David Schwartz Stanford Blockchain 2020

[mandelbaum]

David Schwartz at the Stanford Blockchain conference february 20th 2020 held a lecture called: The best incentive is no incentive.

Warning: David issued a trigger warning to Bitcoin maximalists in the introduction.

 

Edited February 22, 2020 by mandelbaum

[Guest]

Three things stood out for me...   David was rushed through his talk because of time constraints,  so the meat of it was blurred a bit by the speed.  Also he didn’t really fully make the case for ‘no incentive’ (as a result of that shortness of time).

The third thing is that he said on multiple occasions that without a node incentive there is therefor no incentive for bad actors to attack the network.  I think that is actually wrong on multiple counts...  which is odd because I’m certain that he is hundreds of times smarter and more knowledgeable than me...  so my conclusion can only be that either I am wrong (I don’t think I am though) or this presentation was a bit rushed in preparation as well as presentation.

In any case he cuts a startling figure as he passionately speaks on something that fully engages him.

[jn_r]

I kinda agree that the correctness / validity of the network is sound and that there is no incentive from within the network to be gained by attacking the network.

But considering the liveness of the network: An attacker can sabotage the network if he wants to. Reasons why to attack can be as simple as 'hate' (still some of that available in toxic blockchain universe), political (not in my country) or incentives from outside the network. DDOS-ing or sabotaging the network (disable >20% of the validators from the dUNL) could devalue the price of XRP, an event that can be made profitable in other ways outside the XRPLedger network.

That said, an argument could be made that this 'liveness' issue can also be solved in others ways than an incentive model like POW or POS. If you are ok with

- Ripple singlehandedly changing the d(ynamic)UNL such that >80% of the validators are available again

or 

- Allowing the network to be down for a longer period until all important nodes adjust themselves manually to another agreed upon UNL,

then we're fine and we do not need incentives. But I think there is still room and need for improvement, so imo, argument made, but not conclusive.

 

 

Edited February 22, 2020 by jn_r

[Guest]

1 hour ago, jn_r said:

But considering the liveness of the network: An attacker can sabotage the network if he wants to.

Let me start by saying that I think you know far more about this than I do.

That said...   I’m unsure what you mean.

I don’t understand how the attackers sabotage the network?  Spam it?  That’s handled by fee escalation.  
DDOS?  That’s handled by decentralisation.  

Badly formed trans or continuous disagreement?  Consensus ignores the bad actor and the good nodes stop listening.

So you mention this liveliness and I’m assuming you mean dragging the network consensus down in some way...  what way is that? 

[jn_r]

Ok, I'm not a professional consensus designer, just having a interest in the topic, so I hope my wording is ok. But what I mean is:

2 minutes ago, Tinyaccount said:

I don’t understand how the attackers sabotage the network?  Spam it?  That’s handled by fee escalation.  
DDOS?  That’s handled by decentralisation.  

What I understood is that not that long ago almost 6 validators were brought to their knees because of some spam attack and some other coincidental happenings (which I don't know what they were). So sabotaging could possibly be done by either spamming the network with messages, or otherwise you could try to a classic ddos on the IP-ports of the validators in question or target their locations (granted you must know where they are located or on which IP, but that seems not entirely impossible, knowing who the owners of the validators are). 

10 minutes ago, Tinyaccount said:

Badly formed trans or continuous disagreement?  Consensus ignores the bad actor and the good nodes stop listening.

I don't think badly formed transactions will bring the network down, but if not enough validators are available to reach a quorum of 80% then the network will halt, because no new ledgers will be created.

14 minutes ago, Tinyaccount said:

So you mention this liveliness and I’m assuming you mean dragging the network consensus down in some way...  what way is that? 

If the network doesn't continue or is halted, then it is no longer 'alive'.It could be revived though after some time, e.g. by changing the UNL

 

If you compare with the current POW from Bitcoin or Ethereum, you could argue that for POW it doesn't matter if you bring a lot of mining-pools down, the liveness of the network is guaranteed because every node can become a miner, it doesn't need intervention from a central party. Network might slow down, though, because the needed hash-power is less available. 

In that regard it is comparable with the liveness of XRPL, both XRPL-consensus and POW are in danger of temporary halting the network by bringing elements in the network down, which might take some time to fix. But the difference is that in POW there is no need for additional agreement, new miners just start mining. in XRPL-consensus we need agreement on a new or adjusted UNL

[karstnDE]

First thanks for sharing, I found this to be a brilliant talk, unfortunately with too little time. Regarding @Tinyaccount's concerns I agree with @jn_r - nicely summarized! 

