Capped Duration Multiplier: FIP-0056/0036 SUPERSET proposal

[ribasushi]

The excitement around Filecoin's foray into smart contracts, starting with direct EVM support on its L1 chain can not be overstated. At the same time the network continues losing RawBytePower, driven by the confluence of multiple external and internal factors. FIP56 and its older twin FIP36 are proposals put forward by CryptoEcon Lab @ PL designed to enhance and stabilize the Filecoin economy. We propose a draft FIP: superset of the ideas developed during SDM research, resulting in the "Capped Duration Multiplier" (CDM). We believe CDM allows for a radically more balanced and fair Filecoin network, unlocking more use cases and as a result requiring fewer adjustments in the future.

Unlike virtually any other major L1, the Filecoin "proof of probability"-driven consensus derives from a SNARK-based, energy-efficient account of posession of physical goods: storage space. The primary authors of this FIP hold a strong belief that this unique feature of the Filecoin consensus is extremely valuable. This proposal provides a solution to allowing for two coexisting business models:

1. Storage Providers can continue to specialize in the evolution of decentralized data services, thus maximizing their Fil+ exposure.
2. Storage Providers also have the rational option to specialize in securing the consensus of a smart-contract-capable L1 blockchain, while also providing sealing as a service and potentially discovering novel ways to monetize the security substrate itself (reminder: what counts is how much capacity there is, not what is stored in it).

Additionally SPs are able to mix and match the two business models, which in turn creates a viable market for small piece-size storage contracts. This eliminates one of the significant hurdles data DAOs face at present.

Simple Summary

• A Capped Duration Multiplier (CDM) policy is introduced for all new and existing sectors on the filecoin network.
• A sector committed for a longer time receives progressively higher Quality Adjusted Power, compared to a sector with shorter commitment, all else being equal.
• The Duration Multiplier is a capped multiplicative on the existing Quality Multiplier incentive (Filecoin Plus incentive).
• The maximum Quality Multiplier available to any sector remains 10x. This FIP merely introduces an additional source of multiplication, which can be combined with the existing one, up to a limit.
• Sectors with higher Quality Adjusted Power will require higher initial pledge collateral.
• The termination fee cap will scale with the duration multiplier.
• The minimum sector duration time will increase from 6 months to 1 year and the maximum sector duration will increase from 1.5 years to a little over 10 years. The upper bound of Deal Duration Bounds will increase to a little over 10 years as well.
• The CDM policy will apply at sector commitment, extension, and replica update

Incentive Considerations

The solution space of combining the two modes of operation is illustrated by the following simplified table: an approximation of the optimal behavior of a rational SP with variable access/exposure to Fil+ data-holders.

Fil+ Exposure	Min Rational Duration	Effective QAP
100%	MIN_SECTOR_DURATION	10x
80%	2.72 years	10x
75%	2.80 years	10x
50%	3.32 years	10x
33%	4.02 years	10x
25%	4.58 years	10x
20%	5.08 years	10x
15%	5.76 years	10x
10%	6.77 years	10x
5%	8.40 years	10x
2%	9.98 years	10x
1%	MAX_SECTOR_DURATION	9.43x
0%	MAX_SECTOR_DURATION	8.65x

The table demonstrates a path to sustainable CC operations even in face of the enormous gravity of Fil+, revitalizing the no longer viable CC business model, and preserving the current Fil+ centric business-model at the same time. This in turn reverses the trend of raw-power loss bringing us back above the baseline (crossed on Feb 9th) and simultaneously dramatically relieves pressure on Fil+ abuse and governance.

Product Considerations

The CDM exponentially increases the experimentation space just in time to be able to support the opportunities afforded by the F(E)VM. It provides space for complementary business models, with ability to transition between them by design. A short summary of benefits from discussion above:

• Providing capacity (filled with data or CC), and associated expertise remains a viable business model
• Ability to compete with Fil+ on near-equal footing instantly opens doors to paid storage contracts
• Avenue to rationally grow the consensus security and TVL of the network until data-transfer technology catches up
• Enhances decentralization by making it viable to operate an SP in a space-rich yet bandwidth-constrained environment
• Pressure on the Fil+ program reputation and governance is drastically reduced (both from legitimate clients and otherwise)
• Archival needs of business clients worldwide are adequately met (🇺🇸 IRS: 7y, 🇬🇧 legal: 6y, 🇩🇪 business records: 10y)
• The partially-Fil+ sector model becomes very lucrative, creating a viable market for small piece-size storage contracts

Design Rationale

The core insights about Fil+ leading to the CDM are:

• The Fil+ multiplier was introduced at network genesis as an efficient mechanism incentivizing storage provider engagement with an independent data owner. While the 10x parameter was justified at launch, it is extremely overpowered in the current state of the network. The use of Fil+ at about 5% of the overall capacity has already set off a race to the bottom, leading to rampant abuse of the Fil+ system as amply documented in recent Fil+ disputes. Giving Fil+ the ability to potentially be further SDM-boosted by an effective 45x factor compared to a plain Data/CC sector, would make the network completely uneconomical and irrational for anything but Fil+, which in turn severely weakens the incentive's original purpose.

• Engagement between SPs and legitimate Fil+ holders remains extraordinarily difficult. About 80% of the hurdles are technological, in the form of inefficient or outright non-existent software and nascent-at-best playbooks and brokers that a medium-to-large data-holder could leverage. By most optimistic estimates the status quo will persist for at least the remainder of 2023. The remaining 20% are economic in nature. It is believed that many of the liquidity problems can be alleviated by smart contracts, but until the F(E)VM is widely available and stable it is difficult to predict an outcome.

• Even if the current state of tooling was not a factor, converting the current capacity-carrying part of the network to data-carriers would require an enormous movement of data between holders and SPs. The majority of the raw power securing the network is not physically equipped to host high volume market nodes, placing further pressure to seek out schemes abusing Fil+, just to stay afloat.

• As use, legitimate and otherwise, of Fil+ grows, the network will continue losing raw capacity, posing a risk to the security required of a F(E)VM-capable L1 blockchain. It is a strong belief of the CDM proponents that consensus-pledge staking as implemented in Filecoin is insufficient to provide the required level of security, posing a significant challenge to the future of the F(E)VM.

Additionally sourcing ⨎ for pledge collateral remains difficult, contrary to what one would expect at current circulating supply levels. We believe this is caused to a large extent by the presence of multiple circulation spheres: one aligned with decentralized storage, while another aligned with traditional block-producers. Provided this hypothesis is correct, this would add another avenue to increased network TVL, and thus increased price stability.

Specification

This proposal is a strict superset of the specification outlined in FIP56. Common parts are not reproduced here during the discussion period to simplify things. The only additions are:

1. Increase MAX_SECTOR_EXPIRATION_EXTENSION, and the corresponding max_prove_commit_duration to 3700 fildays (~10.13 years)

2. Cap the maximum QAP available to a sector, by defining CDM as:

    FILP_QAP_FACTOR=10          // same as mainnet
    MAX_SECTOR_QAP=10           // !!!NEW!!!
    SDM_SLOPE=1x-per-360days    // same as FIP56/36
    SDM_LAG=540days             // same as FIP56
    MIN_SECTOR_DURATION=360days // same as FIP56

    SECTOR_QAP = MIN(
      MAX_SECTOR_QAP,
      (
        MAX( MIN_SECTOR_DURATION, ( SECTOR_DURATION - SDM_LAG ) )
          *
        SDM_SLOPE
          *
        (
          SECTOR_SIZE - FILP_PIECES_SIZE      // CC or Data
              +
          FILP_QAP_FACTOR * FILP_PIECES_SIZE  // Fil+ client data
        )
      )
    )

[SBudo]

I love this!!
Awesome idea to create a "new" type of Storage Provider (call it "Consensus Provider") along with the associated incentives.
One question that I'm sure will be asked:
Does this apply to newly created CCs or can existing CCs be extended to MAX_SECTOR_DURATION?

It's going to be VERY important to apply the appropriate penalties in case of early termination...

