[ekzyis]

Based on #303

A channel is tracked from the moment we build its `funding_created`, so
`ChannelReadyOracle` accepted a `channel_ready` from a target that never
answered with a `funding_signed`. Without that signature we hold no valid
commitment transaction, so a channel the target treats as ready leaves our
funding output spendable only with its cooperation.

BOLT 2 mentions nodes MUST wait for `funding_signed` before sending
`shutdown` but not for `channel_ready`. This might have been an
oversight.

The two-channel `channel_ready` programs skipped `funding_signed`, so they
now await one for each channel.

[morehouse]

Looks good; will need rebase once #303 is updated.

[ekzyis]

ebed335: addressed feedback + rebase on master (74d7514)

Do you think the lightning developers will be interested in a PR to mention this:

MUST wait for funding_signed before sending out channel_ready

It's an issue with funds at risk for the peer, though, not for the misbehaving node. Maybe SHOULD instead of MUST?

[morehouse]

Do you think the lightning developers will be interested in a PR to mention this:

MUST wait for funding_signed before sending out channel_ready

It's an issue with funds at risk for the peer, though, not for the misbehaving node. Maybe SHOULD instead of MUST?

The spec already says not to broadcast the transaction until funding_signed is received. And would someone really send channel_ready before they even broadcast?

There's probably a gap that could be closed here, and it would probably get merged if the change is small, but I'm not sure it will really make a difference for anyone implementing the spec.

[morehouse]

LGTM