Fees and decentralisation don't necessarily prevent the liveliness of the XRPL decreasing. We had the case with the BTC  IOUs for a few weeks, then the number of tx per ledger increased to ~400. That increased the operating cost for validators, and if a validator is not operated from a well funded institution but e.g. a private person that might lead to less validators ensuring robustness of the network. Same for DDoS.

That's why you need the natural stakehodlers of the XRPL (e.g. FlashFX, moneygram) contributing to it's robustness and not free-riding. I would expect though that the more well-funded institutions, i.e. large businesses, rely more and more on the XRPL, the higher the risk will be for the XRPL halting, the more incentivized they will to stop free-riding and actually contributing. 

 

Edited February 22, 2020 by karstnDE

[Guest]

6 hours ago, karstnDE said:

I would expect though that the more well-funded institutions, i.e. large businesses, rely more and more on the XRPL, the higher the risk will be for the XRPL halting, the more incentivized they will to stop free-riding and actually contributing. 

Yes this.  This is exactly how I expect it will pan out.  Any enterprise will do what it can to safeguard its expected possible points of failure. 

 

6 hours ago, jn_r said:

If the network doesn't continue or is halted, then it is no longer 'alive'.It could be revived though after some time, e.g. by changing the UNL

Thanks for explaining your concern.  So if I understand correctly,  the main attack that concerns you is against the UNL validators.  If they drop out because of say every member attacked with simultaneous DDOS,  then the ledger halts.

I’m not sure if that is true or not.  I think that if validators had a couple of non-UNL nodes in their list then it wouldn’t stop forward progress.  However I get the feeling that most validators do not have anything other than the default.  So maybe at the moment it is true.  So possibly that is something to address.

However...  it can be addressed without resorting to monetised incentives.   If I understand it correctly,  (and there is a significant chance that I don’t, )  adding a few randomised nodes into everyone’s list will mean that the ledger progress continues even though the UNL nodes are not contributing.   I would love to see a few expert opinions on that.  
 

Irrespective though of whatever technical mitigation might be done...  I’m certain that David is correct and the best incentive is no incentive.  Which pithy statement hides inside itself the fact that there IS an incentive...  the shared interest of all participants (especially the overtly commercial ones, )  to have a robust XRPL.

[jn_r]

2 hours ago, Tinyaccount said:

adding a few randomised nodes into everyone’s list will mean that the ledger progress continues even though the UNL nodes are not contributing

It's an interesting thought, but a list of randomised nodes might not be a good idea. It has been found that in order to be save with your UNL, an overlap of 90% in UNL validators is necessary. But who knows, maybe the current validators have secretly made an agreement together to include a fixed list of secret validators next to the well known UNL  That would work, but it would be security by obscurity, not the best .. 

I was thinking of keeping up a backup list of validators, ordered somehow by quality. If a validator on the UNL would fail for some rounds his place could be filled with the first on the backup list. But I haven't worked it out further, not sure if it would work. 

The proposed solution described in the link below is the opposite, it creates a negative list on the ledger:

 

[Mpolnet]

On 2/21/2020 at 6:59 PM, mandelbaum said:

David Schwartz at the Stanford Blockchain conference february 20th 2020 held a lecture called: The best incentive is no incentive.

Warning: David issued a trigger warning to Bitcoin maximalists in the introduction.

 

Really solid video in my opinion. Definitely a bit rushed and I love David for his intellect and passion. Not sure how technical the audience was (I hope they were predominately computer scientist) because man was this convo high level on a lot of technical aspects. I do agree with his view that the initial promise of blockchain was to eliminate rent seekers so if that's the promise of this new technology it's hard to argue against no incentives and how incentives introduce friction. However, arguing for incentives is also understandable so it's definitely a worthwhile topic of discussion. Definitely recommend watching!

[karstnDE]

I really liked DS talk, so I wrote a small piece about the general topic. In case you're interested: https://coil.com/p/karstnDE/Blockchain-Ecosystem-Design-The-Case-for-the-XRP-Ledger/yRJZlHY0

[JASCoder]

29 minutes ago, karstnDE said:

I really liked DS talk, so I wrote a small piece about the general topic. In case you're interested: https://coil.com/p/karstnDE/Blockchain-Ecosystem-Design-The-Case-for-the-XRP-Ledger/yRJZlHY0

Nice write up friend, thanks for sharing.

You surly saw BG's related tweet yesterday, including it here for anyone that somehow missed it...

https://twitter.com/bgarlinghouse/status/1234997098568798208