Author [EN] [ZH] [ES] [PT] [IT] [DE] [FR] [NL] [TR] [SR] [AR] [RU] [EN] [ZH] [ES] [PT] [IT] [DE] [FR] [NL] [TR] [SR] [AR] [RU] [EN] [ZH] [ES] [PT] [IT] [DE] [FR] [NL] [TR] [SR] [AR] [RU] Topic: drltc's Trustee Technical Discussion Thread  (Read 1686 times)

0 Members and 1 Guest are viewing this topic.

Offline theoretical

drltc's Trustee Technical Discussion Thread
« on: March 29, 2014, 11:28:02 PM »


I dislike the trustee concept.  But the more I thought about it, the more I realized that we really don't know any technical details.  So far, the only information I've seen is that the new system will involve a single centralized entity. 
Countless words on this forum have already been devoted to the various strengths and weaknesses of the trustee from legal, regulatory, and marketing perspectives.  Those are surely worthy issues, but if you'd like to pursue them, please do so in a different thread.  I'd like this thread to be solely about technical specification and analysis of cryptocurrency designs that incorporate the trustee concept (with perhaps a little economic analysis as well).

This is a very recent direction change, and I'm not sure that bytemaster or anybody else has publicly discussed detailed technical mechanisms that might be used in a trustee-based BitShares X.  The goal of this thread is to begin that discussion.  Legal, regulatory, marketing, and economic perspectives are off-topic (but may be tolerated if they inform technical decisions).  The primary focus is discussing the practical engineering concerns of trustee-based cryptocurrencies, with a focus on how BitShares X should work.

If you like my analysis in this thread, and want to know my thoughts on how a non-trustee-based BitShares X should look, you can read all about that at https://github.com/drltc/xts-proposal -- and you can also donate PTS to the address there.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #1 on: March 29, 2014, 11:28:39 PM »

To start off with, I'll examine the technical reasons for a trustee design.  My understanding is that the trustee design is bytemaster's attempt to solve two problems with a random method (Poisson process):  (1) Sometimes blocks are generated too far apart.  This is annoying; users would have a better experience if the wall-clock time between blocks was predictable.  (2) Sometimes blocks are generated too close together.  If the interval between two blocks is smaller than the network propogation delay, this is the beginning of a fork.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #2 on: March 29, 2014, 11:28:56 PM »

In this post, I'll discuss two broad design flavors for trustee elections:  "political" elections and "technical" elections.  In "political election," the trustee is chosen by what is basically a traditional democratic process (well, more like a corporate election process as the weight of your vote is determined by the number of coins you hold).  A trustee is a human with a known identity (or at least a pseudonym), and convinces shares to vote for him (or her) by convincing humans to manually enter the candidate's public key in the "I vote for this person" field.  A political election scheme would probably have any trustee in office for months or years, barring impeachment.  Political elections suffer from many problems.  I believe bytemaster's intent is to have political elections.  By contrast, in a "technical election," the trustee is chosen by some automatic process.  I propose the trustee should be chosen randomly from all online nodes (weighted by number of coins), and serve for a limited term (10-1000 blocks, a term measured in hours or days).  If a trustee loses network connection, a replacement should be elected after some timeout interval.  A traditional proof-of-stake random selection method could be considered an example of a technical election with a one-block term.

I believe technical election is superior to political election in the context of cryptocurrency trustee designs.  Much of the analysis to date -- including many of the objections to the trustee concept that have been raised -- apply only to designs involving political elections.  Political elections require some amount of conscious, manual actions by humans representing a large share of the network, but technical elections can happen entirely without human intervention.  Thus, systems using technical elections can quickly replace an offline, malfunctioning, or malicious trustee, without waiting for the humans to notice the situation, arrive at a decision that's acceptable to a majority, and convince that majority to actually vote.  (Anyone who pays any attention at all to real-world politics should realize that each of those steps may prove to be a major obstacle in practice.)  Further, since the choice of candidates and the election process are fully automated, it is practical to hold technical elections very frequently.  Therefore the term length can be set fairly short, which greatly reduces the influence that any one trustee can have on the system.  Finally, technical election can set unambiguous rules for resolving ties, or results where no candidate receives a majority.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #3 on: March 29, 2014, 11:30:43 PM »

In this post, I'll discuss two potential attack vectors for trustee-based systems.  A trustee may pretend that they haven't received certain transactions, either because they wish to censor those transactions, or to gain some advantage.  I call any attack based on ignoring network communications a "selective deafness attack."   A trustee may also use their control of, and information about, the unpublished exact details of the next block in order to gain some advantage.  The simplest example of this is that a trustee can avoid a mismatch between their order and what the market is willing to pay; they know exactly what the best bid / best ask will be, and can construct their transaction to match.  I call this "trustee front-running."

