scaling bitcoin – Data Science, Data Analytics and Machine Learning Consulting in Koblenz Germany https://www.rene-pickhardt.de Extract knowledge from your data and be ahead of your competition Tue, 17 Jul 2018 12:12:43 +0000 en-US hourly 1 https://wordpress.org/?v=4.9.6 Virtual Payment Channels for the Lightning Network – A Lightning Network Improvement Proposal draft https://www.rene-pickhardt.de/virtual-payment-channels-for-the-lightning-network-a-lightning-network-improvement-proposal-draft/ Tue, 17 Jul 2018 11:07:50 +0000 http://www.rene-pickhardt.de/?p=2131 I suggest to add another feature to the lightning network which I call a virtual payment channel (or in short VPC) to be described in this article. Such a virtual payment channel will NOT be backed by the blockchain and thus cannot operate in a trustless manner. However I see several use cases for virtual payment channels. The most important one seems to be the fact that two parties which agree to create a virtual payment channel can instantly create a channel at basically arbitrary capacity. This will help to increase the max flow properties of the network and allow for routing to become much easier. A virtual payment channel basically allows both channel partners to access the funds of the other channel partner which yields some flexibility and obviously a high risk in case of fraudulent behavior. While writing down the article I realized there are also some dangers to virtual payment channels. The main one is that to some degree VPCs resembles the fractional reserve banking system even though it is still only credit based on positive money (I think the English language should borrow the German word “Vollgeld-System“). At first I considered dropping the article because the last thing that we need is a cryptocurrency crises similar to the banking crisis that we have been through. However I think when regulated and used properly VPCs will add more benefit than danger to the lightning network. Despite the drawbacks with this article I will still propose this feature.

I came up with the concept of virtual payment channels since I finally started my work on a splicing protocol specification for BOLT1.1. On paper I have already laid out the data flow for asynchronous, non blocking splicing in and splicing out operations – which resemble pretty much the ones discussed on the github issue. I will write about them soon. Before digging deeper into laying out my specification I wanted to increase the usefulness of the splicing operations. So I started to wonder if there was a way that splicing could take place without the need to ever hit the blockchain? While playing around with some ideas I realized (though not proven) that splicing without hitting the blockchain seems impossible if the trustless property for payment channels is mandatory. Hence I asked myself what would happen if we dropped this central property? As a result I accidentally came up with the concept of virtual payment channels.

Having talked to larger exchanges and potential hubs of the lightning network in the past and having seen that they wonder a lot about rebalancing their funds and about how they could provide capacity without locking their liquidity uselessly I realized that virtual payment channels might be a solution to some of their core problems. During the article and at the end of this article I will provide use case stories from the perspective of a central node in the network as well as from the perspective of a mining pool.

Let us begin by looking at one of the two core elements of the lightning network: The (standard) payment channel. A standard payment channel operates in a trustless manner which allows two parties to create a payment channel without the need for them to trust each other. The ability to create trustless payment channels is certainly a very useful property that the lightning network must have. However this trustless property comes at a rather high price bringing us the following three drawbacks (which should still exist even if we upgrade payment channels to follow the eltoo protocol proposal):

  1. Even when Splicing is implemented channels need to lock the bitcoins that create the capacity for the channel. Meanwhile those bitcoins cannot be used differently making the lightning network and its topology somewhat rigid. (In particular if you take into consideration how much topology could at most change within one bitcoin block.)
  2. Creating (and closing) the channels (and splicing funds) are blockchain transactions which in general are slow and expensive. Slow because we have a block time of roughly ten minutes. Expensive because independently of the block size discussion there is a physical limit of how many transactions fit into a block. Remember that space in the block for a transaction is generally bought by providing mining fees.
  3. Bitcoins within the payment channel are in a hot wallet and it is difficult to secure them properly. Even if the output of the commitment transaction script is spendable only by a cold wallet address we have no guarantee that an attacker taking over the lightning node(s) might mitigate a different output for the next commitment transaction (for example if splicing out of funds is being implemented) or even easier the attacker could just make a payment to some address emptying the channel.

