Bitcoin Days Destroyed as block selection heuristic
8 messages, 11 to 12 Sep 2015. Started by Dave Scotese.
Dave Scotese (3), Jorge Timón (2), Christophe Biocca (2), Vincent Truong
Rather than (promising to, and when they don't actually, at least pretending to) use the first-seen block, I propose that a more sophisticated method of choosing which of two block solutions to accept. Essentially, a miner receiving two solutions at the same height would compute a weighted sum of bitcoin-days-destroyed (transactions received earlier get higher weights) of whatever transactions are in a block *and also* were in the miner's mempool *before* the first solution arrived. Whichever block has more wins.
This strategy avoids allowing miners to use private transactions to mess with the blockchain. It also makes an empty block far less attractive because it is easily replaced, all the way until the next block locks it in. Any block-selection heuristic can be gamed, but I believe that using a weighted sum of BTCDD is harder to game than using block propagation timing.
I asked Can Bitcoin Days Destroyed be a better resolution mechanism for competing blocks? <http://bitcoin.stackexchange.com/questions/39226/can-bitcoin-days-destroyed-be-a-better-resolution-mechanism-for-competing-blocks> on the stackexchange bitcoin site in order to collect objections to and problems with this idea, and have not found any that I haven't addressed. The best objection is that *maybe* empty blocks and selfish mining are either good for bitcoin, or else they are so minimally bad that no effort ought to be expended in preventing them.
If anyone here thinks this is a good idea, and no one can offer reasons it's a bad idea, I will probably start working on an implementation. I'm really slow though, so ping me if it looks like fun to you.
notplato
On Sep 11, 2015 12:27 PM, "Dave Scotese via bitcoin-dev" < bitcoin-dev@lists•linuxfoundation•org> wrote:
Quoted text (2 lines)
> > Rather than (promising to, and when they don't actually, at least
pretending to) use the first-seen block, I propose that a more sophisticated method of choosing which of two block solutions to accept.
There's already a criterion to chose: the one with more work (in valid blocks) on top of it.
It's pretty obvious that Dave is suggesting an alternate tie-breaker:
Quoted text (1 line)
> It also makes an empty block far less attractive because it is easily replaced, all the way until the next block locks it in.
I do see a problem with the proposal. Right now, when a miner sees a new block with the most work and there are no ties, it is always a good idea to build on top of it (unless they're in the middle of building a private chain, or other pathological cases).
With this new heuristic (assuming it is actually followed by a good chunk of people), a miner can reasonably know whether or not they can safely mine a sibling of the block instead. When enough widely propagated transactions exist, and the block to orphan is small, there's minimal risk in mining a sibling block instead of a child block (the only extra risk is in someone else mining a child block right around the time we suceed in mining a siblish block, where we'll definitely be orphaned instead of ~50% of the time).
Because the risk can be measured and is sometimes very small, it will then be profitable for a miner to orphan a small non-empty block and double-spend some confirmed transactions whenever the block confirming them is easily replaced. This lowers the security of 1-conf transactions.
Mind you, that risk doesn't apply if we prefer non-empty blocks to empty blocks and leave it at that, or only switch if the new block doesn't double spend transactions in the old one, so it's a fixable issue.
On 11 September 2015 at 12:32, Jorge Timón <bitcoin-dev@lists•linuxfoundation•org> wrote:
Quoted text (17 lines)
> > On Sep 11, 2015 12:27 PM, "Dave Scotese via bitcoin-dev" > <bitcoin-dev@lists•linuxfoundation•org> wrote: >> >> Rather than (promising to, and when they don't actually, at least >> pretending to) use the first-seen block, I propose that a more sophisticated >> method of choosing which of two block solutions to accept. > > There's already a criterion to chose: the one with more work (in valid > blocks) on top of it. > > > _______________________________________________ > bitcoin-dev mailing list > bitcoin-dev@lists•linuxfoundation•org > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev >
On Sep 11, 2015 1:18 PM, "Christophe Biocca" <christophe.biocca@gmail•com> wrote:
Quoted text (2 lines)
> > It's pretty obvious that Dave is suggesting an alternate tie-breaker:
I thought he was proposing a new consesnsus rule. I see, this would be just a policy validation that everybody would be free to ignore (like the "first seen" spend conflict tx replacement policy).
