Emin Gün Sirer, Cornell Professor with years of experience in distributed systems, highlighted a number of problems with the Lightning Network (LN) in one slide presented at the Satoshi’s Vision Conference this past 23rd of March.
“The first and biggest problem is that it has limited capacity. The capacity of that network is unknown, despite so many years of noise around it, nobody has actually gone and done a scientific study,” Sirer said while having the following slide on the background:
“That emergent capacity is going to depend on the credit relationship between people and we don’t know what that’s going to be like,” he said before adding:
“That credit graph is going to look nothing like the Facebook social graph. The fact that I know and love Taariq does not mean I am going to extend to him, you know, $100,000 worth of credit. It’s just, sorry, but it’s not gone happen. So that network may or may not have any carrying capacity.
It’s economically broken… there is no benefit unless the end-points have frequent interactions. So Taariq and I end up interacting once a year, that’s not enough of a reason to open up a channel between us even if he were credit worthy for $100,000.
And worse, an exchange needs to have funds on hand tied into channels that are proportional to the float per person, time the number of people they want to support.
So, suppose somebody could receive up to $10,000 from Coinbase, and Coinbase has, I don’t know, ten million users. So ten million times $10,000, erm $10 billion… $100 billion dollars get’s tied up? Is that right? I don’t know.”
He did get there and fairly quickly. In this scenario $100 billion would be tied up into channels. Sirer continues:
“In that future world where they wanna do that, LN really takes off, a lot of coins are going to be tied into these payment channels.
The system is insecure, requires new intermediaries. You know about the malleability fix, the kitchen sink known as segwit, that if I were to propose it I’m sure it would have not been accepted, yet somehow it got into Bitcoin Core.
It requires keys to be kept online. You need this new thing called Watchguards to watch the chain 24/7 or else you lose funds. You blink, you lose. I close the channel with an old state and you lose your money.”
“Counter-party misbehavior can lock up funds. Some bad thing happens, and now we have to wait for time-outs.
It erodes privacy. Routing involves an inherent trade-off between finding out what paths are available in the network… an inherent trade-off between efficiency and privacy.
To find a good money route to you, I need to know about everybody’s credit relationships, and I need to do this in real time as it changes. And if I were to do that, then I’m actually monitoring you.
I find out that gentleman over there has a credit relationship with the person behind him, well I do it over and over again, and suddenly if the capacity changes, I know someone paid someone else over that link.
That is what gives me the ability to do what we call network topography. And that is not at all a good outcome.”
Sirer initially emphasized that LN is not the only way to do second layers, taking on his talk towards providing an overview of Teechain, an LN like hardware based second layer.
Sirer said he can’t fix all of the highlighted problems as some of them are inherent to second layers, but some of them can be fixed.
“One way is to look at it as a problem, another way is to look at it as a problem to solve,” Patrick McCorry, who describes himself as UK’s first PhD crypto graduate, said before adding:
“The slide is dead-on that there is a tradeoff between privacy/efficiency. “Concurrency and Privacy with Payment-Channel Networks” by pedro has taken a stab at the problem. (and lightning’s sphinx approach). Lots more to do.”
“There’s a privacy/effeciency trade-off, but I think the tradeoff isn’t balanced. If fees are anywhere near where folks anticipate them, adding a few extra hops costs very little. Adding extra channels is more complicated, but also doable,” Matt Corrallo, who is working for Chaincode Labs, a company implementing a version of LN, said.