[dd45e640b42e6da7da96faee3996ef7c]

it applies as the original FIP56#sector-extensions to extensions

[dd45e640b42e6da7da96faee3996ef7c]

for terminations, see the original FIP56#termination-fee too :)

the termination fee will scale with the SDM, FIP56 or CDM

[ribasushi]

@SBudo note the last point in the Summary above:

The CDM policy will apply at sector commitment, extension, and replica update

also

A Capped Duration Multiplier (CDM) policy is introduced for all new and existing sectors on the filecoin network.

[SBudo]

ah, yes, thank you

[ZenGround0]

This seems promising.

How critical is it to the design for the maximum sector duration to go all the way up to 10 years as opposed to FIP 56's currently agreed upon 3.5 years (at most 5 years if you go off of the unupdated text)? There are at least two reasons why it is significantly problematic to extend this beyond 5 years.

Most importantly crossing the threshold of 5 years invalidates the current max sector lifetime guarantee which kicks off the need for new proofs security analysis. I don't know the details here and it might be possible under incoming FIP 47 but this is non trivial.

Second we will be increasing the network carrying capacity as now every sector added will now be able to live twice as long with extensions. This makes some existing protocol design technical debt which is already potentially bad about twice as bad.

[ribasushi]

@ZenGround0 what you are missing is how ridiculously overpowered Fil+ is. With a 10.15y SDM a CC-exclusively operation can obtain an 8.65x multiplier. In other words a setup having no market node whatsoever is still competitive on the network and provides the currently falling benefits: physical consensus security + decentralization.

If we cap things at 5y max, without changing the slope, the table above looks like this:

Fil+ Exposure	Min Rational Duration	Effective QAP
100%	MIN_SECTOR_DURATION	10x
80%	2.72 years	10x
75%	2.80 years	10x
50%	3.32 years	10x
33%	4.02 years	10x
25%	4.58 years	10x
20%	MAX_SECTOR_DURATION	9.80x
15%	MAX_SECTOR_DURATION	8.23x
10%	MAX_SECTOR_DURATION	6.65x
5%	MAX_SECTOR_DURATION	5.08x
2%	MAX_SECTOR_DURATION	4.13x
1%	MAX_SECTOR_DURATION	3.82x
0%	MAX_SECTOR_DURATION	3.50x

The entire argument of "you can do things just like a Fil+ SP" goes out the window: an SP can not function on the network without having a decent fraction of Fil+ in every sector. Thus my remark: if we are still bound by the 5y max - we would need to raise the slope for this to work.

[ZenGround0]

Yeah this makes sense. To understand the other approach could you write down the function with the raised slope? I think I understand but want to make sure.

[ribasushi]

@ZenGround0 sure - here it is with a slope of 2.

Fil+ Exposure	Min Rational Duration	Effective QAP
100%	MIN_SECTOR_DURATION	10x
80%	MIN_SECTOR_DURATION	10x
75%	MIN_SECTOR_DURATION	10x
50%	MIN_SECTOR_DURATION	10x
33%	2.76 years	10x
25%	3.04 years	10x
20%	3.29 years	10x
15%	3.64 years	10x
10%	4.14 years	10x
5%	4.95 years	10x
2%	MAX_SECTOR_DURATION	8.26x
1%	MAX_SECTOR_DURATION	7.63x
0%	MAX_SECTOR_DURATION	7.00x

Note how it breaks down both the start and the end 😞🐼

It does look ( empirically, no specific insight ) that the 10y vs 10x axis with a slope of 1 is somehow the sweet spot.

Also DISCLAIMER: I am not a crypto-econ person. All of the computations above are done by hand in a spreadsheet by candle light. It would be very much appreciated if someone who knows how to drive observable or even fancy stuff like The Graphulus could help make this more approachable.

[jsoares]

It does look ( empirically, no specific insight ) that the 10y vs 10x axis with a slope of 1 is somehow the sweet spot.

That definitely depends on your objective function. What should we try to maximise here? Would we be happy with 1% FIL+ data, or should we aim for 50% FIL+ data? This is not trivial to model, as it obviously depends on stakeholder preferences, but the choice of parameters will have a major influence on the outcome.

[anorth]

@ribasushi Zen is right that increasing the maximum sector lifetime is a big deal, and hinging the proposal on this will at least slow it down a lot. I strongly suggest that you try to formulate the best version of a capped duration multiplier with a 5 year max first, and perhaps signal how it could be improved with a longer max lifetime, which we might be able to realise later.

FIP-0047 does nothing to mitigate the risk of increasing the maximum sector lifetime. This change needs a whole new security analysis. It should really be a stand-alone FIP (it has other motivations too) and I think the most practical path forward will be to remove it from this proposal, propose it separately, and expect it to take some time.

For your intuition: my understanding is that the 5-year lifetime limit exists in order to reduce the risk of technological progress (faster/cheaper CPUs/ASICs etc) eroding PoRep security. In brief, as progress makes the sealing process some multiple faster and/or cheaper, then a malicious SP gets closer to being able to fake storage capacity. The 5-year lifetime limit is an assertion that we will see this threshold approaching from at least 5 years out, at which point we will need to implement a new proof algorithm and deny the old one. Then all sectors sealed with the old algo are expired before tech catches up such that the SP can remove their hardware but still claim power. Increasing this parameter to 10 years requires us to make assertions about technological capabilities 10 years from now, which I think is much more than twice as hard as 5 years. This would require much more conservative (read: expensive) proof generation.

As an interesting aside, we are actually seeing that technological progress happening already, with proof hardware costs reducing (and more in the pipeline). My understanding is it's roughly on trend with the predicted rate of progress, so we still think sectors sealed now are good for 5 years, but can see the point in time where we stop thinking that approaching.

Beyond the two implications of proof security and network carrying capacity, another consequence of doing this is that the pre-commit deposit for all sectors must double in order to maintain current security margins.

cc @nicola

[bobdubois]

This seems a very promising proposal that would benefit all parties in this ecosystem. It stabilizes the blockchain at consensus layer, makes more providers profitable, decreases the run on Fil+ and DataCap, and allows more providers to enter the ecosystem.
I'd love to see such a proposal implemented.

[stuberman]

i am highly supportive of this proposal and would like to see it implemented as soon as possible.

[flyworker]

hey, I am trying to understand it in a easy way.
CC 10 years = 86.5% DC 1 Year in QAP
Right?

[ribasushi]

@flyworker yes, this is exactly right under the proposed parameters. Note that the parameters are up for discussion as seen above.

[flyworker]

yes, this one line formula is much easier to promote.

[ribasushi]

Also it doesn't have to be CC: it can be your own data sealed for cold storage, or it can be paying customers who rather not go through the Fil+ KYC.

[flyworker]

I know, I am thinking for the big ones.
Chinese SP

[scaseye]

@ribasushi @f8-ptrk and the rest of the small group thank you for this! Count on my vote.

[jennijuju]

Thanks for the proposal!
I’m having trouble understanding where the SDM come into play in this case. Is this like - even the fil+ deal will start with a 1x qap up to 10x base on their duration?

[dd45e640b42e6da7da96faee3996ef7c]

It is hard to design a system that provides incentives and not have it abused. If not impossible.

You either make abusing the system so expensive that it isn't worth abusing it - on the way you will loose a big part of the incentives the system should provide. Or you randomize who gets the incentives to break the connection between abuser and profit - that only works if you have a large enough group to randomize on, we currently do not have that and will not have any time in the midterm future - otherwise the participants can conspire to void the randomization.

[jennijuju]

CDM is an SDM superset - it keeps all its behaviors and considerations. CDM just adds an extra constraint.

I disagree with this point. SDM incentives longer storage commitment for all sectors. While CDM has some of that feature, it also puts a cap on the incentives. Say, assuming a sector is full of verified deals, there is an implication that SP only stores it for the shortest period of time possible to get 10x QAP and nothing longer. And IMHO, this is fundamentally different from SDM's behavior.

[jennijuju]

It is hard to design a system that provides incentives and not have it abused.

It is indeed HARD but yet should be prioritized IMHO.

[dd45e640b42e6da7da96faee3996ef7c]

I disagree with this point. SDM incentives longer storage commitment for all sectors.

as does this fip. it doesn't discriminate on potential sector content.