Additionally an indirect problem will be that companies like exchanges, large scale shops or payment provider or even banks will have many payment channels since they are likely to be central nodes in the network. The amount of channels might exceed the capacity of the hardware running this node. So people providing core financial infrastructure might need to operate several lightning nodes (let us call them A,B,C,…) to serve their customers. Companies then have to decide by collecting and evaluating data which customers to connect to which of their nodes. They also have to connect their own nodes via largely funded payment channels among each other. Even if done in some smart circle like topology (with channels so that each node can be reached in log N hops) it is not the preferable way since way too many funds are unnecessarily being locked to maintain clearing of funds between ones own nodes. Another drawback is that those additional funds are also on hot wallets again and need to be available in the first place.

Let us assume in the above setting some node N1 wants to make a payment to some node N2. The route for the payment happens to be:

N1 –> A –> B –> N2.

In the classical lightning network the routing could only take place if A could literally send over the requested money to node B within a payment channel between A and B. For brevity I omitted the exchange of the preimages and negotiating HTLCs and new commitment transactions.  Those details are defined in the protocol specification so that A and B don’t loose money while they route a payment from N1 to N2.

However if A and B are owned by the same entity (as assumed above) there would be no harm if A received some money and B forwards it without A forwarding the money to B. In particular a payment channel with Bitcoins being locked between A and B would be superfluous. This observation is exactly the idea behind a virtual payment channel.

A high level protocol for operating a virtual payment channel could look like this:

  1. Two nodes would agree on an amount both sides provide to the virtual channel. The money does not necessarily have to be backed somewhere (though having the money available is strongly advisable) and the initial balance also does not have to be symmetric.
  2. The channel balance is maintained by both sides as a balance sheet that needs to be cryptographically signed by both sides for every new update. Even though it will not be possible to enforce the final balance via the blockchain those balance sheets might be enforceable by real courts. (Which technically should never happen as VPCs should only be created if there is complete mutual trust between the channel partners)
  3. The contract between both sides includes the fact that the balance sheet with the highest number of mutually signed updates is the correct one and considered to be used for settlement. (I will repeat myself: Such a contract could only be enforced by a real court and not by the blockchain.)
  4. Channel creation is the first update on the balance sheet  and needs to be signed by both parties. In that way a party can decline to open a virtual payment channel by refusing to sign the balance sheet.
  5. The channel is then announced to the network as virtual payment channel via an extension of the gossip protocol. The extension is necessary because currently the blockchain funding transaction needs to be referenced as a proof of existence of the channel. The transaction is encoded to the short_channel_id field of the announcement_signature message as well as the channel_announcement message. From the perspective of the network it is not any different than the other channels. It can be used for routing payments since every hop is a bilateral contract. The rest of the network doesn’t have to care if that contract is actually enforceable by the blockchain or if there is some other trust provider between the channel partners. I literally see no risk in routing money through a third party virtual payment channel for the sender and recipient of the funds.
  6. The channel announcement should be done by both sides and should at least be signed with their own bitcoin keys. This will prevent spamming channels which can’t be used for routing since the partners do not agree that the channels exist. (c.f. point 4)
  7. Routing along a VPC takes place instead of an HTLC. Channel Partners will just update the channel balance. (in fact they could even resemble the HTLC behavior to their balance sheet but I don’t see the benefit)
  8. When closing the channel one party has to somehow reimburse the other side depending on the end balance in comparison to the start balance. This problem of settlement is the problem of the two partners and not part of the protocol. From a protocol perspective any channel partner can just announce the closing of the channel which takes effect immediately. This is precisely the moment where trust and IOU come into place.
  9. Closing of the payment channel also has to be announced via the gossip protocol.

