The Livepeer protocol tokeneconomic design has always had a historical leaning towards the theoretically pure, sometimes at the expense of the practically applicable. It would assume states that would come to exist when acted upon by rational actors under efficient market conditions and high speed. Some examples of this include:
-
ETH as the payment mechanism for jobs rather than LPT because of the theoretical infinite velocity of money on the blockchain, would lead to no benefit of using LPT, while ETH was far more widely available amongst potential users of the network in the early days.
-
Ignoring the issue of paying fees to oneself on the network because there’s no actual harm in doing so. The main way to prevent this is a network level tax, and that just makes the network more expensive - a worse outcome for users.
-
Ignoring the issue of orchestrators claiming rewards when not actually doing work, because it doesn’t actually harm anyone, and is sub-optimal for those actors vs delegating towards high performing nodes that are actually sharing fees.
-
There is no LPT burn mechanism, because LPT is based on the “network ownership % as proxy for ability to earn fees” model, rather than trying to be an increasing $/LPT currency asset.
Focusing on the behaviors In cases two and three, the cost to acting this way is not rational from a profit-maximization perspective. It’s cheaper to avoid the gas fees on ticket redemptions and just perform the work without settling payments to oneself through the network. It’s more rewarding to stake towards a high performing, low reward cut, low fee share node than it is to run infra, pay for daily reward calls, and earn no fees.
I still stand by the fact that those two behaviors are not actually hurting the network as much as people think, if at all. But there are some side effects that seem to capture people’s attention, that they bind on to in forming negative sentiments about the project. I’d like to focus this post on the narrative and impacts around the recycled fees concept, in which users run gateways and send jobs to their own orchestrators. The narrative goes as follows:
- The circular fees are fake and they’re inflating network statistics and perpetuating a false usage narrative!
Why does this happen?
Some nodes have sent fees to themselves over time for different reasons. Some are testing their PM setups. Some are using their own infrastructure first when available, and then falling over to other providers when extra capacity is needed. And some are just padding their stats (and the Livepeer network stats in general) on the explorer.
Yes, from the very onset of the project, there has been nothing stopping people from setting a 100% fee cut, then sending PM tickets to themselves. I’m actually surprised this doesn’t get acted upon more, since it has always been trivial to send 1, 10, 100 ETH payments to oneself onchain without risk. Of course any node who has a fee cut set below 100% would have a tremendous cost to doing this - they immediately share a portion of their fees to their delegators - so any stake they attract would mean that for each ETH they send to themselves, the less ETH they get back.
A fee cut under 100% serves as a tax against recycling fees to oneself. And a 100% fee share means there is no tax on the fees, though it’s unlikely the nodes will attract stake. Without a tax, there is nothing to stop this behavior.
Network Observability
Observers of Livepeer want a way to evaluate the usage of and utility of the network. The accurate way to do this would be to analyze the users of the network themselves. What are they paying for transcoding or AI inference? Are they happy with the QoS? At what scale can you observe their own usage of it via the sizes of their businesses or applications?
Collecting this data is slow, offchain, and requires hard analytical work. But it’s the most telling. If the network is making users happy, then this is all that should matter, and usage should grow
However the market seeks a way to evaluate the network simply by reading data about usage onchain. The protocol design doesn’t provide any foolproof mechanism for this however. Media data doesn’t hit the blockchain for obvious reasons, the PM payment mechanism is approximated to be gas efficient, and the network is intentionally open access and non-limiting and non-intrustive to users. The best proxy for doing this is to observe fees. However, given the recycled fee scenario, this isn’t perfectly accurate either given that some fees may be self-dealt. An analyst willing to do a little more work could do a better job by observing “fees to non-100%-fee-cut node operators.” And that would probably be a pretty good solution.
Another option to prevent recycled fees
As noted above, any time there is a tax on redeeming fees, it stops people from paying fees to oneself. Any fee share already serves this need, but there is another option: instituting a network wide fee tax as a parameter on the protocol. Then, even in the case of nodes that take 100% fee cut, some percentage of the fees would immediately be taken from them upon ticket redemption.
What would the network do with these fees? There are a lot of options - routing them to the treasury for public goods funding, redistributing them across all stake, or, maybe most interestingly in service of creating an LPT sink to address theoretical mechanism #4 listed above: it could implement a buy-back-and-burn mechanic to impact reducing the circulating LPT supply. Any fees taken in tax could immediately purchase LPT at market on DEX pools and burn the acquired LPT.
Why not do this? It theoretically makes the network more expensive for users….which is a pretty big reason. Node operators relying on fees to cover their operating cost, would now have to raise their prices to cover the same operating cost. UNLESS, the perceived increase in LPT value from their reward cut achieved via the buy-back-and-burn mechanism, more than offset the value of the fee tax.
It’s a long way to go before a tax on the fee value would represent more value on the network than the inflationary LPT issued in reward cuts. However as inflation has been steadily decreasing and continues on a downward trajectory, as it approaches single digits within a couple years, and fees begin to rise, there is an opportunity for an eventual equilibrium where fees outpace inflation and LPT supply begins to reduce.
On the way, would we see higher reward cuts and flat pricing? Is that ok for node operators and worse for delegators? Is that ok considering the node operators are doing the majority of the heavy lifting on the network? Or do the delegators also value the perceived LPT benefits that come with a mechanism to eventually reduce supply?
Do we actually care about the network being perfectly legible from simple onchain observability, or do we only care if users find it useful via the pricing and QoS that they receive?
Simplified option to consider
To simplify: The Livepeer community could consider introducing a network parameter “feeTax” that gets debited from onchain ticket redemptions. The tax could be used to buy-back-and-burn LPT to reduce the circulating supply.
The arguments for this include:
-
Make the network fees more “trustworthy” and legible to the market.
-
Reduce LPT supply.
The arguments against this include:
-
Tax makes the network more expensive (either for users, or delegators), hurting adoption.
-
Onchain legibility doesn’t matter, as it doesn’t increase the utility to end users.
This isn’t a solution to every challenge faced by the Livepeer network. I bias towards the theoretically pure - and stand by the fact that neither of these issues actually hurts the QoS of the network and utility to end users. But I do recognize that this is one simple mechanism that could solve two issues that often get brought up by practical actors in the ecosystem. I definitely welcome people’s feedback and community debate! What do you think?