@fredsbend Mining has real costs, but there is no marginal cost for an extra transaction. The only actual proportional costs are for bandwidth for relay and CPU/memory for validation, but you have to do those anyway. If one miner doesn't include the transaction, another will. And everyone will still need to pay those costs.

The fees you saw years ago were just a result of policies. Those aren't maintainable in the long run when there is money involved.

There is no reason to assume miners would reject transaction below some arbitrary set limit if it loses them actual money.

@PieterWuille That's exactly what I mean. Miners will charge the highest fee they can get, per the law of supply and demand.

Convention was based on policy, but as fees replace block rewards, supply and demand kick in.

@PieterWuille Supply and demand seems reason enough to me. That's why fees are so high now. Blocks are being filled up, so they can charge more now. Supply down, demand up or stagnant, price goes up.

Which could actually lead to interesting operations once fees take over. Things like only mining at certain hours, when more transactions are occurring naturally in the market.

I just gave an argument why the fees will be zero if there is less demand than capacity.

Every miner that requires more than a marginal fee will lose money compared to his competitors by just losing some income, but still perform exactly the same work.

@fredsbend Depends on how it's done. Exchanges yes (although different tradeoffs). But Lightning Network actually increases decentraliztion as people can transact with each other where only a few hops even know about it, yet with very minimal (anonymous) information and as opposed to the whole world having the details stored for eternity.

@fredsbend I have several netflix series to upload, to prove you wrong. Just make it cheap enough. And for a bit higher fee I'll upload my childhood photo/video collection (encrypted). And if they're still a bit higher, I'll just start an online casino that let's you bet 0.000001 and does immediate payouts. Maybe I'll convince Starbucks to accept a Bitcoin payment for each of the billion cups of coffee they sell every day.

@fredsbend Miners don't "charge" fees. Users offer fees to try to outbid other users. Miners simply take the highest fees available until the block is full. If they do anything else, like for example if a miner decides not to take the offer and not fill his block, he simply loses money. The next miner will take that offer.

@fredsbend Have a look at a supply demand curve where the supply is way larger than the demand and see what happens to the price.

@fredsbend Fees aren't actually as high as they seem. All you need to pay is a little bit more than the lowest fee in the next block. There are reasons that that is actually hard to estimate (predict) but also a lot of wallets and services are simply not smart enough to even do a decent attempt. So they end up overpaying a lot. SegWit and other improvements actually make this estimation simpler but that needs time a bit of time.

@fredsbend Another reason why fee estimation is hard is precisely because so many are overpaying, and another is the mempools are sometimes almost cleared of reasonable fee transactions. A consistent pressure on the mempool is a good thing.

@fredsbend That already happens. Which is another reason wallets and services need to work on their fee handling. A lot of transactions can simply wait a few hours or even days and thus take advantage of the lower fees. Smoothing out the pressure on the mempool, making it more predictable.

Hi Guys, I have been reading about bitcoin (coursera, white paper) and the underlying mechanics, and I have some questions about proof-of-work which is used in Bitcoin protocol. From what I understand, proof-of-work is used as a means to select a random node which gets to broadcast its block to everyone else

But is this necessary for preventing an attack? Because if a block has already been accepted by the other nodes, how would a malicious node be able to affect a change in the chain of other nodes, assuming a majority of the nodes are honest

Not sure if my question makes sense! Another way to put it would be, let's say there's a hypothetical algorithm to achieve distributed consensus (which is byzantine fault tolerant) without selecting a random node, would the concept of proof-of-work/stake/anything else be necessary in that scenario?