Hacker News Reader

Comment by tantalor | Hacker News Reader

Can somebody explain without the technical jargon?

Bitcoin's security is based on the idea of a "block chain", which is sort of a mutually agreed-upon list of transactions. As long as all participants in the Bitcoin economy are using the same block chain, then double-spending money is (for all practical purposes) impossible.

It was recently discovered that a change in Bitcoin 0.8 caused an entry in the block chain that old versions can't process. The old versions, unable to verify the 0.8 block chain, automatically started their own block chain based on the last block they could deal with.

  0.8-readable block chain continues
  |
  ---> |
       |  | <- 0.7-readable block chain splits off
       | /
       |/
       *  <- block that old versions can't process
       |
       |
  (history)

For technical reasons, it was decided that the 0.8 block chain would be abandoned, and all clients should switch to the "forked" block chain that was accidentally created by old clients.

  dead    | <- 0.7-readable block chain is now "official"
  |       |
  ---> _  |
       |  |
       | /
       |/
       *  <- block that old versions can't process
       |
       |
  (history)

This causes a problem, because until that switch happens, a person could issue a transaction to a 0.8 client that would be accepted only by clients running the "bad" 0.8 blockchain.

The OP states that he has done this successfully. He submitted a transaction to a BTC<->USD service running 0.8. The service, checking the bad block chain, thought it had received Bitcoins and therefore sent him USD. But his transaction was never sent to the "official" block chain, so according to the rest of the Bitcoin world, he still has his original Bitcoins.

This is significant because prevention of double-spending is a fundamental feature of Bitcoin. If it can be bypassed, then the entire Bitcoin economy is endangered.

What happened to bitcoins 'mined' on the 0.8 chain in the meantime? Did people lose their currency?

Good question. I know that any transactions in forked chains that have been orphaned are included as transaction if it hasn't been double spent on the other chain.

But what happens to the mining reward, I don't know. Anyone?

The mining rewards on the 0.8 branch went poof

A mining reward is actually just a special transaction that the miner includes in blocks they find that let them pay themselves out of thin air.

So, the mining reward transactions on the 0.8 branch only exist on that branch, since the 0.7 chain had its own reward transactions made out to its own miners. When the network reached consensus on the 0.7 branch, money that miners on 0.8 thought they earned just disappeared (other clients stopped acknowledging their reward bitcoins).

Some pool operators lost out too, since they paid people who found blocks on the 0.8 branch.

How would a pool operator lose? They would pay out on the 0.8 branch, in bitcoins. Do pool operators pay in real-world currency like USD?

They do indeed pay in bitcoins. The payment transactions during the fork weren't lost, however-- they were confirmed on both branches of the chain. The only transactions that disappeared were the block reward transactions on 0.8.

If pools paid miners using bitcoins from their own mined blocks, this problem would go away-- that's not a bad idea!

edit: I asked -- some pool operators apparently already do this. The ones that lost out are the ones that don't.

Yes. For the 0.7 chain, it's as if the 0.8 chain never happened. All transactions and all mining in that part of the chain is irrelevant now.

Less technical, and mostly correct (I left plenty of room to nitpick)

Imagine a bank that, instead of having a central computer that all the ATM's call into, relies on each ATM to track every customer's balance. When I try to withdraw money from any ATM in the network, it checks its records of my balance, then broadcasts the transaction to all of the other ATM's.

Since the network is distributed, an evil-minded customer might send his brother out with a second ATM card, and time two withdrawals of the complete account balance, to take place at virtually the same second. Since it takes a few seconds for each transaction to be disseminated, there is a chance that both withdrawals will succeed before either ATM learns of the other transaction -- thus leaving the bank short by that amount, and me and my bro with twice the cash that we should have.

So, the bank institutes a policy of nominating one ATM at random to issue a report of all transactions that it has received once every minute. This report is guaranteed to be self-consistent -- if a customer's balance would go below zero, then that withdrawal is not included in the report. Ties are broken by time stamp down to the microsecond, and then by random chance in the unlikely event of a tie. Now, an ATM can ensure it's not being duped by transmitting the proposed withdrawal, then waiting to receive the next report and verifying that the withdrawal is included in it, then dispensing the money.

Then one day, a major network cable linking the East Coast and West Coast ATM clusters breaks. Both halves continue right on operating (they are fault-tolerant), but now transactions are only spread to half of the network, and each minute two independent reports are issued, again to only half the network each. Now, if you live in NYC and your brother is in LA, you can each withdraw the full balance from your account, and both transactions will be verified since they are being processed by two independent networks.

Later, when the network cable is patched, the two halves have to merge their respective reports and one of the withdrawals will be thrown out as invalid -- but you still have twice the cash.

I can't find a cite, but I read an article somewhere that some transit fare systems work like that. Each node tracks every customer's balance and broadcasts transactions. That's how the turnstile is always able to respond instantly, because the validation is local. It's possible to duplicate a card and swipe it simultaneously at different stations to double-spend. It's considered an acceptable leak in the system given the high effort, infungibility of the good, and near-zero actual cost of a free-rider.

Where subways diverge from Bitcoin is that sold goods can represent a real loss, can be resold thus creating the incentive to scam, and that Bitcoin is anonymous where subway cards aren't (security cameras) so repeated instances of this crime can be caught. So the consequences of double-spending with Bitcoin are vastly higher and the network can't just choose to live with it.

Reply on news.ycombinator.com