Re: [v6ops] PI heresy [draft-ietf-v6ops-ula-usage-recommendations - work or abandon?]

[Philip Homburg <[e-mail]>]

>> I don't think a host should try to figure out which paths work and which
>> don't.
>> 
>> Maybe somebody can come up with a simple algorithm that generalises
>> happy-eyeballs and that is compatible with today's host operating systems.
>> But I doubt it. 
>
>We tried. It's called shim6. Firewalls break it.

I'd call shim6 a protocol and not an algorithm. Happy eyeballs is quite 
compatible with firewalls, NAT, etc. No problem there. I'm sure happy
eyeballs-like approaches can be tried for the multiple-prefix case.

The problem I ran into is that by the time my happy eyeballs implementation
did everything I wanted it do, it was way to complex. Adding support for
many source prefixes would just be too much.

Now, maybe somebody can come up with an approach that doesn't have that
problem. Don't know.

Mean while, I have something that works with every operating system today.
It may not be perfect. So I'm not terribly motivated to add more host
complexity to solve those edge cases.