As I've pointed out in other threads, existing networks like Bitcoin that randomly select "trustees" who serve a one-block "term" have protection against selective deafness attacks, simply due to the fact that the trustee will quickly be replaced, and it's difficult to get all Bitcoin miners to agree that a certain transaction is unacceptable (unless of course the transaction attempts to spend money that doesn't exist, has an invalid signature, or is otherwise corrupt / violates the spec in some way).  So there's a high probability that, within a few blocks, at least one miner will accept any given transaction.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #4 on: March 29, 2014, 11:31:00 PM »
In this post, I'll propose what I see as a practical design for a trustee-based system.

I propose a hybrid two-tiered system.  The higher tier is a single trustee that generates blocks using txsets provided by the lower tier.  The lower tier is responsible for generating txsets using a random (Poisson) process.  All online nodes participate in generating txsets using difficulty-adjusted proof-of-stake.  A txset consists of a proof-of-stake, a parent block hash, a list of transactions, and a signature by the key that generated the proof-of-stake.  When a txset is generated, it's broadcast (of course, to avoid abuse of the broadcast communication path, all nodes must validate the stake, signature, and transactions before propagating the txset).

The trustee must include exactly TXSET_QUORUM txsets in a block; otherwise the block is invalid.  Further, the transactions included in the block must be precisely equal to the set of transactions that occur at least TXSET_PLURALITY times in the block.  TXSET_QUORUM is a tradeoff:  Increasing it increases security, but costs broadcast bandwidth and blockchain storage.  I suggest TXSET_QUORUM = 8.  TXSET_PLURALITY is also a tradeoff.  A trustee with connections to many wealthy coin holders will be able to front-run if they and their allies are able to generate TXSET_PLURALITY different stakes.  Thus, increasing TXSET_PLURALITY makes front-running harder for the trustee.  The tradeoff is that a larger TXSET_PLURALITY delays transactions, since they are required to saturate the network further before they become eligible to be included in a block.  I suggest TXSET_PLURALITY = 5.  (As its name suggests, there is no intrinsic reason why TXSET_PLURALITY >= TXSET_QUORUM / 2, even though my suggested value obeys that inequality).

In order to pull off a selective deafness attack -- especially over multiple blocks -- a trustee would have to control an enormous percentage of the network's online coins.  Indeed, such consensus would be needed that a successful maneuver might be better described as "the trustee's desired protocol change was voluntarily adopted" than "the trustee carried out a successful attack."

Since txsets are generated randomly, it is possible for txsets to be generated too slowly for TXSET_QUORUM blocks to be available when the trustee wishes to generate the next block; under these circumstances, the trustee has no choice but to delay the next block until more txsets become available.  (One possible cause of slow txset generation include bad luck in txset generation.  Another possible cause is a downward txset generation capacity fluctuation that isn't yet reflected in the difficulty.)  To mitigate the effect of slow txset generation, the generation rate of txsets can be increased so that the expected number of stakes generated in the desired block interval is TXSET_OVERGEN * TXSET_QUORUM.  Like the other parameters I've discussed, TXSET_OVERGEN represents a tradeoff:  Increasing TXSET_OVERGEN reduces the probability that blocks are late due to insufficient txsets, but reduces the percentage of the lower layer that a trustee must control in order to effectively dictate which transactions will be included in a block.  I suggest TXSET_OVERGEN = 1.25.

