Rusty Russell

This isn't going to get fixed for 0.12, sorry. I recommend you manually parse $HOME/.lightning/config and $HOME/.lightning//config if they exist, and parse any lines starting with `peerswap-`. Use those results...

> So that's an argument to make them default-on, or even non-optional, to minimize the potential configuration combinations that end users may end up with. Yes. All optional features *will*...

We can't have both "we want people to start using this feature" and "sorry, there's no API stability"; they are opposed in practice. > I think having a feature that's...

@gmaxwell Actually, libsecp is becoming widely used; every lightning implementation uses it, for example. And the schnorr stuff is likely to complete this dominance, since it's just Damn Nice for...

> The way I see x-only pubkeys is that this is mostly an optimization when putting keys on chain. I see things differently from you: compressed pubkeys are a vestigial...

> Hm, I don't think it's strange. So far, we use 33 bytes pubkeys but hash 64 bytes. Is that strange? I guess not. BTW, I think this is wrong:...

> > As a simple crypto API user, 32-byte pubkeys are a minor but clear API improvement. They align nicely, they index ~uniformly, and I expect with encouragement and tooling...

These are not gossip messages, this looks like a complete protocol message index? Anyway, yes, we should enhance the tools/ to generate it!

This fix seems correct, but I'm reminded of Kernighan and Plauger's aphorism "Don't patch bad code, rewrite it". Could we see a third commit that cleans all this up? In...

Slightly neater: ``` diff --git a/gossipd/routing.c b/gossipd/routing.c index 6cf0d7b69..f85cff821 100644 --- a/gossipd/routing.c +++ b/gossipd/routing.c @@ -1409,9 +1409,11 @@ bool routing_add_channel_update(struct routing_state *rstate, return true; } - /* Make sure it's...