[bitcoin-dev] Proposal to reduce mining power bill

Eric Voskuil eric at voskuil.org
Thu Jan 18 05:22:47 UTC 2018


The energy cost of mining cannot be reduced, nor is it rational to consider it “too high”.

e

> On Jan 18, 2018, at 06:34, Enrique Arizón Benito via bitcoin-dev <bitcoin-dev at lists.linuxfoundation.org> wrote:
> 
> Thanks "nullius" for your remarks. Notice my first post was not an RFC but just a blur idea to inspect if something similar had already been discussed in the group. In fact your post has helped me a lot to improve my first mail.
> 
> > Observation:  This totally destroys Bitcoin’s transaction-ordering security.  A “51% attack” could be executed by any miner who has >50% of the hashpower *proportionate to miners who are allowed to mine a particular block*, rather than >50% of *global* hashpower.  (I infer that this could be done retroactively, and wave my hands over some of the details since you did not talk about reorgs.)  The same applies as for attacks requiring 33% or 25% of total hashpower. 
> 
> I'm not sure what you are referring to in this paragraph. Imagine for example that there are a total of, let's say, 2^10 available next-coinbase/miners and the algorithm just allow 50% or 2^9 of them to mine, ¿how could it be possible that one among them could have 51% of power by chance? (Please, read comments bellow before replying)
> 
> > Potential attack, assuming that N *must* be based partly or wholly on the existing set of “next-coinbase” addresses:  A large miner could gradually push N higher, by progressively committing new “next-coinbase” addresses which differ in the next bit for all previously seen combinations of bits. Large miners would have a vast advantage over small miners, insofar as deliberately incrementing N by one more bit could only be done by a miner who creates 2^(N+1) blocks (= 2 * 2^N).  By such means, it may be possible for a very large miner to eventually lock out all other miners altogether, and monopolize all Bitcoin mining.
> 
> I do not think it would be easy even for a large miner but that made me think for an alternative algorithm. Let's introduce the concept of "spent" next-coinbase versus "un-spent" one, something like similarly to UTXO. A next-coinbase would only be valid if it has not been previously used to mine a block. Simplifying, with the spent vs unspent a large miner is restricted to a single next-coinbase as anyone else. Being more precise it's allowed a single next-coinbase for each mined new-miner-block mined creating a "thread" of mining blocks for each new new-miner-block. Schematically a thread would look like: 
> new-miner-block:next-coinbase_1 -> mined-block:next-coinbase_2 ->  ... -> (thread expired - see comment below about expiration)
> 
> In this case a large miner A with T times more power than another one B could potentially spent mining power to create T parallel threads for each thread created by miner B. A solution that could fix this issue is to allow a maximum life time for each thread expressed in number of blocks. After a given number of blocks have being mined the miner is forced to create new new-miner-block to continue participating. The algorithm to choose the life-time must be such that if a miner tries to create many parallel threads (many new-miner-block), by the time it start mining transaction blocks the first new-miner-block will start to expire, so he will be punished.
> 
> If the famous phrase "a degree of indirection solve all programming problems" I think this is an example applied to blockchain. First the consensus chooses who can participate in the next round, then selected miners participate in the round.
>  
>> Now, questions:
>> 
>> How is N determined?  By a wave of the hands?
> 
> Great question. I left it unspecified in the first mail. An algorithm comes to my mind (You are welcome to propose others). Let's imagine the list of registered non-expired next-coinbase addresses  is 2^10. The consensus checks that for N=1 there is *about* 1/2^N == 1/2 of unspent next-coinbase addresses that match (it must be close to 1/2 of the total 2^10 addresses with maybe an small +/- 1% statistical deviation). Then N=1 will be accepted. Check now for N=2. If there are 1/(2^N) = 1/4 next-coinbase addresses matching then N=2 is accepted. The algorithm continues until some "++N" fails. Initially N=0 and so all miners are welcome to the game. They all will start producing next-coinbase addresses and when there are enough different ones N will become 1, then 2, ... This system will will keep an equilibrium naturally. If new miners stop producing new new-miner-blocks, eventually the threads will expire and N will be automatically be decreased. Miners will act rationally to keep enough threads open in their own interest since that will decrease the electricity bill.
> 
> > What part of which block hash is matched against N bits?  You were quite unclear about this, and other important details.  (Much of what I say here is based on assumptions and inferences necessary to fill in the blanks.)
> 
> Thinking about it, the hash must run over "many" different blocks and it must include the next next-coinbase (to force calculating the hash after choosing a next-coinbase). Since it's not expected that many blocks are mined by the same miner in a row (maybe no more than 2 o 3) the "many" number must be for example twice as much as the expected maximum numbers of blocks that a large miner can mine in average.
>  
> > How, exactly, are reorgs handled?
> I think reorgs are not affected by this algorithm. The next-coinbase objective is just to randomly choose which miner will be allowed for the next round.
>  
> > How does this interact with the difficulty adjustment algorithm?  Indeed, how is difficulty determined at all under your scheme?
> As I see it, if the network wants to keep the same pace of new blocks each N seconds, and N=1 (only half of the miners are allowed to mine)  then difficulty/power-bill is lowered by two, for N=2 by 4, ...
> 
> > What happens to coinbase fees from a “new-miner-block”?  The way I read your scheme, the “new-miner-block” must necessarily have no payout whatsoever.  But you discuss only “new bitcoins”,which are a diminishing portion of the block reward, and will eventually reach zero.  Coinbase from fees must go somewhere; but under your scheme, a “new miner” has no payable address.
