Generalized proof-of-stake (bonds) as a potential distribution allocation mechanism

Cryptoeconomics
1

Editor’s node: for now, this is less a specific proposal and more an invitation for general brainstorming.

There are many different ways to conceptualize what it is exactly that proof-of-stake systems are doing, or should be doing. To wit:

  1. A simple, consensus-mechanism-focused understanding would take them to be simply an anti-Sybil mechanism, almost a direct replacement for proof-of-work, where stake needs to be locked in order to guarantee the possibility of slashing, and thus so-called “cryptoeconomic security”. This is the understanding from the original “Slasher” blog post, and it’s still probably the dominant one. From this perspective, proof-of-stake just means some way to allocate voting weight in consensus, usually combined with a short lock period and slashing for misbehaviour in said consensus.
  2. The concept of “locked stake providing cryptoeconomic security” can be generalized to cover more services, as described e.g. in the rainbow staking post, and logically modeled as in the message logic paper. From this perspective, proof-of-stake just means some kind of token locked for some time that can be partially or fully burned (slashed) if a proof of misbehavior w.r.t. a service commitment is submitted to the appropriate location.
  3. From a more economic perspective, the “locked stake” or bond aspect of proof-of-stake can be understood as a way to redistribute ownership to participants who lock up their tokens (and perhaps provide services to the network), as compared to participants who don’t. This also gives the network more control over the liquidity of its token supply over time. This can be generalized into a kind of target maturity curve as described here. From this perspective, proof-of-stake just means that tokens can be locked in “bonds”, which typically (a) mature after a fixed period of time and (b) pay interest.

These understandings are mostly separable and complementary, but should be understood as having distinct goals and thus design criteria. I think (3) can be further split into two distinct sub-goals:

  1. Give the network the ability to control its own liquid token supply over time through regulation of interest rates. This has some similarities to sovereign bond markets and central bank interest rate controls.
  2. Give the network the ability to redistribute more ownership to parties who have made more long-term commitments which align their interests with those of the network. The aim here is to indirectly encourage those parties to take actions which benefit the network as a whole, as they stand to benefit, and to discourage actions which sacrifice long-term sustainability for short-term benefit, as it would be more difficult to exit. A great advantage of this method, as compared to other methods for rewarding contributions, is that it does not require measurement.

(2) contains some interesting design questions, such as:

  • Should these bonds be transferable (before term), non-transferable, or both?
    • Intermediate options also exist, e.g. governance-approved transfers, periodic auctions, etc.
  • Should interest rates be fixed or variable?
  • Should interest be paid periodically, at redemption, or only on exit?
  • Can these bonds be used as collateral for service commitments, borrowing, or other things?
    • An interesting option is to let them be used as collateral to borrow from the network, if the network itself has a supply of other assets. Some control mechanisms would be necessary to allocate limited borrowable assets in this case.
  • Should it be possible to sell a bond early (back to the protocol, or to a third party), and what should the penalty be for doing so?
    • A cubic-slashing-like mechanism could potentially be used to set penalties here.
  • What kinds of network-legible identity mechanisms could be incorporated, and what would be the benefits and drawbacks of doing so?
  • Should it be possible to permanently forfeit the right to exit?
2 Likes
2

I think we need to distinguish between committments and actual contributions: Without assuming measurement, we would only be able to take the outcomes of credible mechanisms into account, and I assume we will need at least some mechanisms outside of this set.

Example:

  • Automatically credible:
    • Immediately: A commitment to stake all my tokens forever (if this is an option)
    • Per interval: A commitment to not equivocate TXs
  • Needs measurement: Commitment to provide routing, solving, etc.

Including all of the latter class, requires the following perspective,

but I think it makes sense to not treat other services different from consensus provision, and we can still separate governance for everything.

Regarding the design questions, for the redistribution mechanism:

This can be fixed at the protocol level, but as far as I understand re-staking would enable choice for this property out of band, albeit with weaker guarantees. Regulating this seems hard.

These are also downstream of restaking, even if we make an in protocol decision.

