Chris,

> So here is my summary of the situation.
> 
> 1) If DHCPv6 with extensions in RFC7078 is used to help the host to select 
> the correct source address, then we could handle even walled garden use cases 
> with default-only SADR.  See details for how this works in 
> http://www.ietf.org/mail-archive/web/rtgwg/current/msg05472.html .  Another 
> assumption would be that DHCP relay could be used so we wouldn't need the 
> (D!=::/0 , S) routes in to reach R7 and R8.
> 
> 2) However, if we don't use DHCPv6 with RFC7078 for host source address 
> selection, then Neighbor Discover Router Advertisements need to be extended 
> using a new mechanism like draft-sarikaya-6man-sadr-ra-03 or 
> draft-pfister-6man-sadr-ra-01.  In this case, default-only SADR is not 
> sufficient to support walled gardens because we are relying on the (D!=::/0 , 
> S) routes to be carried in the IGP so that R7 and R8 can then put the 
> (D!=::/0 , S) information into the new ND Router Advertisements.
> 
> Is this more or less accurate?

Yes, that's reasonably accurate. Although of course not the full picture since 
we have decades of innovation in this space.
- You also have ICMPv6 Destination Unreachable, Source address failed 
ingress/egress policy that could be used as a trigger by the host to try a 
different source, independent of what the network signals.
- Or MP-TCP which presumably would also try some subset of all combinations.
- Or draft-baker-happier-eyeballs-00 which throws spaghetti on the wall to see 
what sticks.

Cheers,
Ole


> -----Original Message-----
> From: [email protected] [mailto:[email protected]]
> Sent: Tuesday, April 12, 2016 2:37 PM
> To: Chris Bowers <[email protected]>
> Cc: David Lamparter <[email protected]>; [email protected]
> Subject: Re: Implications of default-only SADR (was: Re: multi-homing for 
> provider-assigned IPv6 addresses)
> 
> Chris,
> 
>> Consider the topology below.  See 
>> https://www.ietf.org/mail-archive/web/rtgwg/current/msg05472.html for a more 
>> detailed description of the topology.  For H31 to send a packet to 
>> destination B3, H31 must choose a source address from within subnet A3x.
> 
> [very nice ASCII art ruined by my stupid MUA]
> 
>> For this example, we assume that the R1-R4 originate the following (D,S) 
>> routes in the IGP.
>> R1 originates a route for (D=::/0, S=A1).
>> R2 originates a route for (D=B2, S=A2).
>> R3 originates routes for (D=::/0, S=A3) and (D=B3, S=A3).
>> R4 originates routes for (D=::/0, S=A4) and (D=B4, S=A4).
>> 
>> R7 and R8 receive these routes via the IGP.  With the existing mechanisms in 
>> Neighbor Discovery Router Advertisements, R7 and R8 can advertise the 
>> following PIOs and RIOs.
>> PIOs = A4x, A2x, A1x, A3x
>> RIOs = B2, B3, B4, B1
>> 
>> I have intentionally changed the order of the prefixes in the set of PIOs 
>> and RIOs to emphasize that there is no required ordering or relationship 
>> between prefixes in PIOs and RIOs.
>> 
>> With only this information, I do not see how H31 can correctly choose a 
>> source address in A3x when it needs to send a packet with destination 
>> address B3. If this analysis is correct, then it seems like a mechanism like 
>> draft-sarikaya-6man-sadr-ra-03 is needed.
> 
> I take it B1, B2, B3 and B4 are walled gardens.
> 
> What I think you are suggesting is that the host could use a S,D FIB for 
> source address selection. At least type D hosts could. 
> https://tools.ietf.org/html/draft-pfister-6man-sadr-ra-01
> 
> RFC7078 is meant to solve this case though.
> 
> Best regards,
> Ole

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

_______________________________________________
rtgwg mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/rtgwg

Reply via email to