Re: [IPv6] SADR for MHMP is useless yet

[Ole Troan <[e-mail]>]

Eduard,

Two border routers and a stub router is the scenario presented in snac and that triggered the discussion about updating rfc7084. 

Up to you if you “believe” in the use case or not of course. But that’s a use case that would require SADR.

(Personally I think it’s highly unlikely that we will get MPMH to work.)

O. 

On 21 Nov 2022, at 15:58, Vasilenko Eduard <[e-mail]> wrote:



Hi Ole,

 

Let’s define “the simple topology”. It means that any host sees both MHMP routers on one link.

SADR would not apply here. The host should properly choose “the default router” at ND.

 

Let’s define “the more complex topology” that has 2 routers in a short chain: 1^st router is connected directly to Carrier, 2^nd the router is connected to 1^st router (and host after this).

For this topology to start working with PA addresses should be some protocol on routers to split and re-distribute prefixes received from FBB Carrier. It does not exist yet (HNCP is not adopted). Hence, the second router does not have PA addresses. Hence, no need for SADR yet.

 

In general, PA prefixes need to be automatically propagated deeper (at least one additional hop) in the site for SADR to become needed.

SADR should follow HNCP, not vice versa. No HNCP -> no SADR, it is simple.

Hence, https://www.rfc-editor.org/rfc/rfc7084" rel="nofollow"> RFC 8678 is completely useless yet. We could not have a situation.

Of course, the lab can appoint PA prefixes manually to test that https://www.rfc-editor.org/rfc/rfc7084" rel="nofollow">RFC 8678 is right.

It is not relevant for any production network yet.

 

Additionally, It is possible to mention that Mobile carrier is not capable to deliver prefix in general.
DHCP is blocked at the modem level on all smartphones.

For the mobile environment, the problem is double protected.

 

Hence, no need yet for the SADR. It is a far future, if ever.

 

Eduard

From: Ole Troan [mailto:[e-mail]]
Sent: Monday, November 21, 2022 4:57 PM
To: Vasilenko Eduard <[e-mail]>
Cc: [e-mail]
Subject: Re: SADR for MHMP is useless yet

 

Eduard,

 

What was discussed here was a topology with two independent border routers and a stub router attached to them, with a stub link southbound. 

 

The stub router has to, as far as I can tell, do SADR.

(Or other similar policy approaches to routing)

 

Do you see it differently?

 

O. 

On 21 Nov 2022, at 14:32, Vasilenko Eduard <[e-mail]> wrote:



Hi Ole,

 

SADR is not important yet. Maybe in the far future (5+) it would become important.

It is useless for the PA case yet because it is not possible to come to this situation.

How PA prefix would be split and distributed between many routers and links on the complex site?

(something like HNCP or DHCP-PD). Automation here is mandatory, PA could be dynamic.

The need for SADR would happen only after this.

 

SADR is not needed in principle for other use cases: PI, NAT+ULA, NPT+ULA. Normal destanation-based routing is enough.

 

Eduard

 

Ole Troan <[e-mail]> Fri, 18 November 2022 14:52 UTC

Ted,

 

There are a few reasons why MHMP have failed. 

 

1) it requires SADR routing in the intermediate routers. SNAC routers in young case

2) It leaves the choice of egress path to the hosts. Expecting every host/application to be smart enough to handle that correctly especially given the hosts on a SNAC link sounds optimistic. 

3) Hosts selecting exit is often at odds with network policy.

 

PVDs doesn’t seem to be particularly important to the problem, but I don’t mind if you think it is.

Great if you can make MHMP happen.

It might be a distraction for this group and not necessarily something that can be limited to v6ops. 

 

Anyway, I think this is something for the chairs/ADs. 

O.