Michel Py wrote:
I realize now that this is where we erred: by shifting the multi-homing problem from the ISP to the end-user, we made a less-palatable protocol that is adoption-challenged.
Then shift it back to the ISP's [1]:
- Encourage RIR's to allocate /32's to large multihomed end sites (connected to 3 or more upstreams)
- Use something like alternate path encoded addresses/routing for trivial multihoming (2 upstreams)
Allocating anything longer than /32 is asking for a massive swamp.
It's bad enough that ARIN is issuing /48 microallocations as many operators are only filtering routes longer than /48 right now.
Alternate path routing allows you to set the bar a bit higher for getting PI space and allows anyone with two connections to multihome (including people who couldn't multihome in IPv4).
Kevin Loch
[1]http://www.ietf.org/internet-drafts/draft-loch-multi6-alternate-path-encoding-03.txt
"While many IPv6 multihoming methods require changes to host software, this method only requires changes to routers. This will make widespread implementation far more practical and likely in the near term than methods that require upgrading host software (provided that router vendors support this feature)."
_______________________________________________ Ietf@xxxxxxxx https://www1.ietf.org/mailman/listinfo/ietf