Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Bisq v2 #118

Closed
chimp1984 opened this issue Sep 9, 2019 · 23 comments
Closed

Bisq v2 #118

chimp1984 opened this issue Sep 9, 2019 · 23 comments
Labels
an:idea https://github.com/bisq-network/proposals/issues/182#issuecomment-596599174 was:stalled

Comments

@chimp1984
Copy link

chimp1984 commented Sep 9, 2019

I would like to give a rough overview and the more broad picture how the off-chain trade protocol would impact and change Bisq. In fact I think it will be such a radical change that it will require to be deployed as a new application and justifies the upgrade to version 2. The current trade protocol can continue in parallel and might fade out one day when most user have migrated.

The core change is that the trade protocol is decoupled from the security protocol.

Security for the trade is based on a BSQ bond. Once a bond is set up it can be reused for multiple trades in parallel up to the amount of the bond. E.g. You set up a 10 000 BSQ bond and start one trade with 4000 USD against 3600 EUR (no need that the trade contains any crypto currency). Lets assume 1 BSQ = 1 USD. So your available bond for further trades is now reduced to 6000 BSQ. You do another trade with 5000 USD and you still have 1000 BSQ left. You cannot do any trade > 1000 USD now until you have completed one of the open trades. After that, the amount from the completed trade increases your available bond again.

The bond/trade tracking is based on cryptographic pimitives like hashing and signatures and uses the Bisq P2P network. It is described at the end of the #32 proposal (and require still some work to make sure it is solid).

The way how we use bonds has some similarities to opening a channel in Lightning network.

The trade protocol itself is trivial. You make an offer and define the mediator to be used in case of a dispute. A taker agrees to your contract and then both have to exchange their currencies over the payment method defined in the contract.
In case of a dispute the mediator investigates and delivers both traders a signed result which states who violated the contract. That can be used for a confiscation request and the BSQ stakeholders would take the mediators result as orientation for their vote. This human barrier gives extra security against malicious mediators (they risk their bond).
Confiscation is considered an exeption and should happen very rarely. It is similar like ending up in court in normal business relationships. Its an exeptional outcome and both partied usually try to avoid that as far as possible.

With this new model we get tons of benefits and it resolves so many problems which seems basically unsolveable with the current model:

  • You can trade anything again anything: XMR-EUR, USD-EUR, BTC - lightning to LTC - lightning, Apples-Bananas....
  • There is no blockchain tx required for the trade, so no miner fee and no privacy leaks via connected BTC transactions.
  • Trade execution can be instant - no waiting for blockchain confirmation.
  • Offer maker don't need to be online anymore as no automatic tx signing is done.
  • We don't have the requirement for a trading wallet and automated tx signing anymore. This removes our biggest security risk. We should discourage traders to use the Bisq wallet for anything else than for miner fee for BSQ tx and converting BTC to BSQ. We likely can get rid of BitcoinJ.
  • Exchangin BSQ can be done in an atomic tx. 1 BTC input, 1 BSQ input, outputs are swapped. Totally trustless and secure as it is one atomic tx. Trading BSQ-BTC can be presented to the user as conversion rather than a trade. No need to show complexity of offers, etc. You want to get 1000 BSQ? The apps shows how much BTC it costs based on current offers. You click ok and you execute the offers to get your BSQ. E can probably use presigned txs so no need for the BSQ offer maker to be online.
  • It gives as well more security against chargebacks as bonds are locked up long term.
  • Mediator model is flexible. Can be a Bisq mediator, lawyer...

There are also some downsides and risks:

  • New user need to buy BSQ to cover trade amount (there might be some ways to mitigate that by using some sort of reputaion for security, e.g. use certificates like in Self certification of bank account details using ID verified digital certificates #79).
  • It changes the economic model of the DAO. A small feature will become major price factor. We need to think very well about possible unintended consequences and risks.
  • BSQ price volatility becomes a security issue at some extent.

