> A use case I wanted to explore is where homenet router 1 has an Ethernet
> port connected to an Ethernet to powerline adaptor; elsewhere in the home
> is a powerline to Wi-Fi brick (pure AP, no routing), and then homenet
> router 2 has a Wi-Fi to Ethernet adaptor plugged in to its Ethernet
> port.
eth powerline wifi eth
HR1 ----- switch +++++++++ switch ...... switch ----- HR2
> The 2 routers each think they are using Ethernet to communicate with
> each other. If we consider algorithms like in Appendix A of RFC 6126
> (babel), where it's suggested that different cost computations be used
> for wireless and wired, the routers 1 and 2 would use a "wired" cost
> computation,
Yes. Babel would absolutely love to have a library that provides one bit
of information about each interface: "this is an Ethernet, don't bother
doing anything smart" or "be careful, this is something weird, disable
split horizon, perform link quality estimation, and assume worst-case
radio interference".
(Right now, we ask the kernel, and have a simple taxonomy of
wired/wireless/bridged. We'll get it wrong if there's a switch in the
way, of course.)
Of course, we would be able to use more information than that -- the
slowest Phy rate on the link and the exact set of radio frequencies being
used come to mind --, but we'd be perfectly happy with just the one bit
described above.
-- Juliusz
_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet
