I added hybrid category to manually configured border discovery interface types 
(MAY):

https://github.com/fingon/ietf-drafts/commit/0fb83030fd2c2b8d933417aaca2409a043320713

It is usable _within_ home, if (for example) CER is (legacy) RFC7084, and two 
HNCP routers are plugged to it. By default, they are unable to learn of each 
other (and do the HNCP services with each other such as prefix assignment, 
service discovery, and so on). IPv6 on them will probably require some manual 
configuration and-or RFC7084bis behavior on part of the CER.

I’m not sure this is still a good idea, as there’d be duplicate prefixes 
retrieved from legacy router (if 7084bis), but at least it would enable HNCP 
router 1 -> HNCP router 2 clients to communicate on IP level, and to discover 
services. 

Another thing I'm wondering about is how to handle services that do _not_ need 
to have any explicit state. Should there be some sort of feature 
announcement(/support) mechanism? A concrete example: PCP or UPnP IGDv2 proxy 
on inner router needs to know _only_ about the CERs. It does not need to know 
anything more, and it can _try_ to use the protocol as-is (as almost mandated 
by PCP proxy discussions currently, I guess), but given extra information about 
support for the feature on the CER, the response to the host requesting 
services from a router that does not support it could be better. Instead of 
effectively blackholing packet for service that the CER does not support, proxy 
on the first-hop inner router could provide (possibly) more useful error 
response immediately.

_Currently_ we implemented PCP proxying just blindly, hoping the CER does it, 
and otherwise not returning any error. As in case of multi-homing the error 
state would depend on the source address chosen by the client in any case, I'm 
not sure non-source-address-specific error messages would bring any extra value 
here anyway. For example, in case of PCP, returning (e.g.) NO_RESOURCES may not 
make client try different source address.

Cheers,

-Markus

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

Reply via email to