Fee discovery | Bitcointalksearch.org

Peter Todd

legendary

Activity: 1120

Merit: 1152

Quote from: DannyHamilton on June 03, 2014, 08:56:40 PM

Quote from: Peter Todd on June 03, 2014, 08:54:18 PM

Just to be clear, the above is a complete, working, implementation of the idea. Any wallet software that handles double-spends and mutability correctly can implement it without much difficulty.

I haven't had a chance to look at it yet, but I wonder how you get around the problem of the vast majority of nodes refusing to relay or accept a transaction that spends the same inputs as a transaction already in their unconfirmed transaction queue?

I don't have to actually! The replace-by-fee supporting nodes preferentially peer with each other, so there's always peers to relay your transactions. Since those nodes are well connected they have a pretty good shot of getting your tx to a miner who is accepting a higher fee but not the lower one you tried first. In practice this already works really well when bumping fees past the 0.8/0.9 fee drop, or between the various 0.8/0.8 tweaks to fee acceptance.

kingscrown

hero member

Activity: 672

Merit: 500

http://fuk.io - check it out!

decent idea but not sure if its that easy to count

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

Quote from: DannyHamilton on June 03, 2014, 08:54:51 PM

Quote from: piotr_n on June 03, 2014, 08:46:54 PM

What I proposed is simple, very easy to implement and it achieves the goal.

Great!

Implement it. I look forward to seeing it in action.

Thanks for all your hard work.

Get yourself together, man.
I told you that I look for allies in this project. And we had already established long ago that it wasn't you.

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

Quote from: Peter Todd on June 03, 2014, 08:54:18 PM

Quote from: piotr_n on June 03, 2014, 08:51:58 PM

Quote from: Peter Todd on June 03, 2014, 08:42:05 PM

See my replace-by-fee-tools, specifically bump-fee.py for an alternative based on increasing your bid. Not always convenient by itself - fee estimation is still useful - but being able to increase fees after the fact lets you be much more aggressive in your fee estimates.

Yeah, but this isn't trivial to implement and also creates additional traffic.
Sure, adjustable fees is a solution, but there is a long way for it.
Besides, one approach to the problem does not exclude another.

Just to be clear, the above is a complete, working, implementation of the idea. Any wallet software that handles double-spends and mutability correctly can implement it without much difficulty.

I understand. But it is a different approach to the solution.
Some people want to just send transaction with a fee that the minters require at a certain moment in time.
And then they want to forget about the unconfirmed transaction, knowing that it will be mined as soon as any of the miners, asked for the minimal fee, brings in a new block.

Your proposal is about waiting for a transaction to not get confirmed - and only then increasing the fee.

DannyHamilton

legendary

Activity: 3472

Merit: 4801

Quote from: Peter Todd on June 03, 2014, 08:54:18 PM

Just to be clear, the above is a complete, working, implementation of the idea. Any wallet software that handles double-spends and mutability correctly can implement it without much difficulty.

I haven't had a chance to look at it yet, but I wonder how you get around the problem of the vast majority of nodes refusing to relay or accept a transaction that spends the same inputs as a transaction already in their unconfirmed transaction queue?

DannyHamilton

legendary

Activity: 3472

Merit: 4801

Quote from: piotr_n on June 03, 2014, 08:46:54 PM

What I proposed is simple, very easy to implement and it achieves the goal.

Great!

Implement it. I look forward to seeing it in action.

Thanks for all your hard work.

Peter Todd

legendary

Activity: 1120

Merit: 1152

Quote from: piotr_n on June 03, 2014, 08:51:58 PM

Quote from: Peter Todd on June 03, 2014, 08:42:05 PM

See my replace-by-fee-tools, specifically bump-fee.py for an alternative based on increasing your bid. Not always convenient by itself - fee estimation is still useful - but being able to increase fees after the fact lets you be much more aggressive in your fee estimates.