Attention! If for some reason the other party turns out not to be trustworthy one could loose as many funds as the other side virtually provided. In the best case the two nodes are owned by the same person or company or there is some written agreement which backs the virtual payment channel. However obviously one cannot enforce such circumstances and prohibit arbitrary people to create a virtual payment channel. That is why virtual payment channels have to be used with extreme caution. As soon as you accept an inbound virtual payment channel from a third party you vouch with your own funds for the payments of your channel partner. Besides that there are a couple of other drawbacks I see with this proposal.

The 3 biggest drawbacks that I see are:

  1. A virtual payment channel can be seen as an IOU resembling to some degree the mechanics of a demand deposit. One side can provide funds that it doesn’t possess to a VPC and have the other side spend those funds. One cannot tackle this issue by implementing a rule that the network would not accept virtual payment channels if the sum of the amounts in the virtual payment channel is bigger than the sum of the capacities of the standard payment channels. The reason is that the network cannot prove after a virtual payment channel is being closed and reopened that settlement between both partners took place. Hence the potential channel capacity is really infinite. This does not mean that an infinite number of bitcoins can be created out of thin air.
  2. Infinite channel capacity is not a direct problem as in the worst case when routing takes place from A to B the standard channels of B might eventually be dried out in the sense that no further routing away from B can take place. However a network or circle of virtual payment channels could lead to weird chain reactions in which I pay someone and the money is only being routed through virtual payment channels creating literally a money flow out of thin air. I did not calculate this through but I guess at some point in time it will lead to imbalances with real channels but in the first place creating more virtual payment channels is not prohibited.
  3. Since the balance sheet is signed by both parties leaking of the key used for signing is like leaking keys of a traditional payment channel. So even in a virtual payment channel the bitcoins are somehow hot. One cannot directly steal bitcoins but one could initiate sending the virtual balance to nodes one controls and the virtual balance on the routes eventually becomes real balance. So again there need to be complete trust among channel partners.

Things to take into consideration:

  • It is always safe sending money over a virtual payment channel.
  • There is no risk to provide virtual balance in a virtual payment channel.
  • The risk is receiving money on a virtual payment channel because the recipient needs to be sure she will get settled later. This holds true in particular if receiving money was linked to a task like routing it further through standard channels or selling some goods. Both of which should be the standard use cases.
  • One could adopt routing fees that reflect those risks depending in which direction one uses the virtual payment channel or who provides it.

I will now describe a realistic setting which would benefit hugely from the chance to create virtual payment channels.

A use case for virtual payment channels: Payouts in mining pools.

Let us assume you are operating a mining pool. You want to do daily payouts for your – let us say 100k  – customers. Obviously your customers already trust you since they commit their hash power to your pool and trust on the fact that you will eventually do a payout.

The payout scenario without virtual payment channels and without the lightning network.

Every day one has to fit 100k transactions to the blockchain. With batching obviously one could use less transactions but still the amount of transactions and fees will be considerable.

If we have the lightning network the situation is not even getting better: 

For the payouts on day one we have to create payment channels. Obviously a mining pool cannot overfund those channels by too much since it would already need to have the liquidity to do so. So on day 2 the channels have to be either spliced in or the channel has to be closed and reopened. Unless of course the money was already being spent by the customer over the lightning network. (This however will only be possible if additional funds of the mining pool will be used to fund other external channels giving customers the possibility to spend their funds in a meaningful way). We see that using lightning in that case actually seems to increase the amount of blockchain transactions and increases the amount of liquidity that is being used.

Obviously the answer to the sketched problems could be that the mining pool does not have to operate channels to all its customers but can just rely on the lightning network to work in order to find routes to the customers while doing payouts. I predict that in many cases it won’t be possible for the pool to find a route to the customer when she creates an invoice and a channel has to be funded anyway.