My proposal has some protection against network splits:  If a split occurs, then txset generation capacity will greatly slow on the "wrong" side of the split.  If the split's resolved before the timeout interval, no fork will be created.  If a substantial fraction of the online coins prior to the split are on the trustee's side of the split, then block generation will continue at a reduced rate.  If the split's resolved after the timeout interval, there will be a fork, which will be resolved in favor of the side of the split that generated the longer chain.  Which, due to the timeout interval, will be the pre-split trustee's side of the split if the split is resolved quickly.  A longer split would be resolved in favor of whichever side of the split has more online coins, hence txset generation capacity.

A couple other details:  Each txset is required to be consistent, e.g. cannot contain double spends.  But since the block is a combination of txsets, the block may end up containing double spends; the trustee can't remove them because they can't override the decision of the txsets.  I propose the following solution:  Double-spend transactions resulting from txset combination are allowed in a block, but they turn into no-ops, and may be pruned once they are old enough to no longer be needed to validate the actions of the trustee.

Since transactions are not necessarily commutative and associative, transactions should be sorted according to some fixed deterministic order.  In particular, the trustee should not be able to control the order in which transactions execute.  I propose executing transactions in order of descending fee per byte (thus higher fee-paying transactions go first), with H(cur_txn + prev_txn) used to break ties, where cur_txn is the hash of the current transaction and prev_txn is the hash of the previously executed transaction (or all zeros if the tie is for the first transaction to be executed).

Also, it is possible to optimize the storage of txsets in a block, since we know that each txset's transactions are a subset of the block's transactions.  For each transaction in the block, create txset_inclusion_mask, a value of TXSET_QUORUM bits indicating which txsets included this transaction.  (With the suggested parameter value TXSET_QUORUM = 8, txset_inclusion_mask will be a single byte.)  Then the txsets and their Merkle hash can then be reconstructed from the block's transaction list and the txset_inclusion_mask values.  The Merkle hash is dependent on the ordering of the txset's transaction list, but we can simply require txsets to list transactions in sorted order.

With the above optimization, the overall storage requirements of my proposal are quite modest:  A constant per-block overhead of 8 proofs-of-stake and 8 signatures, along with a one-byte increase in transaction size.
« Last Edit: March 29, 2014, 11:36:11 PM by drltc »
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline Agent86

  • Sr. Member
  • ****
  • Posts: 471
  • BTSX: agent86
    • View Profile
Re: drltc's Trustee Technical Discussion Thread
« Reply #5 on: March 30, 2014, 12:33:33 AM »
Hi drltc,
I'm not sure you are accurate on the reasons for the trustee.

You wrote:
"My understanding is that the trustee design is bytemaster's attempt to solve two problems with a random method...  (1) Sometimes blocks are generated too far apart. (2) Sometimes blocks are generated too close together."

I don't think this is the reason for it.  I think the problem is that the only "random methods" involve both mining with all the potential centralization and waste that comes with mining, and also require people to leave their wallets open on a networked computer to prove their stake.  Both of these have big drawbacks; very few people want to risk their coins on a networked computer to participate in the mining/security.

You also wrote:
"A traditional proof-of-stake random selection method could be considered an example of a technical election with a one-block term."

I don't think there is such a thing as a problem free "traditional proof-of-stake random selection method"

Offline bytemaster

Re: drltc's Trustee Technical Discussion Thread
« Reply #6 on: March 30, 2014, 02:17:44 AM »

In this post, I'll discuss two broad design flavors for trustee elections:  "political" elections and "technical" elections.  In "political election," the trustee is chosen by what is basically a traditional democratic process (well, more like a corporate election process as the weight of your vote is determined by the number of coins you hold).  A trustee is a human with a known identity (or at least a pseudonym), and convinces shares to vote for him (or her) by convincing humans to manually enter the candidate's public key in the "I vote for this person" field.  A political election scheme would probably have any trustee in office for months or years, barring impeachment.  Political elections suffer from many problems.  I believe bytemaster's intent is to have political elections.  By contrast, in a "technical election," the trustee is chosen by some automatic process.  I propose the trustee should be chosen randomly from all online nodes (weighted by number of coins), and serve for a limited term (10-1000 blocks, a term measured in hours or days).  If a trustee loses network connection, a replacement should be elected after some timeout interval.  A traditional proof-of-stake random selection method could be considered an example of a technical election with a one-block term.