While CDM has some of that feature, it also puts a cap on the incentives. Say, assuming a sector is full of verified deals, there is an implication that SP only stores it for the shortest period of time possible to get 10x QAP and nothing longer. And IMHO, this is fundamentally different from SDM's behavior.

the idea that was at least discuss is to remove the stacking functionality. let multipliers not multiply. there is actually no reason to couple sector duration and sector "content" multipliers. this still leaves the overall cap at x10 thou.

for each part of a sector (whatever a part of a sector is is another question) lets have power = size * max([multiplier1], [multiplier2])

the dangerous idea this superset of fip56 is trying to address is letting multipliers multiply each other. today its x45. tomorrow its x57463...

[ribasushi]

assuming a sector is full of verified deals, there is an implication that SP only stores it for the shortest period of time possible to get 10x QAP and nothing longer

@jennijuju this assumption is faulty, it does not consider client power. A client with datacap has ( and will have for a long time ) disproportionate sway over how they set their deal terms. The network has no mechanism for "bidding": even in NV18 datacap remains a 1:1 transfer. The client offers it to the SP under specific terms and this is the end of the negotiation: take it or leave it. Clients can, and will, find the lowest "bidder", who will be represented by a large amount of SPs who say "shut up and take my capacity". The scarcity of Fil+ for legitimate purposes is here to stay, and it will only become worse if we, as you say, "just fix Fil+".

This comment from the other thread remains applicable to the CDM just as much as it is applicable to the SDM: SPs have no say over Fil+ contract duration. Clients will dispense datacap on their own terms. and if the SP doesn't like it - they will go to their peers. At some point market dynamics will set in, but this will only happen when Fil+ stops being the only game in profitability-town.

[xinaxu]

I really like this idea. It does give CC miner same opportunities as FIL+ deal taker, though I would love to get more input from them to see if 10y makes sense to me. IMO, it probably make more sense to pledge for 3-5y instead of go for 10y.

1. The collateral for the same capacity will be significant lower in a few years due to lots of factor, i.e. growing QAP, reward halving, etc. This make 10 years a bit too long to get reasonable ROI for the last few years.
2. All commodity HDD has 5y warranty, meaning the miners will need to migrate to new hardware after 5y mark, significantly raising the cost for the last 5 years
3. To actually compare with most FIL+ deals, most FIL+ projects require SPs to retain an unsealed copy, so that per each sealed sector, the effective multiplier is 5x. To compete with them a 5y CC would be good enough

[xinaxu]

I'm not opposing 10y, just saying that the parameters need to be debated with input from big CC miners.

[ribasushi]

@xinaxu if you think 5y can cut it, even with the numbers as listed in #636 (reply in thread) this is GREAT. Authors assumed this will not fly / keep too much focus on Fil+, therefore we never proposed it.

[SBudo]

I would still keep the 10 years as an option.
SPs who want to do CC can decide what makes sense to them economically and pick the length/multiplier.
The 10 years sector duration is important for deals (regular and FIL+) as businesses out there want long-term/archive retention for their data

[anorth]

See #636 (reply in thread)

Separately, deal durations are not fundamentally connected with sector commitments. Once we fix things up (#298) so that user-programmed markets are first-class, SPs will be perfectly able to make 10-year deals, tho then will need to move the data between sectors over time to satisfy them.

[flyworker]

general speaking the proposal is good.

Do we know when will we have another change?
It is just a concern as SP

[dd45e640b42e6da7da96faee3996ef7c]

one of the intentions of this FIP is to stabilize future developments regarding multipliers. by capping the overall multiplier to x10 and having future developments play out within that cap will provide some planning security.

[jcace]

interesting approach! I think this is a very creative way of solving these problems.

What affect (if any) does this have regarding Snap deals?

For instance if I had a 100% CC sector pledged for 10 years and I snap it up after 1yr- can I put in a Verified Deal with length 1 year? Then, the sector's end epoch would be shortened, yet still have the 10x multiplier for its lifetime (2 years total)

Additionally, could this theoretically result in a kind of disincentive for verified deals longer than 1yr?

[ribasushi]

What affect (if any) does this have regarding Snap deals?

The CDM policy will apply at sector commitment, extension, and replica update

For instance if I had a 100% CC sector pledged for 10 years and I snap it up after 1yr- can I put in a Verified Deal with length 1 year?

Based on my reading of https://github.com/filecoin-project/FIPs/blob/master/FIPS/fip-0045.md: yes you can ( @ZenGround0 would have to keep me honest though 😅 ). Note that at this point no power-recalculation will take place: your sector is already at max-QAP.

Then, the sector's end epoch would be shortened, yet still have the 10x multiplier for its lifetime (2 years total)

Sector lifetime is never shortened, this is not a thing. Also see the "already at max-QAP" above.

Additionally, could this theoretically result in a kind of disincentive for verified deals longer than 1yr?

As pointed out elsewhere #554 (reply in thread): SPs have no say over Fil+ contract duration. Clients will dispense datacap on their own terms. and if the SP doesn't like it - they will go to their peers.

[anorth]

For instance if I had a 100% CC sector pledged for 10 years and I snap it up after 1yr- can I put in a Verified Deal with length 1 year?

Based on my reading of master/FIPS/fip-0045.md: yes you can ( @ZenGround0 would have to keep me honest though 😅 ).

No, you cannot snap in a verified deal with a maximum term of less than the remaining 9 years. Note that data cap allocations have a MaximumVerifiedAllocationTerm of 5 years, unless a FIP changes that parameter. You can snap in a verified deal that has a minimum term no greater than the remaining commitment, and a maximum term that is no less than it.

[momack2]

Interesting proposal! Thanks for making it and pushing the conversation forward! I had a few initial concerns while trying to digest this I'd love to get opinions on:

• I'm concerned the proposed "max cap for FIL+ incentives" harms the existing incentive to store useful data cheaply on Filecoin. The incentive to always upgrade an empty sector with data is a feature, not a bug. Doing the effort of actually engaging in a FIL+ deal (with clients, data transfer, retrievals, etc) is more costly to an SP from both an overhead and operational perspective than solely sealing a CC sector for multiple years. This proposal would hurt FIL+ SPs, because the subsidy doesn't account for the additional work they expend into data storage, AND they don't get to benefit at all for longer-term deals (which will imply additional expenses around longer client support / retrieval expectations / etc).
  • Potential mitigation - could cap the "MAX" for FIL+ bytes at a higher value than solely CC sectors (but maybe a less than 10x higher value), so that there is always an incentive to fill CC sectors with deals (as per the current protocol, which incentivizes things like snapdeals to put existing sectors to use).
• I don't think the termination fee is sufficiently hardened to apply the right stability incentives for 10 year sectors. We want sectors and data committed to Filecoin to be robust. If CC sectors could get up to a ~10x multiplier for 10 year commitment, the termination fee would need to be raised to ensure there was no incentives for SPs to commit a 10 year sector, and then terminate it and collect the inflated rewards after just a few years. Current termination fee is not nearly high enough to ensure that with a larger multipler, and would need to be raised significantly to enforce protocol robustness and avoid SPs gaming the system.
  • IMHO - we should do this anyway - but becomes a hard blocker for adding large multipliers for even longer sector duration.

[hyunmoon]

#554 (reply in thread)

[momack2]

@f8-ptrk - correct - I'm not confident that termination fee (even increasing proportionally) is enough to create sufficient stability incentives for 10 year sectors.

[ribasushi]

the proposed "max cap for FIL+ incentives" harms the existing incentive to store useful data cheaply on Filecoin. The incentive to always upgrade an empty sector with data is a feature, not a bug.

The proposal with the original numbers does in fact address this:

• as an SP you either have some markets exposure ( at which point why not take more )
• OR you take a haircut relative to your peers: a 100% non-Fil+ sector has a max 8.65x QAP at max duration, as opposed to the potential 10x.

One can not eclipse the other: reexamine the table in the body of the post.

I don't think the termination fee is sufficiently hardened to apply the right stability incentives for 10 year sectors.

Actually I agree with you here: hardening this further would be quite welcome. However this is the limit of analysis our small group can do alone: we need CEL's help to make further progress on this front.

[benjaminh83]

I like this proposal although it hurts FIL+ because it actually takes a step in the direction of getting real clients and "charging for storage", wheres FIL+ is no silver bullet. Therefore it makes sense to have different approached that is incentivized. With this proposal, I as an SP can focus on doing paid client deals, storing data for 10yr, get a decent multiplier and forget all about FIL+. This will be unverified data, as the client is paying, and we have no desire to come up with 45x collateral, just to earn more block rewards.
So as @Angelo-gh3990 is also mentioning - lets also incentivize SPs serving real paying clients. We do not need FIL+ for that.

