Mining - exponential backoff | Bitcointalksearch.org

Altoidnerd

sr. member

Activity: 406

Merit: 251

http://altoidnerd.com

Quote from: justusranvier on January 08, 2014, 05:00:57 AM

How about you invent your own cryptocurrency where miners need to be approved by some authority to participate and just leave Bitcoin alone?

Here's the thing, OP. It may strike you as "form over function" that we reject solutions to problems that require a trusted 3rd party. However the conclusion reached after thousands of hours of thinking about any of bitcoin's problems is always the same.

There can be no "miner ID" issuing authority. Whether or not miner collusion can be discouraged keeping mining decentralized IS worth discussing. The exponential back off would certainly be nice, but as was stated:

1) suppose you could diminish a miner's reward for mining "too much". The miner then has an incentive to create a new/several identities

2) if you could (in principle) make those identities expensive, gmaxwell suggested that the complicated strategy which balances (1) and (2) is for miners to clump together nonetheless.

The above problem, which is contained in gmaxwell's reply to you, still presents itself even if you think of a method of identifying miners by a means other than a central authority. Hopefully this can clarify some of your questions.

yamamushi

newbie

Activity: 10

Merit: 0

Quote from: Rannasha on January 08, 2014, 04:47:26 AM

Falling back on a central authority, such as the Bitcoin Foundation, like you suggested is not acceptable. The Bitcoin Foundation is not in control of Bitcoin and plenty of people don't even like the Foundation (plus it's a purely American group).

Though discussions like this make it pretty clear that plenty of people believe the Bitcoin Foundation is the central authority.

I fail to see why so many newcomers see the lack of centralization as a fault with Bitcoin.

justusranvier

legendary

Activity: 1400

Merit: 1013

Quote from: deuteragenie on January 07, 2014, 06:07:52 AM

So, how about registering yourself with the bitcoin foundation as a miner, providing contact details, and a phone number, receiving an SMS to confirm the authenticity of your credentials ?

How about you invent your own cryptocurrency where miners need to be approved by some authority to participate and just leave Bitcoin alone?

Rannasha

hero member

Activity: 728

Merit: 500

Quote from: deuteragenie on January 08, 2014, 03:39:44 AM

My apologies if you feel that I am wasting your time and/or others. I was under the impression that it could be a useful discussion to have.

It could be a useful discussion, but only if you can come up with a way where miners/pools can not easily create new identities without relying on a central authority.

Like gmaxwell said, if it is economical for a large pool to create new identities to work around any form of block-finding-speed-restriction (such as the proposed exponential backoff), they will do this.

Falling back on a central authority, such as the Bitcoin Foundation, like you suggested is not acceptable. The Bitcoin Foundation is not in control of Bitcoin and plenty of people don't even like the Foundation (plus it's a purely American group). But any central authority being able to control who can mine a block is a polar opposite to the design goals of Bitcoin.

While centralization in some cases, such as exchanges, is hard to avoid, in those cases it typically involves optional services. Mining is at the foundation of Bitcoin. If that is controlled by a single authority, Bitcoin loses its purpose.

Altoidnerd

sr. member

Activity: 406

Merit: 251

http://altoidnerd.com

>>Hence, a very interesting question is the following: "Is it possible to "expensively" create new identities in a P2P way" ?"

The answer is implicit in gmaxwell's replies - perhaps what you are missing is that mining is a hopelessly parallel, additive process. Two miners can be combined into a single identity at a low cost. Any "expensive identity" can be agreed upon to be shared by miners, leaving the network unable to distinguish miners from their pools.

deuteragenie

newbie

Activity: 36

Merit: 0

Thanks.

Quote from: gmaxwell on January 07, 2014, 02:55:23 PM

Making it expensive to create identities directly undermines the goal of having mining not controlled by particular parties.

I do not understand this. The two issues appear disconnected to me: one is how to create "expensively" identities for miners. The other question is how to preserve decentralization.
Hence, a very interesting question is the following: "Is it possible to "expensively" create new identities in a P2P way" ?"

Quote from: gmaxwell on January 07, 2014, 02:55:23 PM

Moreover, any such scheme has some optimal strategy where the cost of identities vs exclusion are counter balanced, and in general complicated strategy favors larger parties (who can afford to figure out and implement the strategy).

I also do not understand this. I gave above one possible example on how to create identities expensively" that would not favor larger parties. Other centralized examples could be: prrof that you run a marathon, that you achieved a certain ELO at chess, that you have a piece of art exposed in a museum etc.

