This page is a workspace to aggregate, clarify, and present motivations for implementing the FIL+ premium (fka subsidy) proposal. We might aim to turn this into a presentation we can make publicly to educate and build enthusiasm for the change.
This page is not about technical challenges or specific policy analysis. It’s a motivation for why to attempt solving those problems in the first place. It is one-sided.
The contents of this page is expressed slightly better in this presentation.
Background
- FIL+ premium discussion issue https://github.com/filecoin-project/FIPs/discussions/243
- First draft of FIP PR
“Deal” below always refers to FIL+ verified deals.
Make FIL+ deals more attractive
- FIL+ premium enhances Filecoin Plus, it does not wind it back. Improves the predictability and reliability of the reward.
- FIL+ premium will be more reliable, especially for small providers, because payment doesn’t depend on the consensus lottery. Smaller providers have a relatively harder time getting their blocks included.
- FIL+ premium pays sooner, aligned with the storage deal. The QAP mechanism spreads reward out over the whole sector lifespan. In general it delays rewards, sometimes far after the deal has completed.
- The FIL+ premium pay-out period is smooth, no lottery, and aligned with the deal lifespan.
Make FIL+ performance and reputation more transparent
- With QAP, deal-seeking providers hosting useful data compete with capacity-maximising providers on the same leaderboard, and are completely overshadowed.
- With FIL+ premium, a provider’s verified deal hosting performance is directly observable on chain. The “useful data provider” is a distinct leaderboard upon which the deal-seeking providers can compete, without being overshadowed by capacity-maximising providers who take no deals.
- The QAP mechanism diminishes the apparent contribution by spreading it out over sector lifetime.
Make storage more flexible
- Best storage API would be to treat it as a big fungible bag of sectors: add, remove, update storage freely to satisfy new deals. QAP opposes this, but the FIL+ subsidy would remove all these restrictions.
- The QAP mechanism spreads out the deal’s influence on power over the sector’s lifetime (regardless of deal lifetime). This makes it difficult to change sector content, e.g. to take new deals. FIL+ subsidy removes QAP, which makes sector capacity flexible.
- With QAP, it’s possible to add new data to a sector, but impossible to remove or replace data in a sector because the provider has either been overpaid or underpaid for the deal, depending on deal lifespan. Expired data cannot be removed from a sector. Data cannot be replaced even if the deal is complete.
- Thus, QAP restricts storage to write-once. A provider cannot utilise the full space-time of committed storage for deals; they can only use each byte of space once, regardless of deal duration.
- This limits possibility of features:
- Extending deal by transferring to new sector: only possible exactly at end of original sector lifespan
- Transferring deal to another provider: not possible except at sector expiration
- Taking new deals (”snap-on-snap”): not possible to re-use storage that was already used for one deal, even if the deal is complete. Cannot have a short, full-sector deal, and then replace with another full-sector deal (unless we’re prepared to allow things like 15x or higher power boost).
- Capacity deals: client cannot remove or replace data at will, though there might be specific cases that work.
Make storage more valuable
- Storage that is more flexible is more valuable. With FIL+ subsidy, a provider can expect to extract more reward per sector because they can use the full space-time for verified deals, even if those deals are not known when the sector is committed.
- Flexibility incentivises longer sector commitments. QAP incentivises short sector commitments, because some rewards are delayed until sector expiration. QAP then disincentivizes extending sector lifespan, because the storage cannot be re-used for a new deal. A QAP miner needs to re-seal in order to take new deals. With FIL+ subsidy, a provider faces no penalty for making a long commitment up-front, even if they can’t fill the space-time yet.
Enable innovation and iteration speed
- The built-in storage market actor is privileged as the arbiter of deal weight, and hence QA power. The miner actor is tightly coupled to the market actor. E.g. the miner invokes and trusts the built-in market to compute weight/power. The market is consensus-critical, inefficient, and very slow/hard to change (network upgrades).
- This privilege precludes alternative markets built on the FVM from competing (or even being possible at all, without changes to miner). Alternative markets might be more efficient, offer different features, make different trade-offs. Alternative markets will be a critical building block in “storage Lego” (c.f. DeFi “money Lego”).
- FIL+ premium reduces consensus-critical, slow-moving, network-level code. It is a step toward reducing trust in and privilege of the built-in market. It reduces miner actor complexity. It moves code out of the critical trust base and is a step towards markets as pure smart contracts, evolving as fast as their independent development teams desire.
Bonus points
- More scalable. If we want to do exponential storage capacity later, decoupling deals from capacity accounting makes it a lot easier.
- More secure. Consensus power requires linear economic investment in storage hardware. Collusion with FIL+ notaries cannot subvert consensus security.
We probably can’t do this later
- Verified deals occupy 0.39% of storage (0.49% of QA power). So changing now is a relatively small operation. Stakes are limited. Any effects (e.g. changes to deal reward timing) will be small and we have lots of time to observe and react. More design freedom is available.
- If we want these advantages later, changing in the future will be a lot harder if deals occupy a substantially larger portion of space / power.