Re: [Isis-wg] Tsinghua work on source/destination routing

["Templin, Fred L" <[e-mail]>]

Hi Fred,

It is good to see this discussion, but an alternative approach that should
also be considered is tunneling. In the IRON approach at least, the end
user network gets a stable IPv6 prefix that is independent of the access
network IP addresses it gets from its ISPs. So, there is no need for source
address-based forwarding to ensure that packets sent via ISP A will not
have a source address from ISP B.

The use cases for tunneling are very broad, and probably overlap with the
ones you are considering in this approach. The relevant documents are here:

http://tools.ietf.org/html/draft-templin-ironbis
http://tools.ietf.org/html/draft-templin-intarea-vet
http://tools.ietf.org/html/draft-templin-intarea-seal
  
Thanks - Fred
[e-mail]

> -----Original Message-----
> From: [e-mail] [mailto:[e-mail]] On Behalf Of Fred Baker (fred)
> Sent: Thursday, November 07, 2013 8:45 AM
> To: Routing WG; [e-mail]; [e-mail]
> Cc: [e-mail] Group; [e-mail] WG
> Subject: [v6ops] Tsinghua work on source/destination routing
> 
> I'd like to draw your attention to a talk that will be given this morning in homenet. The context is:
> 
> http://datatracker.ietf.org/doc/draft-baker-rtgwg-src-dst-routing-use-cases
> http://tools.ietf.org/html/draft-baker-rtgwg-src-dst-routing-use-cases
>   "Requirements and Use Cases for Source/Destination Routing", Fred Baker,
>   2013-08-13
> 
> http://datatracker.ietf.org/doc/draft-xu-homenet-traffic-class
> http://tools.ietf.org/html/draft-xu-homenet-traffic-class
>   "Traffic Class Routing Protocol in Home Networks", Mingwei Xu, Shu Yang,
>   Jianping Wu, Fred Baker, 2013-10-21
> 
> http://datatracker.ietf.org/doc/draft-xu-homenet-twod-ip-routing
> http://tools.ietf.org/html/draft-xu-homenet-twod-ip-routing
>   "Two Dimensional-IP Routing Protocol in Home Networks", Mingwei Xu, Shu
>   Yang, Jianping Wu, Dan Wang, 2013-08-22
> 
> http://datatracker.ietf.org/doc/draft-baker-ipv6-ospf-dst-src-routing
> http://tools.ietf.org/html/draft-baker-ipv6-ospf-dst-src-routing
>   "IPv6 Source/Destination Routing using OSPFv3", Fred Baker, 2013-08-28
> 
> http://datatracker.ietf.org/doc/draft-ietf-ospf-ospfv3-lsa-extend
> http://tools.ietf.org/html/draft-ietf-ospf-ospfv3-lsa-extend
>   "OSPFv3 LSA Extendibility", Acee Lindem, Sina Mirtorabi, Abhay Roy, Fred
>   Baker, 2013-10-15
> 
> I had breakfast this morning with Shu Yang, who has been writing Quagga code for several years in the
> course of his PHd. He first implemented a source/destination model, reported on in draft-xu-homenet-
> twod-ip-routing, which was an MTR scheme. He tells me he found that very complex. He also listened to
> my talk in homenet around draft-baker-fun-routing-class, and has now implemented (if I understand him
> correctly) draft-ietf-ospf-ospfv3-lsa-extend and draft-baker-ipv6-ospf-dst-src-routing. The FIB
> implementation has a limitation: the source prefixes must be disjoint. However, given that, he has two
> FIB implementations, one of which has separate FIBs for each source prefix in play including ::/0 (so
> if there are M prefixes in the network, M+1 FIBs), and one of which is a single hierarchical M-Trie
> that looks up the destination and then the source. He has tested the code in simulation; the next step
> is testing in live networks.
> 
> Examples of use cases are generally around multi-prefix campus networks. There is a security use case
> that could be of value; at IETF 87, George Michaelson of APNIC reported on ULAs seen in his darknet.
> The short report is that he sees a fair bit of traffic with a ULA source address on the backbone. An
> interesting potential use of source/destination routing would counter that, and perhaps mitigate the
> need for ISP BCP 38 if generally deployed; in a case where a network is using a ULA and a global
> prefix (e.g., is not multihomed but has two prefixes, one of which is intended to only be used within
> its network), the default route to the network egress would use the global prefix as a source, and as
> a result traffic sent outside the network with a ULA source prefix would in effect have no route. The
> network could literally only emit traffic from its correct prefix.
> 
> I think this is relevant to the discussion of
> 	draft-baker-rtgwg-src-dst-routing-use-cases
> 	draft-ietf-ospf-ospfv3-lsa-extend
> 	draft-baker-ipv6-ospf-dst-src-routing
> 	draft-baker-ipv6-isis-dst-src-routing