I think this should be left up to choice per service. The further down in the dependency stack a service is, the higher the penalty should probably be (with infinity penalty basically forbidding early withdrawal.).

For many services, periodic seems like a choice that would lead to incentives that are easier to reason about, since with the correct structures, we could achieve continuity in iterated games.

On redemption or exit could provide interesting features, but effects of the discontinuities introduced should be analyzed in each case.

This would be a commitment over infinite time, which could introduce discontinuities as well, depending on the mechanism.

In general, the class of mechanisms for redistribution should probably err on the side of preserving optionality and keeping the set of assumptions minimal, since network and external games will develop over time, requiring the ability to adapt.

Guarantees should also be dinstinguished by assumptions, e.g. for outcomes of credible mechanisms, we can give better guarantees than for the outcomes of mechanisms that require measurement.

1 Like
3

I agree that there are two different classes here. Here by “commitment” I just meant “locked tokens”, which can be understood as forfeiting the ability to exit (perhaps partially, or for a specified duration of time). Maybe we should use a different word for this as to not confuse it with “service commitment”.

Yes, it is difficult to prevent the creation of an extra-protocol mechanism for reassignment of ownership, at least without some more protocol-legible notion of identity. Non-transferable bonds may be valuable as a social commitment, though (i.e. it may be valuable to have the option, even if we don’t expect to be able to enforce it).

I think that there’s a large class of regulatory mechanisms which don’t work without some notion of identity which is legible to the protocol. A topic to investigate.

Good point.

1 Like
4

Musings:

  • One example worthy of conversation as a reasonable approximation that is specific and not general is what Osmosis did with LP positions a few years back. Users could choose to “bond” xToken-OSMO for three durations (1, 7 and 14 days), with the added benefit of superfluid staking for long duration bonds (14 days). Basically, you can borrow against your bonded Osmo (interest free ) and use it to stake it in the P-o-S application, for additional yield (which the protocols handles minting and burning under the hood).You could not transfer the “bonded” LP positions once deposited with the protocol until the lock duration expired. Periodic auctions run by the protocol sound like a favorable idea here.
  • Probably you might want both options. Some agents may prefer fixed rates because then they can hedge any interest rate risk or make any relevant financial plans over a long bonding time period. Variable rates make sense as market conditions change in real time.

Could you burn the interest?

  • All token holders would benefit,
  • side step any income tax issues in various legal regimes, you
  • create less issuance
  • Depends what the crypto economic model for accountability would be on loan default and also if the loans are over / under collateralized. The protocol would probably need some native or built-in liquidation mechanism.
  • Yes perhaps in exchange for something like a soul bound token or other reward.
1 Like
5

One could certainly implement a protocol which burns the interest, but what would the benefit be for long-term bondholders then / what is the difference between having interest and then burning it vs just not having interest at all?

I think interest could be structured in a way which makes it clear that it’s not income (and from a theoretical perspective it’s not, it’s just redistribution of network ownership). For example, it could be structured as an option to take a loan from the protocol in the native token, where the amount of the loan that can be taken increased over time.

That’s the case if these loans are granted by the protocol itself, yes. I guess I was implicitly imaging that other parties would be granting the loans (but potentially accepting these bonds as collateral). The protocol itself granting loans is an interesting angle to investigate as well (in which case a native liquidation mechanism might indeed be necessary).

Another interesting avenue to investigate could be a mechanism whereby bondholders lend to the protocol, perhaps for a fixed term, in exchange for a call option on the token (which the protocol can guarantee), whereby at the expiration of the term, the bondholder could elect either to withdraw the originally lent capital (with no interest), or to exercise the option and convert the lent capital position into a bond of the token (at the pre-agreed price). I think that this is interesting because it’s a way for users to benefit from the network doing well without actually needing to sell their existing asset(s), with a capped downside (modulo a bit of potential risk from whatever is done with the lent funds). For example, lent USDC could be converted to actual tokenized USG bonds, where the network itself earns the interest (which doesn’t change the trust assumptions very much typically), but lenders receive the potential upside of a call option (and if they exercise it, the lent asset is converted into network reserves, potentially contributing to the floor price of a metastability mechanism).

1 Like