Yeah, but this isn't trivial to implement and also creates additional traffic.
Sure, adjustable fees is a solution, but there is a long way for it.
Besides, one approach to the problem does not exclude another.

Just to be clear, the above is a complete, working, implementation of the idea. Any wallet software that handles double-spends and mutability correctly can implement it without much difficulty.

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

Quote from: Peter Todd on June 03, 2014, 08:42:05 PM

See my replace-by-fee-tools, specifically bump-fee.py for an alternative based on increasing your bid. Not always convenient by itself - fee estimation is still useful - but being able to increase fees after the fact lets you be much more aggressive in your fee estimates.

Yeah, but this isn't trivial to implement and also creates additional traffic.
Sure, adjustable fees is a solution, but there is a long way for it.
Besides, one approach to the problem does not exclude another.

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

Quote from: DannyHamilton on June 03, 2014, 08:24:25 PM

Quote from: Cryddit on June 03, 2014, 08:16:06 PM

I think the way to do fee discovery is to use data mining.

You, DeathAndTaxes, and I are all in agreement on this matter.

piotr_n seems to disagree.

More interestingly, he seems to believe that anyone that disagrees with him is not "the bitcoin community".

He keeps going on about "the Bicoin Core - you guys", and how "the bitcoin core, . . . they don't use the bitcoin community to solve the problem."

It is not so much that I disagree.
I am just saying that your solution isn't going to work, after you spend another two years on developing it.
What I proposed is simple, very easy to implement and it achieves the goal.
I would have said that it achieves the same goal, if not for the fact that your data mining solution isn't going to achieve any goal.

Peter Todd

legendary

Activity: 1120

Merit: 1152

See my replace-by-fee-tools, specifically bump-fee.py for an alternative based on increasing your bid. Not always convenient by itself - fee estimation is still useful - but being able to increase fees after the fact lets you be much more aggressive in your fee estimates.

Cryddit

legendary

Activity: 924

Merit: 1132

Quote from: DannyHamilton on June 03, 2014, 08:24:25 PM

how "the bitcoin core, . . . they don't use the bitcoin community to solve the problem."

How is solving a problem by watching to see what the community does, not "using the community to solve the problem?"

DannyHamilton

legendary

Activity: 3472

Merit: 4801

Quote from: Cryddit on June 03, 2014, 08:16:06 PM

I think the way to do fee discovery is to use data mining.

You, DeathAndTaxes, and I are all in agreement on this matter.

piotr_n seems to disagree.

More interestingly, he seems to believe that anyone that disagrees with him is not "the bitcoin community".

He keeps going on about "the Bicoin Core - you guys", and how "the bitcoin core, . . . they don't use the bitcoin community to solve the problem."

DannyHamilton

legendary

Activity: 3472

Merit: 4801

Quote from: piotr_n on June 03, 2014, 07:02:02 PM

add these options to the protocol

This doesn't require changes to the protocol. Any miner that wants to do this already can. That was your point, remember?

Quote from: piotr_n on June 03, 2014, 05:59:05 PM

And the best part is: we don't need to change Bitcon Core, in order to achieve that.

Quote from: piotr_n on June 03, 2014, 07:02:02 PM

Obviously not supported by the Bicoin Core - you guys are to proud, and have too powerful sponsors, to ask some dirty miners how much do they want to mine your so important transaction.

Other than inflating your ego, I don't understand the point of this comment. There's a lot of opinion, insults, and assumption in there, but not a lot of information.

Quote from: piotr_n on June 03, 2014, 07:02:02 PM

Alternative clients would quickly adapt such a simple protocol.

Unlikely, but you are welcome to give it a try and see what happens.

Quote from: piotr_n on June 03, 2014, 07:02:02 PM

Not even clients - you can just as well provide this information to the community as simple web page.

I've always wondered why mining pools don't already report their fee accepting rules on websites. I've also wondered why they don't create a page where users can submit a request to confirm an unconfirmed transaction by paying an out-of-band fee (for receivers that are willing to pay a fee to get their transaction confirmed), and why they haven't implemented "child pays for parent" yet. All three of those things seem likely to increase the total revenue for the mining pool.

