This is regarding a minor point of Adaptive Issuance, following up from a discussion on stack exchange. The current design (in Oxford) is already good. If you are considering voting for AI, my comment below should not form a basis to change your opinion. Besides, this can always be changed later.
To the point:
edge_of_baking_over_staking is hard to make sense of and there is a better alternative.
Under AI, baking rights come from the baker’s stake and their staker’s stake. To reflect that, each time the baker contributes to consensus (bakes, attests…), rewards are distributed between the baker and their stakers.
But bakers are able to collect an extra sum called “edge” from their staker’s reward, in exchange for doing the validation work on their behalf. This, just like reward distribution itself, is enshrined in the protocol.
Baker has the ability to set a value,
edge_of_baking_over_staking, between 0% and 100%. By default, it is set at 100%.
A 0% value means that the rewards are split based on stake distribution alone, therefore the baker does not earn anything extra from providing the staking service. This is the simplest case.
When this value is set to anything else than 0%, the following happens each time the baker participates in consensus:
- the protocol reserves a percentage of the reward corresponding to the
edge_of_baking_over_stakingand distributes it to the baker’s free balance
- the rest is divided between baker and staker based on stake distribution, and remains frozen
Another notable case is 100%: in this case, stakers earn nothing (their frozen balance does not change) and all the rewards go to the baker’s free balance.
- a public baker who is accepting third-party stakers, wishing to maximize their earnings must periodically run the
stakeoperation, otherwise their edge is treated as delegated balance and earns less
- the new staking primitives are supposed to decouple the technical skills of baking and the financial risk taken by stakers, in exchange for the baker getting an extra percentage of rewards. However, the current design still gives an advantage to the solo baker over the staker: indeed, the solo baker is able to set their edge to 100% and have their rewards unfrozen instantly. Whereas, the staker has their rewards from staking always frozen and has to wait 6 cycles to access them.
Instead of having the edge go to the free balance, have it go to the frozen balance. This is doable with a minor change to the internal accounting used by the protocol.
- the baker never needs to run the
stakeoperation again if they want to maximize their baking rights
- rewards from delegation are always free, rewards from staking are always frozen: this makes more sense overall. It’s cleaner and easier to reason about.
Q: instead of changing it unilaterally, how about giving the baker the choice between this model and the original oxford model?
A: the oxford model actually is a “middle of the road” approach: from a baker’s perspective, the edge from staking goes to the free balance, but the staking rewards from the baker’s own stake are still frozen.
You could imagine an on/off switch where all rewards due to the baker (from own stake and edge) are unfrozen. Or even a new parameter that goes between 0% (all baker’s rewards frozen) to 100% (all baker’s rewards free). But it has problems:
- this preserves the unfair advantage described above: the baker has immediate access to their rewards, while the staker doesn’t
- besides, it’s not possible to give the same advantage to the staker, while keeping the current lazy reward accounting model
- a baker opting for all their rewards to be free might become overstaked over time, even though they are originally not, through no fault of their stakers.
- the generally accepted goal of adaptive issuance is to encourage staking (up to a point), therefore it makes sense for the protocol to default to having the rewards frozen in any circumstance.