> 
> This new-miner-block will have NO transactions inside.
> 
> > What if no existing “next-coinbase” address matches?  Is N constrained to be sufficiently short that a match is guaranteed from the existing set, then that makes it trivial for large mining farms to collect addresses and further dominate (or even monopolize) the network in the attack described above.  If it isn’t, then the network could suddenly halt when nobody is allowed to mine the next block; and that would enable *this* attack:
> 
> I think the previous algorithm I mention to replace the "wave of hands" (test N=1, then N=2,... ) plus the "expiring threads" would suffice to fix it.
> 
> >  What stops a malicious miner (including a “new miner” creating a “new-miner block”) from deliberately working to create a block with a hash which does not have N bits matching any of the existing “next-coinbase” addresses?  Contra what you say, block hashes can’t be “considered random”.  Indeed, partial preimage bruteforcing of block hashes is the entire basis of mining POW.
> 
> Again, that is fixed by the previous algorithm
> 
> 
> > Asking here more generally than as for the attack described above, what stops mining farms with large hashpower from submitting many different “next-coinbase” addresses in many different blocks?  If N be small, then it should be feasible for a large mining farm to eventually register a set of “next-coinbase” addresses which match any N.  **This increases mining centralization.**  If N be large, then this creates the possibility (or raises the probability) that no address will match, and nobody will be allowed to mine the next block.
> 
> Fixed by the expiring thread model?
>  
>> How could miner anonymity be preserved under a scheme whereby each “next-coinbase” address can be linked to a previous “next-coinbase” address?  The only way to start fresh would be with a prohibitively expensive no-payout block.  Mining can be totally anonymous at present, and must so remain.  Miners are only identified by certain information they choose to put in a block header, which they could choose to change or omit—or by IP address, which is trivially changed and is never a reliable identifier.
>> 
> The anonymity decreases in the sense that if you know a next-coinbase address owner you know all its related next-coinbase for the expiring (physical-time-limited) thread. The anonymity increases in the sense that miner will consume fewer energy. Electricity bill is the easiest way today to trace miners.
> 
>  > How does this even save electricity, when there is much mining equipment (especially on large mining farms) which cannot be easily shut down and restarted?  (Allegedly, this is one reason why some big miners occasionally mine empty blocks.)  Though I suppose that difficulty would drop by unspecified means.
> 
> As explained above, the difficulty is reduced by 1/2^N for each round. And of course, miners that want to save more energy will have to adapt to put their systems on stand-by while they  are not chosen for the next round. I think based on my limited experience with ASIC mining that just by not sending new orders to the miner the power comsumption will decrease dramatically even if the equipment is still on.
>> 
>> Further observations:
>> 
>> This scheme drastically increases the upfront investment required for a new miner to start mining.  To mine even one new block all by oneself, without a pool, already requires a huge investment. 
>  
> Once introduced the concept of "expiring" thread I think he will be pretty much in the same condition. To obtain bitcoins he will first need to mine a new-miner-block to enter the game and then an standard block before the thread expires. Notice the expire time/block-length start just after the new-miner-block has been mined so the probabilities to mine before the expiration time will be proportional to its mining power, as for everyone else.  
>  
> > Add to that the uncompensated energy cost of mining that first block with *no* payout,
> 
> I think it could be clearly compensated by the save in energy for standards blocks.
> 
> >and I expect that the bar would be prohibitive to almost all new entrants.Mining costs and incentives are delicately balanced by the design of the network.  Whereas incumbents are much favoured by your scheme, further increasing miner centralization.
> 
> I don't think so after the new fixes.  What do you think? My opinion is that, based on the "theory of games", miners acting in their own interest will try to maximize "N". 
>   
> > Large incumbents could also use this to produce a mining permissions market, by selling the  private keys to committed “next-coinbase” addresses.  
> 
> With the addition of thread expiration, nobody will be really motivated to shell/buy "next-coinbase" addresses since their utility is limited.
> 
> Just a remark: Notice this algorithm reduces the electricity bill, but the hardware needed stays the same, since for each round the miner participates in, it will try to mine as fast as possible and so use as much hardware as possible. No reduction costs are expected in hardware.
> 
> 
> Best Regards,
> 
> Enrique Arizón Benito
> 
> 
> 
>> I have not even tried to imagine what oddball attacks might be possible for any miner with sufficient hashpower to deliberately cause a small reorg. 
>> 
>> -- 
>> nullius at nym.zone | PGP ECC: 0xC2E91CD74A4C57A105F6C21B5A00591B2F307E0C
>> Bitcoin: bc1qcash96s5jqppzsp8hy8swkggf7f6agex98an7h | (Segwit nested:
>> 3NULL3ZCUXr7RDLxXeLPDMZDZYxuaYkCnG)  (PGP RSA: 0x36EBB4AB699A10EE)
>> “‘If you’re not doing anything wrong, you have nothing to hide.’
>> No!  Because I do nothing wrong, I have nothing to show.” — nullius
> 
> 
> 
> 
> _______________________________________________
> bitcoin-dev mailing list
> bitcoin-dev at lists.linuxfoundation.org
> https://lists.linuxfoundation.org/mailman/listinfo/bitcoin-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linuxfoundation.org/pipermail/bitcoin-dev/attachments/20180118/d2a32353/attachment-0001.html>


More information about the bitcoin-dev mailing list