I believe technical election is superior to political election in the context of cryptocurrency trustee designs.  Much of the analysis to date -- including many of the objections to the trustee concept that have been raised -- apply only to designs involving political elections.  Political elections require some amount of conscious, manual actions by humans representing a large share of the network, but technical elections can happen entirely without human intervention.  Thus, systems using technical elections can quickly replace an offline, malfunctioning, or malicious trustee, without waiting for the humans to notice the situation, arrive at a decision that's acceptable to a majority, and convince that majority to actually vote.  (Anyone who pays any attention at all to real-world politics should realize that each of those steps may prove to be a major obstacle in practice.)  Further, since the choice of candidates and the election process are fully automated, it is practical to hold technical elections very frequently.  Therefore the term length can be set fairly short, which greatly reduces the influence that any one trustee can have on the system.  Finally, technical election can set unambiguous rules for resolving ties, or results where no candidate receives a majority.

Thanks for starting this thread... I look forward to seeing the results.   Some quick thoughts:

1) Lets use the notary name rather than trustee because it conveys the responsibility.
2) The technical elections can be manipulated by the false assumption that large stakes indicate trust.
3) Political processes only get 'personal' and thus result in infighting when there is something to be gained by getting in office.   In this case the Notary incurs a cost, but otherwise can gain nothing by having the role and they can lose it if the fail to do a very specific, unambiguous job.   These elections would not be about the 'future direction of the company'.

The problem that I am attempting to solve is not so much just the 'blocks too close' or 'blocks too far' but the fact that an attacker can use short-term uncertainty regarding the chain to attack the network.  The technical challenge is to reach *irreversible* consensus as quickly as possible with out giving anyone opportunity to steal funds.

The challenge with random selection is that you get random failure. 
The challenge with voting pools is that you can get network splits and have synchronization times.  Also choosing who is allowed in the pool is political.

Lets consider the two primary attacks on a trustee system and find technical solutions to them:

1) Notary selectively blocks certain transactions, but continues to operate.
2) Notary stops producing blocks all together.

In the 2nd case the process is very simple.  Every shareholder votes for one or more backup notaries which may only produce a block if the primary notary fails to produce one for a set period of time (1 hour).  After this point the backup notary takes over.  There can be a robust chain of command based upon votes in the past that can automatically take over.

In the 1st case we have two sub cases:
  1) the notary blocks attempts to vote them out of office.
  2) the notary is performing filtering that is 'politically acceptable' and thus they do not get voted out.

I would argue that case #2 is effectively consent of the governed and the only solution is a hard fork with a new trustee for the minority that are against the politically acceptable filtering.

The result is that we only have to deal with case #1.  A notary that is doing partial filtering and is obstructing the election of a new notary.   We can mitigate this particular case by having N notaries that take turns producing blocks.   In the event that any 1 notary fails to produce a block on their turn, then the remaining notaries can vote out that particular notary after X minutes and the next notary in line will take their place.

The key thing with notaries is that they must be prepared to have servers with very high uptime and redundancy. 

The focus should be on how to perform the handoff in a deterministic manner if the primary notary goes rogue.   


 






For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline bytemaster

Re: drltc's Trustee Technical Discussion Thread
« Reply #7 on: March 30, 2014, 02:21:00 AM »
A notary that can do 'front running' would be easily detected by the frequent inclusion of never-before-seen transactions.   If this was observed they would be voted out in short order and the attack reverts to the case of subverting the voting process.

For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #8 on: March 31, 2014, 08:21:01 AM »
2) The technical elections can be manipulated by the false assumption that large stakes indicate trust.

Any system that doesn't give large balances a proportionally large chance of getting elected will simply encourage large balance holders to split up their balance among multiple addresses.  If everyone does this, the result will be massive waste of blockchain storage.  But if only some people split up their balance, the ones who split up will have an advantage relative to those who don't.

