Quote from: TPTB_need_war on September 09, 2015, 08:14:16 PM

Quote from: smoothie on September 09, 2015, 07:43:12 PM

Delegated is not the same as decentralized.

Nor is it necessarily an antithetical concept. Delegated can be decentralized and effectively as trustless as Satoshi's design is (with some differing assumptions that have differing tradeoffs).

Quote from: monsterer on September 09, 2015, 06:48:59 PM

I haven't noticed the word trustless anywhere - is this design trustless?

Yes and in some respects more so than Satoshi's design, but there is a change in the security model assumption.

You've actually hit upon the key distinction between my design and Satoshi's design. Almost everything else follows from it. But still even if I share with you this, there are many details that you'd still have to invent to get a holistically sound design.

I will excerpt from the rough draft of the white paper and bold the relevant phrases...

Quote

1 Decentralized transaction consensus

A Byzantine fault is defined as a disagreement by system participants on the state of the system. A centralized system is inherently Byzantine fault tolerant because it doesnt disagree with itself. Such faults are only system failure if decentralized agreement on system state is required.

For example, the Byzantine fault of differing opinions of voters is not a failure if democracy by majority rule is acceptable. Systemic failure results from obstruction of voting, undetected ballot box stuffing, or purchased votes. These are respectively denial-of-service (DoS), Sybil, and resource-capture attacks.

Protection against the double spending of an account balance is an example of a contract that depends only on the relative order of spend events, and not on the events timestamp. The latter spend must be discarded otherwise the first payee would be defrauded.

In a decentralized network of observers of spend events, a majority vote on event order would fail due to DoS, Sybil, and resource-capture attacks on the observers.

DoS attack on relayed events over the network.
Sybil attack creating unlimited observers or network relay peers.
Resource-capture of any resource required to be an observer.

These attacks can be squelched if for observers and peers their system function is verifiable truth, they can prove a trusted reputation, or they can expend or risk sufficient resources which exceed the gain from cheating.

2 Non-delegated transaction verification by longest chain of proof-of-work

In Satoshi Nakamotos longest chain of proof-of-work decentralized consensus system[Nak09], the participant nodes are distrusted and not capable of independently verifying the event order. They risk proof-of-work resources in exchange for block rewards with low probability of gain from cheating, unless the adversary possesses greater than 50% of the total network proof-of-work resources, or 25 - 33% in the case of selfish mining[ES13].

In addition to this Byzantine failure due to capture of at least 25 - 50% of the system resources, Satoshi Nakamotos system has other weaknesses.

Reliability of the event order requires a proof-of-work confirmation to prevent a double spend due to the Finney attack[Fin11] or gaming of differing mempool heuristics[Hea15]; yet remains unreliable if there is network fragmentation or with too few confirmations given a significant orphan rate combined with some network propagation opacity.
Variance of mining[Ros11] incentivizes a limited number of pools of nodes thus concentrating the control of system resources.
The scaling tradeoff that all nodes incur the bandwidth and processing load to relay and verify all system-wide transactions, otherwise they delegate to pools which concentrate the control of system resources.
Satoshis protocol doesnt verify if pools implement getblocktemplate to enable a node to insert transactions into its winning block to mitigate the pools power to censor transactions.
Satoshis design depends on concentrated control and monolithic network coherencewhich sacrifices censorship resistance and network fault tolerancein order to provide scalability and reliable instant transactions.

Concentrated control accruing naturally or via a resource-capture attack could alter the protocol such as increasing the money supply of cryptocurrency (even gifting the debasement to any publicly acceptable entity such as government¹) or only validating events which accompany some KYC (know your customer) proof of identification. Some argue that a political outcry would move away from such an attack on the protocol, but in reality the preoccupied masses tend to continue to the use the clients they are told to use and are accustomed to such as popular web clients Coinbase, Blockinfo, etcetera. The masses didnt abandon the dollar in spite of its malevolent holistic effects. Most people would not agree the dollar is malevolent.

The security model of the longest chain of proof-of-work in Satoshis system is ultimately founded on the principle that each participant minion will act unselfishly in the short-term to protect the long-term collective benefits of trustless consensus yet this has proven not to be the case throughout all recorded human history when the individual selfish incentives are great enough.

¹ Some influential people such as Martin Armstrong have simultaneously outlined an expectation of a one-world currency monetary reset solution to the current prolifergate nation-states sovereign debt crisis; and called for (a world?) government to tax the money supply instead of income taxes[Arm08].

3 Delegated transaction verification by longest chain of proof-of-work

To overcome the weaknesses in Satoshis design, we delegate the verification of event order to _________ nodes which provably record their system functions in the longest chain of proof-of-work.

Delegating to these _________ nodes is harmless because these nodes obey the fundamental end-to-end principle of networks in the sense that they make no discretionary choices of significance, persist the minimum state possible, are fungible with each other, and cant be monopolized. _________ nodes do set the transaction fee they charge which is based on transaction data size and not value since transaction values may be concealed.

...

Quote

3.10 Security model

The security model assumption of Satoshis design is that every minerwhether alone or in collusion with other minershas a greater opportunity cost when mining on an incorrect or shorter block chain. Additionally unlike other reputation-based or proof-of-stake schemes, it is implausible to game the order of block solutions because the source of the entropy for proof-of-work is external thus an open instead of closed thermodynamic system. Also each miner can verify the entire history of the block chain, including every transaction.

The security model weaknesses of Satoshis design detailed in the prior sections derive from the generative essence that miners cant autonomously determine the correctness of the longest chain w.r.t. to double spends, censorship resistance, mining centralization, collusion, and selfish mining.

Our security model retains the assumption of greater opportunity cost when mining on incorrect or shorter correct block chain, while adding objectivity to the correctness of double spends and censorship resistance. The model is founded on the principle that published data cant be unpublished and that nodes listening for compliance will need to do so for other reasons, such as _________ nodes maximizing the efficiency which payers can send non-instant transactions...

Although block chain miners dont verify for themselves every transaction in the chain of hashes for each hash stored on the block chain, the nodes listening and checking for compliance insure that the block chain is verified against the distributed and independent _________ nodes.

Any entity that downloads an archive of all the historic distributed data could perform a full verification to attain the same security as a full node that downloads the entire block chain in Satoshis design. The distinction from the security model in Satoshis design is that if the compliance checking nodes fail to report cheating, the cheating will become nearly implausible to revert. This is why the incentives offered to compliance checking nodes are lucrative and compliance checking node membership is permissionless. Compare the odds that no party will avail of the profit incentive to the security assumptions in Satoshis design:

Network fragmentation never occurs.
The masses will politically take time from their preoccupied lives en masse to fork away from an insidiously (perhaps undetectably so) malevolent 25 - 51% attack, e.g. one that requires KYC or adds some debasement to fund social welfare in a world government or collaboration of regional governments such as EU, G7, G20, Asian Union, and Mercosur.
Pools with different names, IP addresses, and servers arent controlled by the same entity, i.e. Sybil attacked.
Scaling transaction volume and zero confirmations by centralization is secure.

Choose your poison. I think the security assumptions in my design are much more robust because decentralized, permission-less opportunity is inherent more reliable than centralized control. I would guesstimate the odds of a serious problem arising in my security model on the order of an asteroid striking your house, while the odds of the Bitcoin algorithm failing (to remain decentralized, permission-less, not just a fiat) due to any of those listed items at the end is very palpable.