I don't see how miners would benefit from running this policy so I would not expect them to run it in the long run (like the "first seen" spend conflict tx replacement policy). If miners don't use it, I don't see how users can benefit from running that policy themselves. They will still have to keep waiting some block confirmation to exponentially reduce the chances of a successful double-spend attack with each new confirmation (as explained in the bitcoin white paper).
Quoted text (4 lines)
> Mind you, that risk doesn't apply if we prefer non-empty blocks to > empty blocks and leave it at that, or only switch if the new block > doesn't double spend transactions in the old one, so it's a fixable > issue.
How do you know which of 2 blocks with the same height is "newer"?
Quoted text (8 lines)
> On 11 September 2015 at 12:32, Jorge Timón > <bitcoin-dev@lists•linuxfoundation•org> wrote: > > > > On Sep 11, 2015 12:27 PM, "Dave Scotese via bitcoin-dev" > > <bitcoin-dev@lists•linuxfoundation•org> wrote: > >> > >> Rather than (promising to, and when they don't actually, at least > >> pretending to) use the first-seen block, I propose that a more
sophisticated
Quoted text (11 lines)
> >> method of choosing which of two block solutions to accept. > > > > There's already a criterion to chose: the one with more work (in valid > > blocks) on top of it. > > > > > > _______________________________________________ > > bitcoin-dev mailing list > > bitcoin-dev@lists•linuxfoundation•org > > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev > >
Quoted text (1 line)
> How do you know which of 2 blocks with the same height is "newer"?
From the particular node's perspective. I'm aware there is no possibility of consistent global ordering.
Dave's code is about switching blocks (instead of continuing on the existing one), and, in that context, "old" means the first sibling the node saw, and "new" is any subsequent block. I will disambiguate this in the future, because I'm clearly confusing at least 1 person.
Quoted text (1 line)
> I don't see how miners would benefit from running this policy so I would not expect them to run it in the long run (like the "first seen" spend conflict tx replacement policy).
There's always a default, and if miners don't have any overriding reason to change, they'll likely stick to it. Which is why Dave started his statement with:
Quoted text (1 line)
> Rather than (promising to, and when they don't actually, at least pretending to) use the first-seen block
Clearly recognizing that any changed logic is non-binding.
On 11 September 2015 at 14:37, Jorge Timón <jtimon@jtimon•cc> wrote:
Quoted text (46 lines)
> > On Sep 11, 2015 1:18 PM, "Christophe Biocca" <christophe.biocca@gmail•com> > wrote: >> >> It's pretty obvious that Dave is suggesting an alternate tie-breaker: > > I thought he was proposing a new consesnsus rule. I see, this would be just > a policy validation that everybody would be free to ignore (like the "first > seen" spend conflict tx replacement policy). > > I don't see how miners would benefit from running this policy so I would not > expect them to run it in the long run (like the "first seen" spend conflict > tx replacement policy). > If miners don't use it, I don't see how users can benefit from running that > policy themselves. > They will still have to keep waiting some block confirmation to > exponentially reduce the chances of a successful double-spend attack with > each new confirmation (as explained in the bitcoin white paper). > >> Mind you, that risk doesn't apply if we prefer non-empty blocks to >> empty blocks and leave it at that, or only switch if the new block >> doesn't double spend transactions in the old one, so it's a fixable >> issue. > > How do you know which of 2 blocks with the same height is "newer"? > >> On 11 September 2015 at 12:32, Jorge Timón >> <bitcoin-dev@lists•linuxfoundation•org> wrote: >> > >> > On Sep 11, 2015 12:27 PM, "Dave Scotese via bitcoin-dev" >> > <bitcoin-dev@lists•linuxfoundation•org> wrote: >> >> >> >> Rather than (promising to, and when they don't actually, at least >> >> pretending to) use the first-seen block, I propose that a more >> >> sophisticated >> >> method of choosing which of two block solutions to accept. >> > >> > There's already a criterion to chose: the one with more work (in valid >> > blocks) on top of it. >> > >> > >> > _______________________________________________ >> > bitcoin-dev mailing list >> > bitcoin-dev@lists•linuxfoundation•org >> > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev >> >
Yes, this proposal is a policy that everyone would be free to ignore. I should have introduced the situation in which this *unenforceable* policy makes sense to me. Here it is:
Every miner is listening for valid block solutions but might receive two valid blocks and then they have to decide which one to use. Choosing the one you saw first is the default behavior. In that situation, we'd all like everyone to choose the same block. I propose that a better heuristic than "first seen" is to compare the BTCDD, *but only of transactions you already have in your mempool*, and
*weight the BTCDD so that txns you got earlier are more important.* The heuristic is most useful when the two blocks are received within a small window of time, opting for the first-seen rule otherwise. I assume many miners have an idea of how long it takes for anyone's new block to get across the network, and more specifically, the range of times it takes for new solutions to get to themselves. During this little time window, the chances are 50/50 that they'll choose the right block. If the default behavior were to use BTCDD during that time window (one second? I have no idea!), then the chances would be significantly better.