Quote from: gmaxwell on January 07, 2014, 02:55:23 PM

...it wouldn't work because parties could permit others to use their identity. (E.g. to mine at this pool you must obtain an identity and allow the pool to use it).

As explained above, this would need to be combined with some exponential backoff mechanism, in which case it does not matter whether it is permitted to share identities.

Quote from: gmaxwell on January 07, 2014, 02:55:23 PM

In any case, it's hard to comment with more than these sorts of arm-waving generalities against a proposal which is itself an arm-waving generality.
Please take the time to concretely work out how to intend to issue and manage these identities, and what effect(s) you have them to have before asking other people to spend time considering your idea. Otherwise any reviewers time investment is unboudned as you can continually reveal previously undisclosed features which patch over their concerns one by one.

My apologies if you feel that I am wasting your time and/or others. I was under the impression that it could be a useful discussion to have.

gmaxwell

staff

Activity: 4284

Merit: 8808

Making it expensive to create identities directly undermines the goal of having mining not controlled by particular parties. Moreover, any such scheme has some optimal strategy where the cost of identities vs exclusion are counter balanced, and in general complicated strategy favors larger parties (who can afford to figure out and implement the strategy). Also even if you had a perfect identity scheme (which could trivially be used to impose external control by ordering the identity holders to do things, but we'll ignore that) it wouldn't work because parties could permit others to use their identity. (E.g. to mine at this pool you must obtain an identity and allow the pool to use it).

In any case, it's hard to comment with more than these sorts of arm-waving generalities against a proposal which is itself an arm-waving generality.

Please take the time to concretely work out how to intend to issue and manage these identities, and what effect(s) you have them to have before asking other people to spend time considering your idea. Otherwise any reviewers time investment is unboudned as you can continually reveal previously undisclosed features which patch over their concerns one by one.

deuteragenie

newbie

Activity: 36

Merit: 0

Oh, I realize very well that this is not fully in-line with the 'philosophy'.

However, if the ideal goal of perfect decentralization has to be balanced with the overall security of all users, and especially that provided by the protocol related to double-spending, then, as a temporary solution and until one finds something better, I would go for centralizing registrations.

Kazimir

legendary

Activity: 1176

Merit: 1011

Quote from: deuteragenie on January 07, 2014, 06:07:52 AM

So, how about registering yourself with the bitcoin foundation as a miner, providing contact details, and a phone number, receiving an SMS to confirm the authenticity of your credentials ?

Sounds like the opposite of decentralization. Very bad idea, and totally not in line with the Bitcoin philosophy.

Sukrim

legendary

Activity: 2618

Merit: 1007

That creates a centralized point and once you allow this, mining itself is not really necessary for money distribution to begin with.

deuteragenie

newbie

Activity: 36

Merit: 0

Thank you for your explanation.

I agree that the difficulty is to determine who mines. As a principle, it should be prohibitively expensive to create multiple identities for miners.
So, ip addresses, public keys, dns, etc. won't work as they can essentially be created massively.
Mining specific address/key for miners won't work either because it would form a impossible to jump over barrier of entry for small miners.

So, how about registering yourself with the bitcoin foundation as a miner, providing contact details, and a phone number, receiving an SMS to confirm the authenticity of your credentials ?
Would this be a sufficiently 'expensive' way to create new miner credentials ?

I guess many people would gladly contribute a few BCs to support the extra work/costs of the foundation that this would entail.

Then we can talk about flow control.

Rannasha

hero member

Activity: 728

Merit: 500

Ideas such as this don't work. The problem lies in the inability to properly identify who mined a block. If a miner or mining pool uses a different coinbase-address for each block, does not sign it's blocks with a message and regularly transmits their new blocks from a new IP address, how are you going to enforce such a miner to back off after finding a block?

deuteragenie

newbie

Activity: 36

Merit: 0

It seems that there could be several issues related to excessive share of some miners.

Has a solution like exponential backoff (http://en.wikipedia.org/wiki/Exponential_backoff) been explored, whereby a miner has to wait before it is allowed to mine a block, if it has mined a large number of blocks before ?

Keeping on the TCP analogy, would it be possible to foresee slow start (http://en.wikipedia.org/wiki/Slow_start) for those miners that do not mine often ?

Could the combination of both lead to much more fragmented/decentralized mining situation?

Topic: Mining - exponential backoff (Read 926 times)