[PHABRIC77]

@benjaminh83 can you please connect the dots for me a little? Why does long-term duration CC need to be incentivized to enable SPs to target fee paying clients? I understand that the FIL+ incentives are seen by many as creating the wrong incentives and thus undermining client pricing discipline currently. But this is an argument to level the playing field (or at least to reset the relative incentivization) as opposed to an argument for long-term CC incentivization per se. Why would appropriate incentives not be applied to long-term data storage (for example), with possible additional incentives for robust verification process/other standards? Or there being some balance struck between the competing priorities based on the network's intended goals. Appreciate your input!

[Tom-OriginStorage]

This seems an interesting proposal. However I have some concerns/considerations that the proposal may bring

a. Limits flexibility: The proposal may limit the flexibility of miners to adjust their commitments and storage deals based on changing market conditions or other factors.

b. Increases initial pledge collateral: Sectors with higher QAP will require higher initial pledge collateral, which can increase the costs and barriers to entry for miners.

c. Increases termination fees: Sectors with longer commitments will face higher termination fees if they leave the network early, which can discourage some miners from participating in the network.

[ribasushi]

I will only respond to the points I am qualified to speak on

b: sectors with higher QAP will require higher initial pledge collateral

This is exactly why the cap matches current mainnet. Your calculations remain the same as before: you need to plan for at most 10x the pledge per sector, regardless of which mix of datatype/piece-size/duration you choose. Unlike previous proposals, this is virtually the one thing that remains the same.

c: longer commitments will face higher termination fees if they leave the network early, which can discourage some miners from participating in the network.

SPs who do not want to take a long commitment have the option to go all-Fil+ and do short deals ( provided they find a client interested in providing such, one such source could be the PL Data Preservation DAO ). The core of this proposal is that an SP has two paths to arrive at a certain risk/reward balance point.

[dd45e640b42e6da7da96faee3996ef7c]

a: Limits flexibility: The proposal may limit the flexibility of miners to adjust their commitments and storage deals based on changing market conditions or other factors.

i agree that the outlook of potential 10 year sector duration can look inflexible. but especially with non multiplied deals the opportunity to reuse a sector (technically possible, not implemented thou) for multiple consecutive deals over the life time of a sector should add a lot of flexibility - resnapping (wanted, needed - we most likely all agree on that)

what i think it will for sure provide is a radically increased planning horizon for SPs in regards to sector sealing

[herrehesse]

Could you please explain how the proposed Capped Duration Multiplier (CDM) would impact the reduction in RawBytePower (RBP) on the Filecoin network? It seems like incentivizing longer commitments to storage sectors would lead to a higher Quality Adjusted Power (QAP), but would this not make the reduction in RBP even worse than it already is? It would be helpful to understand the potential trade-offs and benefits of this proposal for the overall stability and balance of the Filecoin economy.

[dd45e640b42e6da7da96faee3996ef7c]

with this proposal it is still impossible to max sector qap without FIL+ (or another not yet existing multiplier). sectors with current max. duration of 1.5 years will still only benefit from FIL+, not the SDM part in this proposal. FIL+ will still be a dominant force after this would be implemented

[herrehesse]

We are interested in your proposal and would like to know if it is possible to assign a dedicated team to perform the necessary calculations to further evaluate its feasibility. It would be valuable to see the results of this analysis.

[dd45e640b42e6da7da96faee3996ef7c]

in my eyes this proposal would allow to restore the balance between FIL+ and CC sectors. I do not think the chain is in balance right now. we have rough;y 5% of FIL+ sectors on the chain. as soon as we go in the direction of 9% more qap will be notary approved than from sealed sectors. thats dangerous to some extend. if we do not allow CC sectors to play a larger role in the consensus mechanism we risk this becoming an authorized chain - mining allowed by notaries.

[dd45e640b42e6da7da96faee3996ef7c]

@herrehesse especially you should know that FIL+ is currently not at all supporting the goal of "storing humanities most important data" as the x10 multiplier became an essential tool to survive as a miner on chain. taking pressure off FIL+ by allowing normal storage, cc sectors to proliferate into feasibility again should help with that too

[herrehesse]

Agreed.

[Angelo-gh3990]

I'm a 100% for this proposal. When this moves forward, sealing data without FIL+ and "charging for storage", or mixing and matching cc/data and FIL+ together makes a lot more sense. It will restore balance like @f8-ptrk says. And like @ribasushi said "you don't have to pull your client through KYC/KYB", opening up the bigger Enterprise market that way.
You had me at : "Ability to compete with Fil+ on near-equal footing instantly opens doors to paid storage contracts"
Love it, thanks everyone who worked on this!!! And @ribasushi for writing this up so clearly.

[TippyFlitsUK]

I would also like to publicly express my support for this proposal. Thank you sincerely for your efforts Riba and Patrick!!

[hyunmoon]

I thought CC sectors were designed to become less sustainable over time?
IMO, data onboarding is much harder than simple CC sealing in many aspects.
With the current incentives dilluted this much, I wonder how many would still want to keep onboarding verified data.
My operation recently started onboarding verified deals in a scale because CC mining couldn't keep us afloat.
To make this switch, we had to make a big investment in the 10Gb/s dedicated fiber line that required us 3 years of lock-in.
Also, we had to overcome a big learning curve and we had to jump through many hoops to make it happen.
All these would not have happened if there was an easier way to achieve the x10 QAP incentives.
If this proposal passes, a simple CC miner who didn't have to go through what we did will be able to earn the same block reward as us just by pleding sectors for a longer period.
That is not what I hope to see in the future as it will discourage the current and the potential data onboarders and data preparers.
Filecoin exists to work as a global decentralized data storage.
I don't think making CC sectors more valuable is on the right path towards accomplishing the goal.

[hyunmoon]

FIP-0033 will negate the concerns for the security of the chain.
Then what reasons are left to implement CDM?

[anorth]

FIP-0033 is unlikely to happen as written. See also #442. I like your sketch of a proposal @hyunmoon , but it needs more thought and detail to determine how viable it is in practise.

However, I also believe that CC (i.e. the actual hardware commitment) is still an important contributor to consensus security, and am worried about its economic viability.

[PHABRIC77]

Enjoyed reading this debate and thinking more about consensus. Dumb question time: why not incentivise compute capacity being committed for consensus purposes separately from storage capacity? Further why conflate incentives for data ingestion (Fil+) with raw power for consensus purposes? Doing so negates any possible revisions for FIL+ and appropriate standards and controls being initiated. If the concern is to limit max multiplier, then just structure the incentives at a level where an optimised SP is receiving incentive multiplier with which you are comfortable. This separation gives the SP (miner) flexibility on its operations, clarity on the incentives being offered (and the purpose for doing so), and increases differentiation of contributors to the network. It gives the network the chance to optimise and rebalance to achieve its core objectives and to respond to macro and micro developments. If there is a reason why it is all mashed together that makes my question as dumb as I fear it is please put me right! Thanks

[dd45e640b42e6da7da96faee3996ef7c]

Until the introduction of this modified proposal, PL was all but certain to go through with introducing FIP36, under the name of FIP56 as-is to mainnet

The authors are not particularly happy with their proposal standalone, but moved forward anyway since it is evaluated in the context: doing nothing == Fil+ turbocharged to 45x

Coincidentally after CDM publication the SDM authors slashed the slope of their proposal by more than half, so now it sits at 20x as opposed to 45x. While welcome, it still takes us further towards the "Fil+ or GTFO" model, which is what we fundamentally oppose for the exact reasons outlined in the draft's Product Considerations. Or more succinctly: see #636 (reply in thread)

That is to say: all your criticism above is valid and should be considered. But it really belongs in the FIP36 FIP56 thread: #554. Your timely feedback there would be much appreciated, as it seems the proposal is moving to community last call this Monday the 6th, without mention of this ( CDM ) proposal at all.

[PHABRIC77]

No criticisms from me. But I agree I think the approach outlined in this cdm should be further refined, and by inference preceding proposals are in need of even more change (ie this is a significant improvement). It follows that all my concerns are valid in the context of those FIPs and I would be happy to restate these in those contexts.