Situation with virtual payment channels on the lightning network: 

The pool could accept a virtual payment channel with funds as high as the pool owns to the customer on the customers side of the VPC. This does not require a Blockchain transaction and saves fees. For the second payout on the next day the balance can just be increased by the amount the pool owes to the user. Obviously this kind of payout is not real as the customer does not directly control the funds. However the pool provides cryptographic proof and a signature to the customer which again cannot be enforced by the blockchain. The customer at any time can pay via lightning using exactly this virtual payment channel.

In order for this to be useful for the customer his virtual payment channel needs to be connected to the “real” lightning network. This is achieved by the mining pool which has to have several outgoing channels funded to rather central nodes on the network (e.g. shops, exchanges, payment provider,…). Since the pool can assume that not all customers will spend or transfer all their funds immediately the pool only has to provide a fraction of the funds allocated in the virtual payment channels as outgoing capacity to the other real channels. This reduces the amount of liquidity that needs to be on the hot lightning network node wallet. In case a channel dries out there will be funds spliced in or the channel is closed and reopened. This will happen way less frequently than payouts to all customers and thus reduces load on the blockchain. Also the customer can already receive bitcoins over the lightning network in this virtual payment channel (if he decides to trust even more funds to the virtual channel) In any case the customer can obviously also accept or demand payments on  a standard payment channel by disallowing inbound routing over the virtual payment channel.

Finally the customer could even get hold of his funds by fulfilling inbound routing requests on his standard payment channels which do outbound routing over the virtual payment channel.

For customers of the pool this mechanism probably also has other great advantages:

  • They could configure their software that spending and routing will first go through the VPC so that their money is quickly being spent.
  • They will probably be connected much better to many shops since the mining pool is maintaining and funding those real payment channels.
  • One could even think of a feature in which the pool could “splice out” money of a virtual payment channel by funding a standard payment channel for the user in case the pool can’t find a route to the recipient of the customer. The pool could obviously just itself open that channel in order to have a better connectivity to the lightning network in the future increasing the service for all its customers.

Conclusion:

We see that Virtual Payment Channels increase the flexibility for groups of users that for some reason trust each other. We can think of a virtual payment channel as an agreement to spend the funds of the channel partner in ones own name. Applications are for services that have plenty of customers and don’t want to reallocate funds all the time as well as for very central power nodes that cannot handle all of their traffic on one computer and need to distribute the node to several computers. However virtual payment channels should be used with large caution as one needs to trust another party. In the case of just distributing ones own node this will never be an issue. In case of the mining pool scenario trust already needed to exist in the first place so it should be ok but still yields a risk for the customer. The biggest downside is that a virtual payment channel can be seen as creating liquidity out of thin air and is kind of a credit that might break forcing chain reactions. However in this case it is just the channel partner who should have known better than trusting blindly. In particular if large operators go bankrupt as a follow up of such a chain reaction everyone who has standard payment channels with those operators won’t loose his or her funds due to the trustless property of the lightning network.

To the best of my knowledge the concept of virtual payment channels for the lightning network has not been proposed before. If I am mistaken I will be happy to receive a pointer. Now I am very excited to see what you as a community will think about my proposal. In particular I am excited since I understand that my proposal violets the core philosophy of the lightning network meaning that payment channels can operate in a trustless manner. Yet I think progress will only happen if creativity is allowed by thinking about all possibilities instead of fixed limitations. Also it will not change the trustless nature of the lightning network as this will only occur if two consenting parties agree to engage to such an endeavor. As outlined in this article they might have good reasons to do so. In the great tradition of Bitcoin Improvement Proposals (BIPs) I suggest that we start a process for Lightning Network Improvement Proposals (aka LIPs). This draft would need to be specified a little bit more if people think it would be a useful feature and could be such a LIP at the end.

Obviously if my proposal won’t make it to future BOLT specifications the idea is out and we cannot prevent companies like a mining pool operator still implementing this within their proprietary lightning supported wallet which they provide to their customers. However I think it would be much more preferable if virtual payment channels either become part of the protocol standard or are never used.

If you liked this article and are interested in the follow ups you can follow me on twitter.

 

 

]]>