From aj at erisian.com.au  Mon Nov 30 01:25:54 2015
From: aj at erisian.com.au (Anthony Towns)
Date: Mon, 30 Nov 2015 11:25:54 +1000
Subject: [Lightning-dev] Return to the Layered Commit Transactions?
In-Reply-To: <20151127212742.GA17286@sapphire.erisian.com.au>
References: <20151125090754.GA29895@debian> <878u5krtts.fsf@rustcorp.com.au>
	<20151127051446.GA4583@sapphire.erisian.com.au>
	<20151127212742.GA17286@sapphire.erisian.com.au>
Message-ID: <20151130012554.GA22914@sapphire.erisian.com.au>

On Sat, Nov 28, 2015 at 07:27:42AM +1000, Anthony Towns wrote:
> (You might want to add further constraints on the timeout based on fees,
> of course -- the above is what you need to avoid being left holding
> the bag)

You might also want to batch the timeouts, so if you have 1000 concurrent
HTLCs on a channel at any point in time [0] and expect 10% of them
to timeout, you'd be better off expiring them in five batches of 20,
rather than individually.

That way you can spend them all to a single output, saving you both
19*34 bytes in that transaction, but also something like 19*130 bytes
when you want to actually spend the outputs. (With 40 satoshi/byte fees @
$350/BTC, that's an overall saving of 2c or more per batched HTLC)

Cheers,
aj

[0] I think the 100kB standardness limit on transactions means you can
    have at most about 3000 concurrent committed HTLCs on a channel,
    assuming your anchor is a single multisig transaction, and your
    outputs are all P2SH.