[tmellan]

Much of this proposal seems predicated on an assumption the Fil+ 10x is overpowered, that SDM will make it even more so, so the duration multiplier to Fil+ needs to be capped, if I interpret the logic correctly.

From an ROI perspective at least this isn’t true since the ROI boost from a multiplier is diminishing as the multiplier increases.

For example, CC going from 1x to 4.5x is a much larger ROI boost than going from 10x to 45x for Fil+.

This implies that capping is not necessary to encourage hardware and raw byte. Do you agree?

[tmellan]

Let me repeat what I understand and add clarifications:

• Typical proportion of pledge being borrowed could be [50,100]%
• Can you give me a per sector value for opex for a current 1.5y Fil+ sector and a 1.5y CC sector?
• I mean what loan rate are you considering (I'm assuming 20-30% per year)

[dd45e640b42e6da7da96faee3996ef7c]

• yes
• 1:~10 FIL+ to non FIL+ (i think the exact numbers do not really matter) - the fixed costs scale out fast
• max. 20% on the borrowed amount (there are lenders taking % on block rewards. we are undecided what we'd prefer - maybe you have an answer ;) )

[tmellan]

Thanks, understood.

Then it seems fairly sure the point I made above is correct (though pls run the calculations yourselves). So we are solidly within the domain of diminishing marginal change to ROI and DMs having the effect that CC can be strongly viable without a cap. Agree or not @SBudo and @f8-ptrk?

(Also just stating, I'm not against capping per se, or for it, yet, but I want to establish shared understanding of what is or isn't needed to achieve goals of this proposal).

[SBudo]

Hey @tmellan,
Thanks for the follow up, and apologies for the delay.
My response had two parts to it though, and while the ROI seems to have become the point of focus, the second part is MUCH more important: by keeping or increasing the gap between FIL+ and CC, SPs are incentivized to continue on their race to the bottom with FIL+. Even if the ROI slopes down, as soon as you introduce a higher QAP for FIL+ compared to CCs, most SPs will continue to do FIL+.

Anyone who is currently NOT doing FIL+ shouldn't be too concerned about ROI anyway (or lack thereof) as they probably will have to close shop soon or pray the FIL price triples overnight, and that's probably why the raw byte power is alarmingly decreasing (4.8% in the last 30 days).

To partly answer your question regarding the ROI, we are doing 100% FIL+, new sectors, with 100% borrowed FILs (20% interest), on leased HPE hardware hosted at Equinix (costs are mounting at every word). This was originally designed to become a long-term Storage Provider system (in the true sense of the word) that would allow paying customers to onboard their data (that seems to be a very distant future now).
I'll have to spend some time putting together an ROI comparison to fully answer your question (which I'll do), but understand that the reason I disagree is stated above. Even if, on paper, the ROI increase of FIL+ reduces in proportion to the ROI increase of CCs, most SPs will simply continue to seek FIL+ deals, as they will still be more profitable.
Part of the appeal of this FIP is to incentivize CC production. I feel that increasing the power of FIL+ in parallel will only have the opposite effect.
What's needed to achieve the goals of this FIP is to level the playing field between CCs and FIL+ by increasing the incentives for CC production, and not incentivizing FIL+ even more.

[PHABRIC77]

Is there a reason why FIL+ multiplier is set in stone? Why not restructure the incentive model entirely to meet network goals through appropriate economic incentive? For eg could incentivise notaries to perform streamlined and efficient (yet effective) KyC etc as well - any legitimate organisation will expect and respect this happening if not unduly cumbersome. My general point is that if you are going to put forward such a comprehensive change in network incentives why not design it to meet broader requirements that align with network aspirations? Could even incentivise functionality like retrieval etc, and other existing barriers to network growth.

[tmellan]

Trying to understand what the duration incentive in the proposal is doing:

SECTOR_QAP = MIN( MAX_SECTOR_QAP, ( MAX( MIN_SECTOR_DURATION, ( SECTOR_DURATION - SDM_LAG ) ) * SDM_SLOPE * ( SECTOR_SIZE - FILP_PIECES_SIZE // CC or Data + FILP_QAP_FACTOR * FILP_PIECES_SIZE // Fil+ client data ) ) )

As mentioned in the comment above, it's probably sensible to err on the side caution and take a weaker slope (e.g. like 2/7 recommended in the SDM proposal.)

Looking at a 3y CC sector with a duration slope 0.3 I get a duration multiplier of around 0.45. Since this is less than 1 it corresponds to a penalty rather than incentive for longer durations.

Someone else confirm? Could be making a mistake or the mechansim could be off in terms of slope.

[ribasushi]

The MAX_SECTOR_QAP cap in the CDM introduces an interdependency between the SDM_SLOPE and MAX_SECTOR_DURATION. One can not independently change one without the other, so just plugging a different slope is not going to result in a favorable outcome. Moreover the solution space is undefined for an SDM_SLOPE < 1, since this was out of scope. If anything a shorter MAX_SECTOR_DURATION would require a higher slope as demonstrated here: #636 (reply in thread)

[tmellan]

CDM introduces an interdependency between the SDM_SLOPE and MAX_SECTOR_DURATION

Understood. And as a follow on one can't just change the slope as defined in this function.

SDM_SLOPE < 1, since this was out of scope

But it's not out of scope anymore? Or if it is why? Would it be possible to check you've read this comment and agree or disagree so we're on the same page. Thanks

[zixuanzh]

This is an interesting proposal and there are a few elements here that I would like to unpack besides some other points that folks have discussed like the max 10-year lifetime. Each of the element speaks to different groups and gather enthusiasm and support for different reasons. However, I would like to question some of the assumptions to further the discussion. Some of the points here might not be what the authors intended but they are what this proposal implies.

• Reversing Raw-Byte Power Loss Trend? One assumption of the FIP is that it will encourage more CC onboarding and thus bringing the network above the baseline. I appreciate the enthusiasm and optimism to bring the network back above the baseline. However, similar reasoning noted in the extreme scenario analysis here also applies. Rational SPs might prefer to use 1/10 of the hardware capacity to get to the same level of QAP for CC sectors. It is the same "dilution effect" of fiat cost from Fil+ as it is today. Without any major change in global macro dynamics, there is a world where this proposal accelerates the decline in Raw-Byte Power of the Network instead of the other way. For those favoring PoW, this FIP is not working in that direction. If making the network more favorable for PoW is a desirable outcome which I don't disagree, there are other ideas and proposals too.

• Putting Fil+ and CC on Equal Footing? CC and Fil+ are created not to be equal in the first place to create a gradient in incentive to favor sourcing and onboarding data. The 10x is not perfect, the governance is not perfect, and I am sure there are many operational reasons that led Fil+ to what it is today. Fil+ as a community can work together to address their own governance and operational issues. However, what this proposal is doing is implicitly removing the differential that favors Fil+ over CC. If the motivation is to sunset Fil+ or lower that differential, let's have a separate FIP that makes it more explicit. What makes us think that Fil+ will still be around in similar volume and remain competitive after this FIP?

• Clients might pay for using CC? I remember having this discussion even before the launch of the network. A simple question is that what gives us the confidence that clients will pay for the storage when few clients are willing to use the storage even when they are paid to do so?

[dcasem]

Reversing Raw-Byte Power Loss Trend?
Rational SPs might prefer to use 1/10 of the hardware capacity to get to the same level of QAP for CC sectors.

Most SPs have fixed costs (space, power, equipment) that cannot be scaled down. Ultimately, the alternative to locking up FIL via long-duration sectors is a continued downward price spiral on FIL, putting even more SPs out of business and further reducing Raw-Byte Power.

Putting Fil+ and CC on Equal Footing

I like that this FIP accomplishes two birds with one stone, encouraging longer lock-ups and sunsetting FIL+. FIL+ has not achieved its objectives. There have been benefits, but it's likely long-term harmful to the ecosystem. That said, I think FIL+ should continue until a reliable means of getting tokens at scale (particularly long tenors) is developed. Right now, you can't find someone to lend you FIL for eighteen months, let alone ten years.

Clients might pay for using CC

We are able to get users to pay to store with us (and ultimately on the Filecoin network). It's probably harder to convince someone that you'll pay them to store their data -- it almost sounds like a scam.