Roadmap

  • First we need to finalize and audit the bond/trade tracking concept. This is the core of the off-chain trade protocol and it need to be 100% solid.
  • If that is completed we should get a good idea how we implement the new trade experience and the BSQ/BTC conversion so that user experience is much better as now. @pedromvpg, @ripcurlx are very welcome to help here.
  • We need to find out what is the best deployment model. Currently it seems the best is to deploy a new Bisq v2 application completely separate from the old app. The P2P network and the DAO are the only connecting elements. As it is a fresh new app we can clean up lots of stuff and think about more radical changes (e.g. get rid of BitcoinJ,...).
  • We need some notifiaction mechanism as traders dont need to be online. Maybe we can use the existing mobile notification but as this feature becomes more important we might need to develop some more robust tool (current has not guarantee for msg delivery).
  • Ensuring that the P2P network is compatible to support both applications might be a bit of a challenge. To start a new network and use a bridge node for transferring DAO data between the networks might be another option. We also plan some bigger changes in the P2P network to make it more ddos resistant which would be easier if we don't have dependencies on existing network.
  • We need to find the best model for trade fees. As the trade does not require a tx we dont want to add a tx for the fee. At unlocking of a bond we could maybe apply a fee. But will require more thought once the protocol is more defined. Should be the easiest problem to solve ;-).

UPDATE:
The protocol for keeping track of bond usage need more time for design and analysis. Until that is completed that proposal is on hold.

@battleofwizards
Copy link

battleofwizards commented Sep 10, 2019

Why do we need specifically BSQ token for bonding? Because we need confiscatability by the DAO.

Ideally though users should be able to bond BTC.

But then we lose confiscatability. Or do we? What if we invented a way to make BTC bond confiscatable by the DAO? Not even sure it's possible - just food for thought.

@chimp1984
Copy link
Author

chimp1984 commented Sep 11, 2019

What if we invented a way to make BTC bond confiscatable by the DAO

Burning BTC is possible but the economic effect is different. With BSQ it goes back to stakeholders. With BTC it goes to miners or icase it get destroyed (e.g. by sending as op_return), then it goes to all BTC holders as the total supply is a bit smaller. As the confiscation of a bond comes usually with a reimbursement for the victim it will be a zero sum game for the DAO (x BSQ confiscated/burned - x BSQ issued by reimbursement). With BTC we would not have that symmetry and all BSQ stakeholders would basically pay for a reimbursement due the supply increase/inflation.

@chimp1984
Copy link
Author

We need to find the best model for trade fees. As the trade does not require a tx we dont want to add a tx for the fee. At unlocking of a bond we could maybe apply a fee. But will require more thought once the protocol is more defined. Should be the easiest problem to solve ;-).

I think the easiest way to apply fees is to use the proof-of-burn feature. After x trades (e.g. 10 trades) the user needs to pay the accumulated trade fees in one batch tx by burning BSQ and delivering a proof-of-burn to the next trader. The protocol for tracking the bond need to support that so that it is required after x+1 trades to attach that proof of the fee payment, otherwise the peer would reject the trade. Better would be that each following trader verifies the fee payment to avoid that a self-trade with a manipulated code could be used to avoid a fee payment.

@battleofwizards
Copy link

battleofwizards commented Sep 11, 2019

With BSQ it goes back to stakeholders

Yes, that is a very good point and very nice property to have (a symmetry of burned vs reimbursed).

Although from an outsider perspective this will look like we force buying and hodling BSQ onto users (which is arguably suboptimal to hodling BTC). I'm afraid that never mind how good is our justification the question "why can't I simply bond BTC" will pop up over and over again. I'm worried that even the Bitcoin OGs may not have the inclination to understand and defend the BSQ position.

Tricky!

@battleofwizards
Copy link

battleofwizards commented Sep 11, 2019

After x trades (e.g. 10 trades) the user needs to pay the accumulated trade fees in one batch tx by burning BSQ and delivering a proof-of-burn to the next trader.

Wouldn't one rather reset the identity than pay the "surprise 10-nth trade fee"?

If we want to batch fees then maybe pre-paid would be better?

Maybe fee payment batching should be optional to avoid trades linkability?

@sqrrm
Copy link
Member

sqrrm commented Sep 11, 2019

The bond is already linking trades together so a fee payment linked to that bond would be equally linkable either as individual fee txs or as batched fees.

I think it makes sense to do it similarly to the proof of bond, by adding a salted hash of the proof of burn such that only trade peers can verify that the fee is paid upon presentation of the salt. After trading with a lot of peers there will however be a lot of peers that can link those trades together, but keeping it at the same level of linkability as the bond seems reasonable.

Maybe there is some crypto magic that could be done to avoid linkability of the trades through the bond (and fee) with zero knowledge proofs. This is outside of my are of expertise.