I think Jorge is right that it doesn't benefit miners. It doesn't hurt them either, unless they are trying to do selfish mining. Well, it benefits them in terms of increased bitcoin stability by A) making it easier for clients to decide which block is valid when they see two competing with each other, B) motivating miners to add transactions instead of mining empty blocks, C) severely decreasing the utility of any global private network of nodes intended to spread selfishly-mined blocks, and D) motivating miners to stay well-connected so that they get transactions quickly.
I sent this to the list because it is only useful if it is set as default behavior since most miners leave the defaults alone, and the benefits don't materialize unless a majority follows the policy.
On Fri, Sep 11, 2015 at 11:37 AM, Jorge Timón <jtimon@jtimon•cc> wrote:
Quoted text (47 lines)
> > On Sep 11, 2015 1:18 PM, "Christophe Biocca" <christophe.biocca@gmail•com> > wrote: > > > > It's pretty obvious that Dave is suggesting an alternate tie-breaker: > > I thought he was proposing a new consesnsus rule. I see, this would be > just a policy validation that everybody would be free to ignore (like the > "first seen" spend conflict tx replacement policy). > > I don't see how miners would benefit from running this policy so I would > not expect them to run it in the long run (like the "first seen" spend > conflict tx replacement policy). > If miners don't use it, I don't see how users can benefit from running > that policy themselves. > They will still have to keep waiting some block confirmation to > exponentially reduce the chances of a successful double-spend attack with > each new confirmation (as explained in the bitcoin white paper). > > > Mind you, that risk doesn't apply if we prefer non-empty blocks to > > empty blocks and leave it at that, or only switch if the new block > > doesn't double spend transactions in the old one, so it's a fixable > > issue. > > How do you know which of 2 blocks with the same height is "newer"? > > > On 11 September 2015 at 12:32, Jorge Timón > > <bitcoin-dev@lists•linuxfoundation•org> wrote: > > > > > > On Sep 11, 2015 12:27 PM, "Dave Scotese via bitcoin-dev" > > > <bitcoin-dev@lists•linuxfoundation•org> wrote: > > >> > > >> Rather than (promising to, and when they don't actually, at least > > >> pretending to) use the first-seen block, I propose that a more > sophisticated > > >> method of choosing which of two block solutions to accept. > > > > > > There's already a criterion to chose: the one with more work (in valid > > > blocks) on top of it. > > > > > > > > > _______________________________________________ > > > bitcoin-dev mailing list > > > bitcoin-dev@lists•linuxfoundation•org > > > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev > > > >
Would this alter the way txns will be prioritised in order to try to win? You would always pick txns with the best BTCDD to maximize your chances of being the block to build on.