[scotthconner]

I don't see the logic behind thinking that its rational for SPs to reduce hardware to keep the same QAP. Game theory would suggest that the only one reducing their raw byte power would get disproportionally penalized on block rewards compared to SPs that opted to keep their fixed cost attached to the network.

With a higher CC multiplier, SPs are likely thinking "it's easier to catch up to bigger SPs if I add hardware now, and they stay the same."

I haven't seen any feedback from SPs (maybe I missed it), that suggested that CDM would incentivize them to reduce their hardware capacity on the network. Can I see the logic behind that?

[tmellan]

Let me give a simplified example to illustrate the logic:

• Assume you have 10 FIL to invest.
• Base case: your hardware costs are 10 FIL/year, and block rewards are 10 FIL/year.
• New option: 5x duration multiplier. Block rewards are still 10 FIL/year, but now hardware costs are 5x lower at 2 FIL/year.

Which is the best option, and what happens to raw byte at network level if everyone does the same?

Similar logic applies to SDM, and FIL+ too. If I recall correctly, it was pointed out by Andrew Leece in the SDM discussion, but has popped up a few times elsewhere.

Ultimately the logic is predicated on assumptions of fixed investment, which is for sure pessimistic and not the ‘central scenario’ at all, but still a necessary thing to think about.

This is why SDM is limited to a maximum duration multiplier of 2x, to balance out the expected upside effects on churn, against unexpected but possible downsides.

[dd45e640b42e6da7da96faee3996ef7c]

so we agree that capping the multiplier is a good idea "to balance out the expected upside effects on churn, against unexpected but possible downside" when adding additional multipliers to the network?

[jhookersyd]

This is a great FIP!

My thoughts on FIL+ today:-

• FIL+ is tearing the ecosystem apart
• The FIL+ decentralised governance model has zero power to drive effective change
• "some" SPs/Notaries are abusing the weak/slow governance model of FIL+ today
• Are we really storing useful data with FIL+? Or is 80% of the data just another copy of data that will never get used? Are we cramming large datasets into FIL+ that have no real long term value?
• Re-snapping sectors after getting rewarded new FIL+ data cap for the same dataset will be hugely resource-intensive and a complete waste of time/hardware/energy
• FIL+ was never going to scale
• Hopefully, this FIP kills FIL+

This FIP allows the network/SPs to find REAL datasets as capex/bandwidth/network functionality allows instead of just fighting over Datacap for questionable projects.

I would keep the max duration at 10 years and let SPs/us decide what we want to commit to.

I would love to see the modelling from the CryptoEconLab on this FIP next.

@ribasushi great work!

[dcasem]

Though much of this has already been addressed, I'll add our two cents.

FIL+ has helped us address various challenges, such as moving data across continents, packing files, and distributing deals. However, the current reality leaves much to be desired. SPs paying for deals to increase their Quality Adjusted Power is neither economically viable nor sensible long-term.

Storage/retrieval should not be free. The vast majority pay for it today and expect to continue paying for it. This model provides a nice transition to a more sustainable business model -- where clients pay to store data.

The proposed structure enables us to continue to learn and iterate around operational processes through FIL+ collaborations. It also ensures that SPs who today borrow FIL on relatively short tenors (i.e., 2-year loans) can continue to do so as new means of liquidity, like Glif Pools, emerge via FVM.

Incentivizing away from FIL+ and towards securing the network (block rewards) and storing useful data (legitimate pay-for-storage deals) feels like the right move.

If this results in long-term staking of FIL for long-duration CC sectors, even better.

Nice work!

[dkkapur]

Hey folks - getting a lot of interesting comments here on Fil+ specifically, as was called out in the notary governance call today. Really appreciate all the honest feedback here, would be great to channel it into ideation, i.e., like what JV proposed above, in an appropriate place. Let's use this space: filecoin-project/notary-governance#841.

As someone who's been involved in Fil+ deeply as well as generally engaged in data onboarding to Filecoin, I'd like to highlight the following about this particular FIP proposal:

• A lot of folks calling out the drop in relative QAP for Fil+ vs. CC are missing the major point that it is a 10y period for the returns to look comparable. Let's not straw man the proposal into this singular point and talk through the rest of the pros and cons as well.
• There is a clear path for an SP to elect into being a network-securing only SP. Not everyone is here to just make deals, and it is useful for us to have long-term incentives to continue retaining compute / hardware resources to secure the network and enable new use cases (FVM, etc.).
• We are still ultimately relying on network subsidies / incentives here with additional adjustments to QAP - whether its CDM based or DataCap based. Though it is a step in the direction of bringing forward the timeline for simpler business models to be constructed.
• Sector lifetimes of 10y are compelling but also complicated. I look forward to seeing more on the feasibility of this.
• For SPs looking to scale up - how does this impact business planning vs. network state today or proposals like SDM? Will this change priorities and behavior (fiat committed, time committed, etc.)

[PHABRIC77]

Hi @dkkapur agree with your first point in particular and broadly with your other points, but think your second point should be elaborated on more fully and better understood through this and other discussions. In particular, a clear articulation of why committing capacity is so important as distinct from it being deployed effectively, and what targets there are for % utilization vs committed capacity (both storage/compute) framed in the context of the network's broader goals. From my perspective, I want FC to thrive and to become both the de facto layer 0 for Web3 and a genuine cloud competitor. I think (hope) this aligns with the broader goals of the network participants and am asking these questions because I care.

[dd45e640b42e6da7da96faee3996ef7c]

@PHABRIC77 don't confuse non FIL+ with empty committed capacity sectors. just because there are no FIL+ deals employed in a sector doesn't mean it's empty

at around 9% sectors filled with FIL+ we are in a situation where most of the networks QAP comes not from sectors anymore but FIL+, a (arguably failing) off chain process build on trust. that's then arguably not proof of space time anymore but something very different - and i doubt anyone has signed up for that or would in the future

[PHABRIC77]

Ok. So when we think about CC in this proposal, do we mean unverified deals? If not, my point stands

[dd45e640b42e6da7da96faee3996ef7c]

we mean non FIL+'ed sectors. that can be anything from empty sectors, sectors with non FIL+ deals or sectors with other data inside.

[PHABRIC77]

That’s what I thought. Reiterate my point that a fuller discussion would be welcome.

[jhookersyd]

As a community, we need to move away from trust based governance models and trust based incentive structures.

CDM gives us/SPs clear guidelines on how to grow and sustain a business long term.

Protocol based incentives is why Bitcoin is so powerful today.

Don't trust! Verify!

[dd45e640b42e6da7da96faee3996ef7c]

it's just hard to quantify the energy usage of fiat currency. and how "useful" it really is depends how close you are to the "king" :)

fascinating is how this has a centuries old term and we encounter it in cutting edge technology.

[stuberman]

One would hope that economic principles are universal and timeless.

[SBudo]

(a bit off topic, but.......) If you haven't read it yet, I would highly recommend to read "The Bitcoin Standard"....
https://www.amazon.com.au/Bitcoin-Standard-Decentralized-Alternative-Central/dp/1119473861

[vukolic]

I would highly recommend to read "The Bitcoin Standard"...

That should (will) eventually be part of elementary school curriculum (required reading)

[Fatman13]

That should (will) eventually be part of elementary school curriculum (required reading)

I fear that schools will hardly ever teach people anything about money or being an economically responsible person. At its current state it's more about indoctrination of people into drones. 😂

[woshidama323]

If FIL+ is a transition plan , What would happen if we create a FIP based on a transition plan?

[Fatman13]

100% behind the motivation of this FIP. Hopefully it could contribute to the eventual sunset of f+ on L1.

If it comes down to choose one between SDM and CDM, I cast my vote for CDM.

[ArthurWang1255]

I get the reason why we want to change the minting rule of Filecoin. But I don't see this FIP changes anything. There is a shortcut that simply abuse Fil+ can get a 10x QAP. Storage providers will always be incentivized to pursue Datacap rather than purchasing bunch of hardwares to be managed.

