[Babel-users] Fwd: [PATCH net-next v3 00/19] Multipath TCP part 2: Single subflow & RFC8684 support

[Dave Taht]

Juliusz Chroboczek <jch at irif.fr> writes:

>>> and wow, mptcp is finally merged. Anyone for source specific routing?
>
> Pro memoria, using source-specific routing with Babel was described by
> Matthieu many years ago:
>
>   http://blog.multipath-tcp.org/blog/html/2015/01/19/babel.html

Happy fifth anniversary.

While mosh did finally gain ipv6 support 4 years back I'm under the
impression the mosh multipath idea didn't make it (?)

I'm totally out of touch also on source address selection rules. In the
case of 5g + wifi, in general you want a primary address to be on the
(cheaper) wifi connection and to just keep the other alive.

Have any sane solutions to this appeared?

Somewhat along the same vein

wifi: 2001:a:b:c::1, :2, :3, :4
5g: 2007:a:b:c::1, :2, :3 etc

connecting to

2006::a:b:c::1

my "ideal" case would be to try the first wifi and the first 5g (on a
privacy address), rather than cycle through all the addresses in a
longest prefix match, and to still have some std way of expressing from the
router or to the application which to prefer.



>
>> Don't open the bottle of champagne too fast...
>
> You've changed, Benjamin, I recall at one time you were quite willing to
> open bottles :-)
>
>> I heard it is only partial...
>
> Yes, from the description it looks like it's partial -- it will establish
> an MP-TCP connection, but won't actively establish any new subflows.
> Still, it's an important step -- it implies at least agreement in
> principle from Miller, which is a big deal.

yep. one more release, I'd hope... but more likely 2-3.

>
> -- Juliusz
>
> _______________________________________________
> Babel-users mailing list
> Babel-users at alioth-lists.debian.net
> https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/babel-users