@bodymindarts
Copy link

@sqrrm are you suggesting that the participants must prove to each other that the fee has been burned before executing the trade? Do you mean for past trades that the bond was used for or the current trade?

@sqrrm
Copy link
Member

sqrrm commented Sep 11, 2019

are you suggesting that the participants must prove to each other that the fee has been burned before executing the trade?

@bodymindarts Yes, that's my thinking. For the bonding to work the peers need to verify the bond of the counterparty. That includes checking the current outstanding liens on the bond which necessitates checking all liens on the bond and verifying which ones are currently in effect. Similarly the batched fees would not be tied directly to a trade so the complete history of trades and their respective fees would have to be calculated to compare to the fees paid for the bond.

The other option is to specify individual trades when paying the fee which would be impossible to do for pre payment of fees. Post payment would be able to specify the trades but it would then be too late for the past counterparties to verify that the fee was paid. At that point the whole history of trade fees need to be shared with future counterparties, essentially having the same effect as the non specific trade fee payment.

@bodymindarts
Copy link

bodymindarts commented Sep 11, 2019

I think 1 fee burn per trade per party is certainly less complex to understand for the user and probably to code. It does however open up potential for collusion.

@battleofwizards
Copy link

battleofwizards commented Sep 11, 2019

I think as a general principle (and not limited to v2) we should attempt to make user trades as unlinkable as possible. At minimum, we should have plausible deniability.

I do recognize I am making already hard problem even harder... :(

@battleofwizards
Copy link

The proposal assumes reusing a single BSQ bond for multiple trades - both in amount dimension and in time dimension. The rationale (I guess) is to cut down on Bitcoin fees. Bitcoin fees are expected to skyrocket so this is very important consideration (and colored transactions are even more expensive).

However, if we forget about tx costs for a second, we could consider a bond-per-trade. Then suddenly everything seems easy. No linkability. Trivial to apply fees. Less potential for security critical bugs. Easier to understand for the user. Easier to code (although not exactly? should be compatible w/ future batching?)

Maybe privacy by default, bond reuse as option?

@bodymindarts
Copy link

@battleofwizards 1 bond per trade is always going to be possible. Users who don't want their transactions linkable can create as many bonds as they like (including 1 per trade).

@battleofwizards
Copy link

battleofwizards commented Sep 11, 2019

@bodymindarts it's all about what your encourage in the UI as default or recommended flow.

@chimp1984
Copy link
Author

@battleofwizards If users don't take care of on-chain privacy (coin merge) they would make one bond linked to the next bond. That is even worse with BSQ than with plain BTC because you cannot use CoinJoin for BSQ.
The way how the bond tracking works need to ensure privacy so that trades are not linkable. The fees should not add more requirements, but it is not thought our yet how it will be implemented exactly. Even the core bond-tracking protocol is not finalized but in early concept phase. Hopefully in a few days we have a version for security and privacy audit ready.

@chimp1984
Copy link
Author

chimp1984 commented Sep 11, 2019

@battleofwizards
Re: Bitcoin maximalists having problems with BSQ.
Do we really care today about people who had problems in the early Internet days that the Internet was used for something else than sending "serious" messages from one university to another (e.g. starting to send images, videos, bittorrent). As the Internet was not just a digital mail or newspaper, Bitcoin is not just digital money but much more and we have not even scratched the surface of what it can be used for in the future. Narrow mindset will not lead to those future applications. The Bisq DAO is already a first step out of the pure money mindset to use a Bitcoin token to build a more complex value transfer and management system.

Update:
An interesting point of view is also if you apply the maximalist mindset to the early Internet. If people would have considered internet only to be used for exchange information between scientists (as it was initially designed for beside the military branch) they likely would have rejected to support transfer of monetary data like Bitcoin. So Bitcoin maximalists would not exist as no Bitcoin would exist if you apply that narrow mindset on the core layer Bitcoin is based on - the Internet.

@flix1
Copy link
Member

flix1 commented Sep 16, 2019

My concern about bonding and DAO confiscation of the bonds is that it is not very scalable. Imagine having 1 million users and 0.1% of them being scammers... 1000 scammers could create 1000s of conflicts a month... would we need a DAO confiscation vote for each? It could easily turn into a DDOS attack vector.

With a pure Mutually-Assured-Destruction 2-of-2 multisig protocol you could have bonds (in BSQ or BTC) without the need for the DAO to intervene.