I prefer the idea to remove the 10x incentives from Fil+ project. Make CC and DC same weight on Filecoin.
Nonetheless, Fil+ project can incentivize storage providers to store their data through a tokenomics that is based on its own DataDAO token. Storage providers will be more than happy to mine Filecoin and a DataDAO token at the same time.
If the Fil+ DataDAO has very strong consensus and is very valuable to the storage clients, then the Fil+ DataDAO token will be very valuable and more storage providers will be incentivized to join the Fil+ DataDAO by storing Fil+ deals while providing storage capacity to Filecoin network.

We shouldn’t confuse the consensus of Filecoin and the consensus of Fil+ project. Let’s make these two consensus clear and independent with each other so that people can pick their side freely.

[ArthurWang1255]

In my view, the consensus confusion is the root reason storage providers are hesitated.

[ArthurWang1255]

I believe the team has sensored this distraction of consensus already:
On the Filecoin webpage(filecoin.io), it says: ”Filecoin is a decentralized storage network designed to store humanity’s most important information”.
However, in the recent Filecoin Master Plan, it has changed the words to "Onboard and safeguard humanity’s data"

The key difference/point here is that, Filecoin as a protocol, is not in a proper position to define what is the “most important information”.

[dcasem]

After being on the North American SP call, I propose making the CC equivalent to FIL+ three or five years rather than ten. We would likely be hard-pressed to find willing participants to lock up FIL for a 10-year period.

Never mind, I understand that SDM is being pursued irrespective and that this entire thread was an exercise in futility.

[LinZexiao]

It seams that we are trying to reduce the impact come from the abuse of Fil+, but I don't think this proposal can solve the problem fundamentally.Maybe we should make some change to fil+ instead.

[AxCortesCubero]

I'm a bit late to this discussion, but I don't think I've seen this mentioned yet:

It seems a main motivation of this proposal is the opinion that FIL+ is currently over-incentivized and CC power should be better balanced. There are some valid arguments in the discussion that support this idea, and there seems to be a good amount of community support and interest in it.

That said, if the problem statement is: "FIL+ is currently over-incentivized",
I wonder why wouldn't we consider first the simplest and most direct solution of just decreasing the FIL+ multiplier from its 10X?

This proposal seems to try to solve this problem in a less direct way, Using duration multipliers to try to make CC more competitive against FIL+.

If we think CC should be more competitive, the straightforward solution should be to lower FIL+ multiplier. Then the next question, "do we want to incentivize longer durations?" could be independent.

[ribasushi]

@AxCortesCubero I can obviously only speak in very broad strokes ( and I DO NOT run an SP ), but I personally think the road from "screw CC miners" last year to where we are today has several causes:

• The realization that having the free market (correctly) price Fil+ onboarding as zero-or-negative, is in direct existential conflict with the ability to charge for services
• Various failure modes of the current Fil+ system becoming acutely apparent ( see Governance Team Statement on December/January Community Discussions notary-governance#819 )
• The baseline crossing event has affected every SP and forced a realization of "oh, hardware security is important"
• The interest in FEVM extended that understanding of needing a security source ( network is not securing just Fil anymore, it will hold financial instruments in multiple diverse currencies )

[AxCortesCubero]

Given this list, I still think the direct answer to the problem you are describing is to propose lowering the FIL+ multiplier from 10X. If the reason to not make that proposal is that we think it would fail, then I would say just to propose it and see what the community thinks about it, not much point in trying to predict when we could just ask the community.

In terms of the Compromise @vukolic is trying to mediate, I think the biggest issue I don't agree with about this proposal, is as I explained earlier, about ordering of incentives. If we agree that having FIL+ in your sector is preferable than not having FIL+ data on your sector and should be incentivized, I don't know why that statement shouldn't hold true for 10 year sectors.

I can understand the concern that we could worry about simply multiplying multipliers together, which can get very big quickly. My two compromise possibilities that I envision could get us some percentage of the way there:

1. SDM FIP plus explicitly lowering FIL+ multiplier from 10X

or

2. SDM with non-multiplicative combination of multipliers. The multipliers can combine in some way that grows more softly, but still having FIL+ in your long term sectors would be the best thing you can do.

[PHABRIC77]

Great to see a constructive attempt to come to a compromise. Could have a duration curve for CC and for FIL+, and the question then becomes the appropriate level of incentivization (and the requirements) for each duration. That is something which will likely need to be determined via the formal process. A softer multiplier (or none at all) could make sense.

[nai2283]

Given this list, I still think the direct answer to the problem you are describing is to propose lowering the FIL+ multiplier from 10X. If the reason to not make that proposal is that we think it would fail, then I would say just to propose it and see what the community thinks about it, not much point in trying to predict when we could just ask the community.

Lower the FIL+ multiplier won't change the fact that CC SPs have negative profit, this one would be a really bad idea, as it on one hand will reduce the overall pledge by de-incentive DC SPs, on another hand won't achieve the goal of better security by higher hardware commitment this FIP wants to achieve IMO.

[Fatman13]

Given this list, I still think the direct answer to the problem you are describing is to propose lowering the FIL+ multiplier from 10X. If the reason to not make that proposal is that we think it would fail, then I would say just to propose it and see what the community thinks about it, not much point in trying to predict when we could just ask the community.

Given how FIP is governed right now, it would be exceptionally draining to have such undertaking by individual or SPs. CEL would be the only body that is resourceful enough to push something like this through. (Like FIP0056 which is about to get accepted even after major pushbacks)

[danlessa]

Hey'all. BlockScience has worked to perform a first-principles-based review and compare the FIP-0056 (i.e., SDM) and CDM proposals. It has been instigated and funded by the Filecoin Foundation due to the controversy generated by recent proposals and the consequent need for an independent technical group to read through the arguments, check them, and evaluate their implications. The full article can be found at Medium. Our Executive Summary, Diagnosis and Recommendations on the whole FIP-0056 / CDM discussions are as follows:

Executive Summary

• On the Proposals Technicals
  • Both FIP-0056 and CDM favor Long-Duration CCs with different patterns and intensities compared to the Current Active Algorithimical Policy (Now). CDM does so heavily, while FIP-0056 does so slightly. FIP-0056 can be better for long / fully-filled FIL+ sectors and CDM for shorter / slightly-filled FIL+ sectors. FIP-0056 doesn't actively disincentivize any sector, while CDM does so: it generates a virtual penalty for having too much FIL+.
  • Both proposals can increase the amount of Locked Filecoin for 1-2 years, with CDM having a more substantial effect.
  • The Termination Fee Cap Modification in both proposals should be removed and proposed as a distinct FIP.
  • CDM can take either the proposed or an FIP-0056 like parametrization. The latter will have an more balanced QAP Share among participants and less effect on the temporarily increased Locked Filecoin .
• On the debate and general recommendations
  • The friction in the process could be improved by segmenting it according to the steps described in the Computer-Aided Governance in Action. If FIP-0056 is to be postponed/rejected, we recommend using it as the source document for the workflow around a new proposal so that there's a soft agreement between requirements, forms, and parameter values on all steps of the process.
  • The debate has exposed multiple concerns. All are legitimate in intent, but not all in factual basis. We provide a extensive review in this document.
  • FIL+ has been a focal point of disagreement among participants. We recommend that the community has space for open discussions around its role and functions and how it fits into the various Vision(s) for the Network.
  • Community Oversight of the analyses surrounding FIPs and additional Educational Efforts are to be suggested.
• On the recommendation around the Decision
  • We recommend Postponing or Rejecting FIP-0056 as per application of the Precautionary Principle on the Technological Hurdles & RBP as Key Advantage thesis as advocated by the CDM Proponents. If that thesis is taken to be Unlikely or False, then the Recommendation is Void towards FIP-0056.
  • An underlying thesis behind CDM relates to technological rather than economic hurdles being blockers to Filecoin Adoption. If Filecoin's RBP size is considered a key advantage, then there's a catastrophic risk of interfering with the CC's economic calculus. With the currently available information, we are unable to disprove that Thesis.

