[Lightning-dev] Lightning fees vs miner fees

Anthony Towns aj at erisian.com.au
Wed Oct 28 02:13:59 UTC 2015


On Tue, Oct 27, 2015 at 10:34:12PM +0100, CJP wrote:
> > Alternatively, maybe it means that
> > miners aren't an effective cartel, and thus can't force fees to
> > rise to maximise revenue; that doesn't seem likely to me though...
> Why not?

Trivially, if a majority of hashpower wanted to act as a cartel, they
could just use the soft-fork protocol to make and enforce any decisions
(eg, versionbits to vote on decisions, then orphan any non-compliant
blocks to enforce it).

> So, how do your calculations work out if the miners are not a cartel,
> but instead completely non-organized? 

I don't see how to come up with a figure for fees or miner revenue other
than than maximising revenue for miners (ie, the same fees they'd charge
if they were a cartel) or minimising costs for users (ie fees at 0 or 1
satoshi per tx). The supply constraints for transactions per block aren't
really tied to miners' costs, so supply capacity doesn't particularly
increase with costs, and so supply/demand curves don't help much.

If you choose somewhere in between those extremes, that might mean the
blockchain gets used for transactions above $20 or $100 rather than only
above $200; but given we're talking about lightning payments being below
about $10 each, that seems fine.

> What if, in the style of the
> prisoners dilemma / tragedy of the commons, each miners' actions are
> targeted towards maximizing its own profits, even if that comes with a
> disproportional cost for other miners?

Hmm. Having miners accept txs with different fees introduces a timing
aspect -- ie, it's not just "is it worthwhile to pay Bob $5 if it costs
45c?" it's "is it worth paying Bob $5 in 10 minutes for 45c; or in 60
minutes for 20c?"

Once you add that complication to the analysis I think you have to figure
a cartel would take it into account too. If every 10 minutes you had a
new set of transactions from people with the following willingness to pay:

  1000 tx worth 100c in 10 minutes or 20c in 60 minutes
   500 tx worth  90c in 10 minutes or 20c in 60 minutes
   200 tx worth  30c in 10 minutes or 25c in 60 minutes
   100 tx worth  20c in 10 minutes or 20c in 60 minutes

Then I think you'd optimise your revenue by accepting any transactions at
90c or more immediately, along with any transactions that are at least
50 minutes old at 25c or more. Or you could just include 17% of 25c txs
from the mempool in each block for roughly the same effect. That would
make you around $1400 in fees, and does better than:

  $1350 in fees if you just accept 90c txs asap
  $1000 in fees if you just accept 100c txs asap
   $510 in fees if you just accept 30c txs asap
   $360 in fees if you accept 20c txs at anytime up to 50m

If a non-cartel miner comes into the picture with 17% hashpower (so gets
one block an hour roughly), they could have a different policy. (If they
had less hashpower, they couldn't affect prices) They'd still get the
1500 90c fees that were set based on assuming a cartel member would
only pick them up immediately at that price. They could pick up all
the 20c fee transactions, without having to let them age 50m as well,
for maybe $1650 in total fees. But if the non-cartel miner accepted 20c
transactions as well, users would notice, realise they can just offer
20c in fees and still be fairly reliably accepted within an hour, and
that miner would only end up collecting $360 in fees, rather than $1650.

(Accepting all the 25c transactions actually means consumers would be
paying 25c to get their transaction accepted in about 55 minutes rather
than 60 minutes which might also affect prices, but the analysis is hard
enough already...)

> For a miner it makes sense to include transactions that have much lower
> fees than the "cartel optimum": if you don't do that, you don't get the
> fees, but some other miner will still include them and "spoil the
> market",

Because you only get blocks in relation to your hashpower, and people care
about timeliness of confirmations, I don't think you can spoil the market
easily just by being a miner with a small percent of hashpower.

*Alternative* payment methods to publishing on the blockchain (like
lightning, altcoins, or fiat payments) can undercut prices and steal
marketshare though. So I think prices will still end up at "free market"
rates rather than anything really extortionate -- ie, lightning fees of
1% and bitcoin txs would cap bitcoin fees at about $2/tx even with an
absolutely effective revenue-maximising bitcoin miner cartel.

(Calling bitcoin miners a cartel is probably unfair; you don't call
"Visa" a cartel just because it sets its own fees at a profit-maximising
level. You'd only really have a cartel if the /industry/ cooperated
like that, so Visa, Amex and Mastercard all coordinating fees. Likewise
you'd only have a blockchain cartel if Bitcoin and litecoin and other
altcoins coordinated fees; switching to an altcoin seems about as easy
as changing credit card providers)

Anyway, this is pretty tangential to lightning at this point. Followups
to bitcoin-discuss or direct mail?

Cheers,
aj



More information about the Lightning-dev mailing list