I may as well use it.

While RAs do support prefix lengths in their PI option, there isn't a
mechanism specified on how to do autoconf in a /127 (or rather, who
gets the 0 or 1 address). As DHCPv6 is going to be run over that link
anyway to acquire DNS addresses etc., it might be better to build
the /127s autoconf mechanism into that. This new DHCPv6 option could
also imply a default route/router, which the eliminates the requirement
for the upstream BRAS to generate periodic, per-PPPoE session/prefix 
specific RA announcements. That could be a useful thing to eliminate, as
generating pre-PPPoE session specific RAs (due to different prefixes),
could start to be a significant control plane load when there are
multiple 10s of 000s of PPPoE subscribers attached to the same device.

Regards,
Mark.


> Frank
> 
> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On Behalf Of Mark
> Smith
> Sent: Monday, March 29, 2010 4:47 PM
> To: Mark Smith
> Cc: Becca Nitzan; [email protected]; [email protected]; George, Wes E [NTK];
> [email protected]; Miyakawa; [email protected]
> Subject: DHCPv6 support for /127s, for ISP subscriber PPP/PPPoE p2p links
> (Re: router vs. host discussion in 6man today for the /127 draft)
> 
> Hi,
> 
> The subject pretty much says it. It's extremely wasteful to be
> allocating a /64 per subscriber PPP/PPPoE session. 
> 
> An alternative model is lay a "virtual" /64 over the top of the 100s or
> 1000s of PPP/PPPoE sessions, and have the subscriber's PPP IID used to
> autoconf the LL and global addresses. However that's vulnerable to the
> /64 issues, and because IPv6 over PPP defaults to allowing disabling
> DAD, could also suffer from duplicate addresses. There is a caveat to
> allow DAD to be re-enabled, but with cheap residential CPE, that
> couldn't be relied on.
> 
> Also it'd be worth making DHCPv6 support announcing a default
> router/route to these end nodes, so that the upstream BRAS doesn't have
> to issue periodic RAs. In either the /64 per P2P link or the single
> "shared" model, with 100s (or more typically 1000s) of subscribers
> attached, the router will have to spend a reasonable amount of
> resources on sending RAs down the individual PPP/PPPoE sessions.
> 
> Regards,
> Mark.
> --------------------------------------------------------------------
> IETF IPv6 working group mailing list
> [email protected]
> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> --------------------------------------------------------------------
> 
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to