Suppose there are 10,000 honest nodes that control 5 addresses on average.  They are all running for notary, and will perform the function faithfully if called to serve.  They haven't created more addresses because they don't care enough about getting elected notary to study the mechanics and engage in abnormal transaction patterns to optimize their chances of getting elected.  Suppose Eve comes along and decides to run for notary with an intent to abuse the position, e.g. she plans to disappear and black out the network until the timeout expires and a replacement is chosen.  Eve runs for notary with 9,950,000 sock puppet addresses.  If each address has an equal chance of getting elected to be a notary, then Eve has a 99.5% chance of getting elected by virtue of controlling 99.5% of the addresses on the network; the fact that each of those addresses only contains a few satoshi is simply irrelevant.  Worse yet, there is a 99.5% chance that Eve's replacement will be another of her sock puppet addresses, which will itself disappear when elected.  Selecting an address Eve doesn't control has one-half of one percent chance of success per hour; the outage will last 200 hours on average.

The above example shows that the chance of getting elected must be proportional to some scarce resource that is intrinsically hard to create (stake).  I chose coins.  Here are some alternatives which I found less suitable:  CDD is scarce, but unfortunately it also encourages sybils, for reasons I've discussed elsewhere.  Hashpower is scarce, but BitShares has already publicly promised to avoid protocols incorporating a computational arms race.  Real-world identities are the scarce resource used to prevent sybils in real-life democratic elections ("stuffing the ballot box"), but obviously real-world identities cannot be verified by an arbitrary node.

Every shareholder votes for one or more backup notaries which may only produce a block if the primary notary fails to produce one for a set period of time (1 hour).  After this point the backup notary takes over.  There can be a robust chain of command based upon votes in the past that can automatically take over.

If a network split has persisted for an hour, you still have a fork.  If the first N backup notaries are offline, your outage lasts N hours.

I suggested the same timeout mechanism, though I didn't specify a suggested value for the timeout.  Let's call the parameter NOTARY_TIMEOUT = 1 hour.  Both your system and mine protect absolutely against forks caused by splits shorter than NOTARY_TIMEOUT.  Both systems generate forks after NOTARY_TIMEOUT has elapsed and the "wrong" side of the split elects a new notary.  My system has the additional advantage that the small side of "lopsided" splits -- where a large number of online coins end up on the same side -- will slow its block production rate proportionally.  Thus, geographically isolated network problems, e.g. a natural disaster that disconnects a single city but leaves most of the Internet intact, will automatically have good failure behavior as long as the isolated area contains a small number of the total online coins -- regardless of whether NOTARY_TIMEOUT expires before connectivity is restored, or which side of the split the current notary is on.  Also, under my system, only nodes that are online at the time of the election can run, avoiding the problem of backup nodes that are offline.

I'll address more of your objections in the next post.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #9 on: March 31, 2014, 09:10:31 AM »

Lets consider the two primary attacks on a trustee system and find technical solutions to them:

1) Notary selectively blocks certain transactions, but continues to operate.
2) Notary stops producing blocks all together.

To this I would add: 3) Notary temporarily halts block production, slows block production far beyond the target rate, or introduces extreme variance into block production times.  4) Notary creates a transaction based on early access to the precise content of a block.  5) Notary re-orders the transaction list or otherwise transforms the block in a way that changes its semantics while preserving its validity.

Requiring the notary to frequently be replaced lessens 3).  You could certainly be more proactive about policing the situation.  For example, maybe clients by default should automatically stop voting for a notary if too many recent blocks are further apart than 1.5 times the target period.  Or maybe the fees the notary receives for performing the service are docked.  Although, to enforce this, you would also have to find a way to deal with a notary that backdates the block timestamp...

An alternate solution to 2) and 3) would be to reduce NOTARY_TIMEOUT to be much smaller, e.g. twice the target block period.  This is only an okay solution; reducing NOTARY_TIMEOUT makes forks more of a risk.  So you'd probably want to require multiple confirmations.  This is why I like exchanging ideas; I really didn't envision NOTARY_TIMEOUT as long as an hour, so I just sort-of assumed you always had to worry about forks and use multiple confirmations.

For 4), you correctly note that it's certainly possible to catch habitual violations by statistical analysis:

A notary that can do 'front running' would be easily detected by the frequent inclusion of never-before-seen transactions.   If this was observed they would be voted out in short order and the attack reverts to the case of subverting the voting process.

