Re: [bitcoin-dev] Safer NOINPUT with output tagging

[Anthony Towns <aj@erisian.com.au>]

On Mon, Dec 24, 2018 at 11:47:38AM +0000, ZmnSCPxj via bitcoin-dev wrote:
> A boutique protocol would reduce the number of existing onchain wallets that could be integrated in such UI.

Seems like PSBT would be a sufficient protocol:

 0) lightning node generates a PSBT for a new channel,
    with no inputs and a single output of the 2-of-2 address

 1) wallet funds the PSBT but doesn't sign it, adding a change address
    if necessary, and could combine with other tx's bustapay style

 2) lightning determines txid from PSBT, and creates update/settlement
    tx's for funding tx so funds can be recovered

 3) wallet signs and publishes the PSBT

 4) lightning sees tx on chain and channel is open

That's a bit more convoluted than "(0) lightning generates an address and
value, and creates NOINPUT update/settlement tx's for that address/value;
(1) wallet funds address to exactly that value; (2) lightning monitors
blockchain for payment to that address" of course.

But it avoids letting users get into the habit of passing NOINPUT
addresses around, or the risk of a user typo'ing the value and losing
money immediately, and it has the benefit that the wallet can tweak the
value if (eg) that avoids a change address or enhances privacy (iirc,
c-lightning tweaks payment values for that reason). If the channel's
closed cooperatively, it also avoids ever needing to publish a NOINPUT
sig (or NOINPUT tagged output).

Does that seem a fair trade off?

Cheers,
aj