I see this as potentially being a bad outcome for bitcoin's fungibility. On Sep 12, 2015 5:26 AM, "Dave Scotese via bitcoin-dev" < bitcoin-dev@lists•linuxfoundation•org> wrote:
Quoted text (104 lines)
> Yes, this proposal is a policy that everyone would be free to ignore. I > should have introduced the situation in which this *unenforceable* policy > makes sense to me. Here it is: > > Every miner is listening for valid block solutions but might receive two > valid blocks and then they have to decide which one to use. Choosing the > one you saw first is the default behavior. In that situation, we'd all > like everyone to choose the same block. I propose that a better heuristic > than "first seen" is to compare the BTCDD, *but only of transactions you > already have in your mempool*, and > > *weight the BTCDD so that txns you got earlier are more important.* > The heuristic is most useful when the two blocks are received within a > small window of time, opting for the first-seen rule otherwise. I assume > many miners have an idea of how long it takes for anyone's new block to get > across the network, and more specifically, the range of times it takes for > new solutions to get to themselves. During this little time window, the > chances are 50/50 that they'll choose the right block. If the default > behavior were to use BTCDD during that time window (one second? I have no > idea!), then the chances would be significantly better. > > I think Jorge is right that it doesn't benefit miners. It doesn't hurt > them either, unless they are trying to do selfish mining. Well, it > benefits them in terms of increased bitcoin stability by A) making it > easier for clients to decide which block is valid when they see two > competing with each other, B) motivating miners to add transactions instead > of mining empty blocks, C) severely decreasing the utility of any global > private network of nodes intended to spread selfishly-mined blocks, and D) > motivating miners to stay well-connected so that they get transactions > quickly. > > I sent this to the list because it is only useful if it is set as default > behavior since most miners leave the defaults alone, and the benefits don't > materialize unless a majority follows the policy. > > On Fri, Sep 11, 2015 at 11:37 AM, Jorge Timón <jtimon@jtimon•cc> wrote: > >> >> On Sep 11, 2015 1:18 PM, "Christophe Biocca" <christophe.biocca@gmail•com> >> wrote: >> > >> > It's pretty obvious that Dave is suggesting an alternate tie-breaker: >> >> I thought he was proposing a new consesnsus rule. I see, this would be >> just a policy validation that everybody would be free to ignore (like the >> "first seen" spend conflict tx replacement policy). >> >> I don't see how miners would benefit from running this policy so I would >> not expect them to run it in the long run (like the "first seen" spend >> conflict tx replacement policy). >> If miners don't use it, I don't see how users can benefit from running >> that policy themselves. >> They will still have to keep waiting some block confirmation to >> exponentially reduce the chances of a successful double-spend attack with >> each new confirmation (as explained in the bitcoin white paper). >> >> > Mind you, that risk doesn't apply if we prefer non-empty blocks to >> > empty blocks and leave it at that, or only switch if the new block >> > doesn't double spend transactions in the old one, so it's a fixable >> > issue. >> >> How do you know which of 2 blocks with the same height is "newer"? >> >> > On 11 September 2015 at 12:32, Jorge Timón >> > <bitcoin-dev@lists•linuxfoundation•org> wrote: >> > > >> > > On Sep 11, 2015 12:27 PM, "Dave Scotese via bitcoin-dev" >> > > <bitcoin-dev@lists•linuxfoundation•org> wrote: >> > >> >> > >> Rather than (promising to, and when they don't actually, at least >> > >> pretending to) use the first-seen block, I propose that a more >> sophisticated >> > >> method of choosing which of two block solutions to accept. >> > > >> > > There's already a criterion to chose: the one with more work (in valid >> > > blocks) on top of it. >> > > >> > > >> > > _______________________________________________ >> > > bitcoin-dev mailing list >> > > bitcoin-dev@lists•linuxfoundation•org >> > > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev >> > > >> > > > > -- > I like to provide some work at no charge to prove my value. Do you need a > techie? > I own Litmocracy <http://www.litmocracy.com> and Meme Racing > <http://www.memeracing.net> (in alpha). > I'm the webmaster for The Voluntaryist <http://www.voluntaryist.com> > which now accepts Bitcoin. > I also code for The Dollar Vigilante <http://dollarvigilante.com/>. > "He ought to find it more profitable to play by the rules" - Satoshi > Nakamoto > > _______________________________________________ > bitcoin-dev mailing list > bitcoin-dev@lists•linuxfoundation•org > https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev > >
Quoted text (8 lines)
> > From the Bitcoin wiki page on transaction fees > <https://en.bitcoin.it/wiki/Transaction_fees#Technical_info>: > > Transaction priority is calculated as a value-weighted sum of input age, > divided by transaction size in bytes: priority = > sum(input_value_in_base_units * input_age)/size_in_bytes >
If I read that correctly, that is directly proportional to BTCDD, so whatever effect concerns you has already been built into the code.
On Fri, Sep 11, 2015 at 3:21 PM, Vincent Truong < vincent.truong@procabiak•com> wrote:
Quoted text (6 lines)
> Would this alter the way txns will be prioritised in order to try to win? > You would always pick txns with the best BTCDD to maximize your chances of > being the block to build on. > > I see this as potentially being a bad outcome for bitcoin's fungibility. >