However, infrequent violations are basically indistinguishable from network hiccups.  Whether this is a serious problem is debatable.  I'm not really perturbed by the idea of the notary occasionally trading for himself at the best bid/ask.  I'm more worried about the possibility that a notary will decide to sell this valuable capability to others, who will then have a financial incentive to vote for him.  My system avoids the problem entirely by requiring all transactions, including the notary's own transactions, to only be eligible for inclusion in a block when they've provably been broadcast to a substantial portion of the network.

The solution to 5) seems obvious, but there's a subtle problem.  (A) Avoid transaction malleability, i.e. any change to a transaction invalidates its signature.  (B) Make the block a deterministic function of the transaction set.  In practice, this means sorting the transactions.  The tricky part is in what the sort key should be.  If you sort by transaction hash, then you inadvertently trigger a computational arms race where everyone searches for a transaction with a small hash.  If you trust the notary to generate a random salt to combine with the transaction hashes, he can instead non-randomly generate a salt that puts his own transaction first.

Here's my solution:  Create a salt by hashing a sorted list of the transaction hashes.  Then the sort key for transaction i is H(tx[ i ].hash + salt).

This makes the salt chaotic -- unpredictably dependent on the entire network's actions, but deterministic so the notary can't change it to a favorable value.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #10 on: March 31, 2014, 10:00:31 AM »

I had really hoped to stay from political discussion in this thread, but it's now become inevitable, because the two proposals being discussed reflect bytemaster's optimism -- and my cynicism -- about the selection of the notary by a political election process.

Political processes only get 'personal' and thus result in infighting when there is something to be gained by getting in office.   In this case the Notary incurs a cost, but otherwise can gain nothing by having the role and they can lose it if the fail to do a very specific, unambiguous job.   These elections would not be about the 'future direction of the company'.

I'm not convinced that the notary shouldn't be compensated with a cut of the transaction fees.  Someone else proposed on this forum that those rewards should vest over time.  That way, the network can punish a notary that disappears or otherwise behaves badly by taking away their pending earnings.

There is something to be gained by getting in office:  The ability to cause the network NOTARY_TIMEOUT downtime by disappearing.  As I outlined in the previous post, your system also allows the notary to censor transactions, and gives the notary the ability to exactly match the best bid or ask.

we have two sub cases:
  1) the notary blocks attempts to vote them out of office.
  2) the notary is performing filtering that is 'politically acceptable' and thus they do not get voted out.

I would argue that case #2 is effectively consent of the governed and the only solution is a hard fork with a new trustee for the minority that are against the politically acceptable filtering.

I would argue that case #2 is effectively indifference of the governed.  Silence is not consent.

Under your system, censorship happens with a unilateral choice by the notary; a veto requires either a hard fork or an active choice of the majority (and in the latter case, can only be implemented by the indirect means of voting for another notary).

My system is the exact opposite.  Censorship requires a hard fork; the unilateral actions of a notary can't accomplish much.  No active choice is required of the majority (which is just as well, since I figure the majority usually won't be bothered to care enough to vote at all).  These features are far from accidental; the main rationale behind my design is specifically to prevent the notary from being able to censor transactions by removing the notary's ability to decide which transactions are included, and to avoid political elections when technical ones are sufficient to the task.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

Offline bytemaster

Re: drltc's Trustee Technical Discussion Thread
« Reply #11 on: March 31, 2014, 02:41:56 PM »
Have N metronodes that take turns in a deterministic manner.  Have block timestamps be deterministic.

Now all attacks must involve collusion.


Sent from my iPhone using Tapatalk
For the latest updates checkout my blog: http://bytemaster.bitshares.org
Anything said on these forums does not constitute an intent to create a legal obligation or contract between myself and anyone else.   These are merely my opinions and I reserve the right to change them at any time.

Offline stuartcharles

  • Sr. Member
  • ****
  • Posts: 281
    • View Profile
Re: drltc's Trustee Technical Discussion Thread
« Reply #12 on: March 31, 2014, 04:35:04 PM »
Have N metronodes that take turns in a deterministic manner.  Have block timestamps be deterministic.

Now all attacks must involve collusion.


Sent from my iPhone using Tapatalk

Just trying to catch up on this development. I get that if you had 100 or so metronodes scattered around the world then this would still be safe, but am i understanding you correctly that you expect these people to have powerful servers ready and waiting to verify a block when its there turn? And not get paid for doing that?

