BobaPandora (OP)
Newbie
 Offline
Activity: 4
Merit: 0
|
 |
April 20, 2015, 05:17:14 PM
Last edit: April 20, 2015, 06:23:47 PM by BobaPandora |
|
In view of some recent discussions, I thought I would write a post here to present my opinion on non-PoW consensus protocols. For the past couple of years there have been dozens of novel and interesting non-PoW approaches to achieving consensus, most based on some combination of trust and stake. For example, we can arrange crypto-coins on a spectrum ranging from stake-based to trust-based as follows:
[STAKE] Ethereum v2 (PoS), Peercoin, etc —-> Nxt (w/ leased forging) —-> Bitshares DPoS —-> Stellar/Ripple [TRUST]
(For purposes of this post, I will ignore ignore the PoW components of Ethereum and Peercoin)
My assumptions:
1) Trust should be dependent upon stake. Breaking or diluting the relationship between stake and trust leads to a perversion of incentives.
2) Somewhat obvious, but a public blockchain (of some form) is necessary. By this I mean that the entire ledger history must be publicly available for verification.
3) As a result of #2, the ledger will grow rapidly and, therefore, there will be *some* non-trivial computational resource requirement to participate in the verification of transactions.
So here are the conclusions I draw:
* #1 above would indicate that Stellar, Ripple, and similar systems are non-ideal (not the mention the complexity-security inverse correlation)
* Non-delegated PoS systems suffer from low participation. For example in Ethereum, without delegation, the punitive PoS system actually disincentivizes participation due to the fact that a well-intentioned PoS “miner” could lose their deposit balance if they are hacked. This means that even large non-sophisticated stakeholders may choose to not participate, and the network may suffer resultantly. This problem would increase in the case that mainstream adoption takes place, as the higher opportunity cost and demographic change would further reduce participation. Punitive PoS is an attempt at solving the nothing-at-stake issue, but it exacerbates the participation problem. If the assumptions hold true, we can conclude that Peercoin and Ethereum are non-ideal. (edit: To clarify, punitive PoS solves a real problem. Just pointing out the side-effect here and pointing out that a constrained delegated PoS would help to mitigate this.)
* Nxt uses a form of delegation called “leased forging” which is very similar to DPoS. The primary difference here is in efficiency and, secondarily, “weight” of participating forgers (which I will not address here). DPoS allows only the top 101 delegates to forge blocks (with equal frequency), while Nxt block forging is performed primarily by a handful of large stakeholders, with a long tail of less frequent forgers.
I hypothesize that Bitshares DPoS, if it were to gain mainstream adoption, would result in the most transparent and efficient system of those discussed. Firstly, the set of active delegates would eventually stabilize in the hands of 101 very public and transparent entities (maybe some governmental) with razor thin (or zero) margins. Most likely these entities would pay a dividend to voting accounts and would have a vested interest in the success of the platform. Any malicious or unreliable delegates would promptly be replaced via rapid, low-friction, trust-based voting.
In the same way that crypto-currency exchanges (which are themselves a sort of “delegate” in the world of Bitcoin) are trending more recently towards public audits and transparency (e.g., known and trusted identities, cryptographic proof of reserves, etc.), DPoS delegates will similarly trend towards openness and accountability, as demanded by voters/stakeholders. In both Nxt and Bitshares, stakeholders have a low barrier to participation (moreso in Bitshares). If the crypto universe were an economy, delegated block production can be thought of as a form of “specialization” that is a natural byproduct of a free market.
Also, unlike Nxt, the set of Bitshares block producers (active delegates) is relatively stable/static under normal circumstances. Nxt is limited by the lowest common denominator of block producers. Operating under the principle that the average user should be able to forge blocks, the Nxt protocol cannot easily be modified to embrace a scalable infrastructure. Bitshares delegates, on the other hand, will most likely build dedicated infrastructure to establish direct communication channels between active delegates and to enable transaction speeds that are simply not possible with (a) a non-deterministic solution or (b) a solution dependent on low computational resource requirements. Already, Bitshares has the fastest block production of any PoS coin (~10 seconds).
Keep in mind that there are other differences between the protocols I’ve discussed in this post that will obviously impact their security or their usefulness in any one application. That is not my point. My point is that, all else being equal, constrained delegated block production is better suited for participation and efficiency over other available options.
* Very obviously, I am a Bitshares stakeholder
|