Cryddit

legendary

Activity: 924

Merit: 1132

I think the way to do fee discovery is to use data mining.

I mean, given a set of assumptions and conditions, we want to predict the correlation of fees and delays? We should be monitoring what the relationship of fees and delays is in correlation with those conditions. IE, instead of getting miners to promise a particular behavior, we should be monitoring what miners actually do.

Important conditions are probably

number and size of tx currently not in a block
recent hash rate to difficulty ratio (block rate)
the priority (bitcoin days destroyed) of the current transaction
time of day (I bet strong repeating cycles show up on a 24-hour frequency)
time of week (I bet strong repeating cycles show up on a weekly frequency too)

Anyway, set up a statistical regression on those variables and I bet that with you'll get a good solid basis for statistical prediction.

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

Quote from: DannyHamilton on June 03, 2014, 06:52:21 PM

Miners can already do this.

Miners already don't do this.

Do you think that by creating a post here at bitointalk, suddenly miners are going to say:

"Hey, piotr_n said we should do this. We'll do what he says."

I'll watch the blocks for the next day or two, but I suspect your demands have been ignored by the mining pools.

Perhaps take this to the "Mining" sub-forum instead of the "Development & Technical Discussion" sub-forum and see what the miners have to say about it.

Note: Bitcoin is a voluntary system. We can't force the miners to implement your idea, all you can do is ask them, and then wait and see what they do.

I think I can certainly find a few people who find this idea interesting, some of them maybe mining pool operators already, others their friends.
And then, we can maybe work out some standard, add these options to the protocol and deliver a useful feature for the community.
Obviously not supported by the Bicoin Core - you guys are to proud, and have too powerful sponsors, to ask some dirty miners how much do they want to mine your so important transaction.

But other bitcoin clients/wallet? They are not as proud and all they care about is progress - to move this project forward.
Alternative clients would quickly adapt such a simple protocol.
Not even clients - you can just as well provide this information to the community as simple web page.
And then even people using the bitcon core would finally know what fee they should put, since they need to send it now.

DannyHamilton

legendary

Activity: 3472

Merit: 4801

Miners can already do this.

Miners already don't do this.

Do you think that by creating a post here at bitointalk, suddenly miners are going to say:

"Hey, piotr_n said we should do this. We'll do what he says."

I'll watch the blocks for the next day or two, but I suspect your demands have been ignored by the mining pools.

Perhaps take this to the "Mining" sub-forum instead of the "Development & Technical Discussion" sub-forum and see what the miners have to say about it.

Note: Bitcoin is a voluntary system. We can't force the miners to implement your idea, all you can do is ask them, and then wait and see what they do.

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

Quote from: NanoAkron on June 03, 2014, 06:44:40 PM

Because it will trigger a positive feedback loop of ever-rising fees.

Only if the miners create a cartel,
But if they do, with of without this feature, they can rise fees all they want.

I think you are too much paranoid about all these your evil miners.
Miners are not evil - they do very important job. And they profit in bicoins so they are actually most interested in keeping a value of this currency up.

NanoAkron

sr. member

Activity: 252

Merit: 250

Because it will trigger a positive feedback loop of ever-rising fees.

piotr_n

legendary

Activity: 2053

Merit: 1356

aka tonikt

I just don't understand why are you so much against miners giving promises on for which fee they guarantee to include a transaction in the next block.

How would such an information hurt anyone? And IMHO it would help a lot - enough to build a nice business model around it,

EDIT:
@DeathAndTaxes, I don't understand why you deleted all your previous post in this topic.

DeathAndTaxes

donator

Activity: 1218

Merit: 1079

Gerald Davis

Quote

If you dot want to buy the guarantee to be mined withing in the next block, use your system; set you fee to ten times less and pray...

I think we are done here and clearly posting was a waste of time.

Topic: Fee discovery (Read 1387 times)