How about if only the top 1000 nodes by balance were allowed to verify blocks and in order for a block to be verified you would have to have agreement from a randomly chosen second node. That way you have the proof of stake and mining.  It would also be difficult and expensive for an attacker to have both of the randomly picked nodes.

these 1000 or 10000 or whatever number would be paid for verifying the block. They would not want to join a pool as it would mean handing over all of their coins to the pool. Like you said bitcoin is paying a fortune for security. Like this we could pay less, still be secure and there are more interesting ways to make money from bitshares than just speculation and mining.

Forgive me if this has already been suggested i have limited time to try and keep up to date on all the projects i see a future for.

Offline bitcoinerS

  • Hero Member
  • *****
  • Posts: 584
    • View Profile
Re: drltc's Trustee Technical Discussion Thread
« Reply #13 on: March 31, 2014, 07:18:18 PM »

I dislike the trustee concept. 

Me too.

Also, I just found this on http://nakamotoinstitute.org/

Quote
I've been working on a new electronic cash system that's fully peer-to-peer, with no trusted third party. Satoshi Nakamoto, 2008
>>> approve bitcoiners

Offline theoretical

Re: drltc's Trustee Technical Discussion Thread
« Reply #14 on: March 31, 2014, 07:34:17 PM »
case #1.  A notary that is doing partial filtering and is obstructing the election of a new notary.   We can mitigate this particular case by having N notaries that take turns producing blocks.   In the event that any 1 notary fails to produce a block on their turn, then the remaining notaries can vote out that particular notary after X minutes and the next notary in line will take their place.

The focus should be on how to perform the handoff in a deterministic manner if the primary notary goes rogue.   

The "N alternating notaries" idea is a really good one.  It can solve the problem of a notary disappearing and triggering a horribly long timeout.  With N notaries, one designated as primary, you can have the primary notary expected to produce a block.  But the other notaries can mitigate the damage if the primary disappears with an aggressive timeout, e.g. NOTARY_AWOL_TIMEOUT = 2 minutes.  If a secondary notary hasn't seen the primary notary within NOTARY_AWOL_TIMEOUT, then they broadcast a signed message saying "Notary appears to be AWOL."  Then select the next notary in the rotation order that's online.

You need a little more logic to avoid deadlock or inconsistent state when the primary notary produces a block at about the same time the secondary notaries start broadcasting AWOL messages.  Here are the cases:

Case 1)  A majority see the block.  This is normal operation; the primary is operating as expected.

Case 2)  A majority say the primary is AWOL.  There are two subcases:

Case 2a) Unanimous vote for AWOL.  The primary has disappeared or gotten so slow it missed NOTARY_AWOL_TIMEOUT entirely.

Case 2b) Non-unanimous majority vote for AWOL.  The primary is still online, but it generated a block just before NOTARY_AWOL_TIMEOUT.  The new block wasn't able to propagate to all secondaries in time for NOTARY_AWOL_TIMEOUT.

Case 3)  A quorum exists, but there is currently no majority.  This might happen if a minority of notaries are offline, and the remaining online notaries are split between AWOL votes and non-AWOL votes.

Case 4)  No quorum exists.

Three of these cases are obvious.  Case 1) is normal operation.  In Case 2) you kick the primary.  In Case 4) you're probably on the wrong side of a network split.  You shut down entirely until a quorum is found or manual intervention occurs.  You could select new notaries in Case 4) after a lengthy timeout like NOTARY_TIMEOUT = 1 hour, but this risks a fork if you're on the wrong side of a long-lived network split.  I think Case 4) is rare and serious enough that shutting down to wait for either a quorum or manual intervention might actually be the right thing to do.

Case 3 is the hard one; you don't know whether the offline nodes have seen the block, and you need their input in order to decide which way the vote went.  I think you need to vote again after a timeout in Case 3.
BTS- theoretical / PTS- PZxpdC8RqWsdU3pVJeobZY7JFKVPfNpy5z / BTC- 1NfGejohzoVGffAD1CnCRgo9vApjCU2viY / the delegate formerly known as drltc / Nothing said on these forums is intended to be legally binding / All opinions are my own unless otherwise noted / Take action due to my posts at your own risk

 

Google+