Diagnosis

• On the technical comparison of the FIP-0056 and CDM proposals
  • Both proposals will have long-duration CCs as their primary beneficiaries and short-duration CCs as the main negatively affected party.
  • Contrary to the discussion, FIP-0056 does not bias QAP towards FIL+. Unless CCs are disproportionally risk-averse, FIL-0056 is CC-biased compared to Now.
  • FIP-0056 will favor mostly long-duration FIL+ sectors, especially those filled with FIL+. CDM will disproportionally shift incentives from fully-filled sectors with FIL+ toward other parties. Partly filled FIL+ can benefit from CDM in several circumstances.
  • Both proposals can potentially increase Locked Filecoin Supply, with CDM being stronger than FIP-0056.
  • An incentive flaw has been encountered regarding the Termination Fee Cap in both proposals, which is even 'quirkier' in CDM.
• On the Debate
  • The debate process exposed multiple concerns. Some were legitimate in intent and factual basis, while others had honest intent but weak or even misleading factual basis.
    • Most of the latter did fall into two categories: 1) FIP-0056 skewing the Token Wealth Distribution even more towards FIL+ sectors and 2) mistaking Network-wide phenomena with Individual phenomena (e.g., lower network-wide rewards means lower rewards for the individual miner that remains / onboard all things being equal, which is not true).
    • The debate made clear that the FIL+ incentives are a central point of contention in the public debate as of now.
    • A misconception is the equivalence between FIP-0036 and FIP-0056. The latter doesn't modifies the Consensus Pledge, and it has a far gentler slope.
  • Issues around study reproducibility and perceived relevance towards stakeholders as consumers of studies were identified.
  • A concerning thesis has been propagated by the CDM proponents, which, if true, poses a catastrophic risk to the Network's Success. Given the impossibility of disproving it and the perceived non-criticality of the FIP-0056 modification, we opt to employ the Precautionary Principle for the recommendations.

Recommendations

• On the Proposal's Technical Structure
  • FIP-0056: Remove the Termination Fee Cap dependence on SDM from FIP-0056 and make it a separate FIP.
  • CDM: Consider FIP-0056 equivalent parameter choices
  • CDM: Remove the Termination Fee Cap dependence on SDM entirely.
  • CDM: Consider soft transitions rather than caps on the Quality Multiplier.
• General
  • Community Oversight on the FIP Analysis should be increased through either external stakeholder involvement during its execution or usage of independent FIP reviewers
  • Educational Content coverage on Filecoin's Economics and Protocol must increase Quantity, Reach, and Quality.
  • The community should be stimulated to discuss the Storage Market in terms of its state, challenges, benchmarks, and required actions & incentives. The inherent risks and the role of Filecoin's endogenous economy around the transition towards a Storage Marketplace / FVM-dominated world should be the topic at large.
• If FIP-0056 is not approved
  • Go back to first principles: formally describe the problem that should be solved (e.g., 'we should incentivize longer sectors,') and what mechanisms can be used to support that. Select the mechanism(s) based on diversity-of-interest criteria: the best one is the one that can be tuned to manage trade-offs across different utility functions.
  • Break down the solution proposal into two FIPs:
    • The first should decide the mechanism (e.g., 'we should have a Sigmoid SDM regardless of the numbers). This determines the Class of Incentives that are achievable.
    • The second should determine the actual numbers (parameters) that tune the mechanism. This instantiates the Actual Incentives that are to be realized.
• On Taking a Decision Immediately
  • FIP-0056 should be postponed or rejected if the "Technological Hurdles & RBP as Key Advantage thesis" validity is considered Unknown or Likely to be True.
    • FIP-0056 does not appear to be intrinsically 'bad' or 'good' for the FIL ecosystem. However, in Engineering work, the precautionary principle always applies. If the expected benefit doesn't clearly outway the potential risks, postponing is always a safer option.
    • The same precautionary principle applies to CDM as of now.
    • A possible concession is to utilize the more general proposed CDM form and set it so that it has FIP-0056 behavior: Ceiling on 20x and slope = 0.287. The definition of those two values could be the subject of a poll.
  • CDM, as being proposed now, should be modified before being considered
  • We could not find a compelling public argument making the case that FIP-0056 is an urgent intervention.

[dkkapur]

@danlessa thanks for sharing this. quick potential correction in this line:

The former doesn't modifies the Consensus Pledge, and it has a far gentler slope.

I think this should be "the latter" since this refers to 0056 not 0036, correct?

[danlessa]

@dkkapur that went unattended for. Thanks for pointing it out

[dd45e640b42e6da7da96faee3996ef7c]

dear SPs, governance team and everyone else:

i want to give you an update on the state of the CDM FIP proposal, the authors current intentions, possible pathways for an implementation and some general thoughts about what was going on in the last few month regarding economic FIPs (SDM in its various forms and CDM).

So, why an update now?

We, the authors of the CDM FIP discussion opening thread (#636) and FIP proposal (#635) got asked, roughly 14 days ago, whether we would like to present the CDM proposal in front of the core devs (https://github.com/filecoin-project/core-devs) in the 5th of May meeting. We declined to do that and i will lay out why we decided to do so:

• we do not think the proposal in its current form could move forward to be implemented and a presentation in front of filecoins off chain decision committee for protocol changes, the core devs, would possibly set things in motion that should not be set in motion with the currently published proposal
• the last few month revealed significant gaps in the governance process, prompting the need to adjust the proposal on our end
• parts of the proposal, namely the SDM of FIP56, got rejected by the core devs
• other FIPs got accepted and implemented in nv19/20 and the coming nv21 upgrade that will have impact on the chains states development

We are still convinced that a capped duration multiplier is a mechanic that would move the chain forward in a direction that is favorable and we will go on working towards it's implementation.

A word on FIP56 (and to some extend about FIP36). the core mechanics of FIP36/56 could be broken up into a few elemental changes to the chain:

• an incentive for SPs to pledge sectors longer (the DM, duration multiplier, core of the series of FIPs - FIP36/56)
• raising the lockup target for pledge (FIP36)
• adjusting the termination fee mechanics to accommodate a DM (FIP56)
• [infinite] stacked multipliers (FIP36/56)

We think that the only reason, and the comments on the proposals and FIPs support that, FIP36/56 got a bad reception from SPs is the stacked multiplier concept, the only constant in the revisions of FIP36/56 with their respective parameter changes.

The CDM addresses that fact by proposing a general multiplier cap - defusing the dangers the stacked multiplier concept would have introduced to the chain. DM incentives are a core part of the CDM as well, adjusting the termination fee mechanics too. why changing the pledge lockup target got left behind with FIP36 is not immediately clear to us. we assume a misconception why FIP36 got "rejected" by SPs plays into this - we had early drafts of the CDM including the mechanic too as it seemed to make sense at that time. Now, with FIP56 rejected, the underlying work/mechanics for the CDM got rejected too - the CDM was always intended as superset of FIP56, borrowing its DM and termination fee mechanics.

That's why we (not necessarily only the authors of the CDM discussion/FIP proposal) intend to break up the CDM proposal into multiple independent FIP proposals to reduce complexity.

1. a multiplier cap that aligns with the current state of the chain: max. 1:10, raw:quality-adjusted power
2. adjustments to the termination mechanism (to prepare for the DM) (we encourage you to participate in the existing discussion here - the termination fee discussion can and should take place independent of a DM - we will certainly engage with the teams already working on adjusting the termination fee mechanics)
3. a duration multiplier incentive

Splitting the CDM proposal in these distinct parts will allow the community to streamline discussions, allow more atomic changes to the chain and, hopefully, allows us to advance the governance process.

A few closing notes:

• while i will certainly have this post pre screened by the authors of this discussion thread it might be that i do not refer to the authors in general when using "we". All of the above is discussed in boarder groups, SP especially, and i/we want to encourage everyone to take part in these important discussions. (if you are a SP and want to work more actively on moving the protocol forward please feel free to contact me on slack f8-ptrk). We will engage with all parts of community in the process of getting a CDM into the protocol again to, hopefully, get the desired governance process on the way - everyone else please feel free to contact me too for discussions, idea exchange, criticism or whatnot!

• while QAP as a concept is a nuisance at times it in itself is not the problem. the way off chain programs are implemented to utilize QAP in line with FIP003 are the problem. these programs, if not fixable by their respective off chain controls, need to be reigned in on via protocol changes. the CDM proposal itself is not intending to do that. It will need careful consideration on potential future FIP proposals in that direction to strike a balance between allowing incentive structures and containing the potential harm power multiplier utilization can do to the L1, the core of filecoin everything else builds on.

At last:

please allow us a few weeks before engaging actively again in the public forum regarding the implementation of the CDM.

Thanks for you time, passion, work and efforts! they are appreciated.