I believe that with:

-2-of-2 multisig security deposit (up to 110% of trade amount)
-Mediation
-Web-of-trust (identity-less reputation)
-Account age filters

You could have a safer, more scalable system which also requires less radical changes to Bisq...

@flix1
Copy link
Member

flix1 commented Sep 16, 2019

Having said that... decoupling the trade protocol from the security protocol has huge advantages on other fronts... for example for the creation of decentralized financial instruments. You could build an uncensorable FX platform on Bisq 2.0 with CFDs, options, futures, etc..

If it can be made scalable...

@mpolavieja
Copy link

DAO confiscation of the bonds is that it is not very scalable

Agree on that, but probably the risk of losing the bond would discourage scammers from such DDoS attack. I am more concerned about legal liability of BSQ holders, as it is almost certain that in the case of legal conflict the Bisq DAO will never be considered a limited liability entity in most jurisdictions, so the BSQ holders would be directly liable.... This is quite disturbing.

Maybe a fee structure like the following would ensure that the DAO almost never has to decide on trade reimbursments?:

  • Cheapest (Default): 2of2 multisig without mediator and without DAO recourse
  • Medium: 2of2 multisig with mediator and without DAO recourse
  • Expensive: 2of2 multisig with mediator and DAO recourse. Only available for trades greater than X BTC.

@chimp1984
Copy link
Author

chimp1984 commented Sep 16, 2019

0.1% of them being scammers... 1000 scammers could create 1000s of conflicts a month...

Those 1000 scammers will quickly learn that it was a huge mistake to try to scam. The DAO stakeholders are not judges, just executors and only in weird cases where the reliability of mediations get questioned they would need more effort/time.
Beside the stolen-bank account issue in April we had basically zero scams as in the current model scammers have no chance. Specially with altcoins it is zero (fiat has more weaknesses in itself).
Traders agree on a mediator and accept his conclusion when entering a trade. So they cannot dispute that mediators decision later.
I think it will be even more safe as currewnt arbitration as the "skin in the game" is higher. Now security deposit is a few % of the trade amount, bonds are > 100% so trying to scam will hurt much more.

@chimp1984
Copy link
Author

@mpolavieja
Legal liability of DAO stakeholder is a difficult question. In a way they are an open permissionless multisig group who can detroy your funds if a super majority agrees. The incentives align that abuse is not realistic, but theoretical and legal interpretation might be different. Hard to find parallel models in real life so likely the legal questions will remain un-answered.

@mpolavieja
Copy link

mpolavieja commented Sep 16, 2019

The DAO stakeholders are not judge, just execution

That is tricky. In order for this claim to succeed all the process and the wording should be carefully designed.

Regarding the process, the decission of a mediator should never be questioned by the DAO. In case of a Mediator going rogue, its bond should be confiscated afterwards, and then another reimbursment requested in next cycle based on the Mediator having been banned (but then the DAO is making a decission on the reimbursment so probably another mediation process would be needed).

Regarding the wording, a reimbursment request should be made to the mediator instead of to the DAO, and the DAO request should be a "Mediator decission execution request" instead of a reimbursment request.

In any case, the bond confiscation processes are always a DAO decission, aren´t they?

@chimp1984
Copy link
Author

The decisions are only executable by DAO voting. Mediators give their judgement (like a judge) and the DAO executes it (like police) if they dont have any doubt on the mediators quality of work. It is a feature that this is not automated and a human layer is in place as a corrective. So DAO voters should make their due diligence but they are not investigating the case. But a mediator with many weird cases and much public disputes around their cases needs close attention.

Wish more policemen would question the decisions of their authorities as well before they beat up peaceful demonstrants (who pay their salaries with their taxes...) - but thats another topic ;-)....

@MwithM MwithM added an:idea https://github.com/bisq-network/proposals/issues/182#issuecomment-596599174 was:stalled labels Aug 14, 2020
@MwithM
Copy link

MwithM commented Aug 14, 2020

Closed as stalled

@MwithM MwithM closed this as completed Aug 14, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
an:idea https://github.com/bisq-network/proposals/issues/182#issuecomment-596599174 was:stalled
Projects
None yet
Development

No branches or pull requests

8 participants
@flix1 @bodymindarts @mpolavieja @sqrrm @MwithM @battleofwizards @chimp1984 and others