Re: [Bitcoin-development] Coinbase reallocation to discourage Finney attacks

[Mike Hearn <mike@plan99.net>]

[-- Attachment #1: Type: text/plain, Size: 3848 bytes --]

On Thu, Apr 24, 2014 at 1:22 PM, Jorge Timón <jtimon@monetize.io> wrote:

> > I'm using it in the same sense Satoshi used it. Honest miners work to
> > prevent double spends. That's the entire justification for their
> existence.
>
> I thought the mechanism they used to prevent double-spends was proof of
> work.
> Therefore dishonest miners where only those who mine on top of a block
> which is not the longest valid chain they've seen.
>

No! This is a misunderstanding. The mechanism they use to prevent double
spends is to *ignore double spends*. The blocks they created indicate the
ordering of transactions they saw and proof of work is used to arrive at a
shared consensus ordering given the possibility that transactions arrived
at different times.

I'm continually amazed at how many people seem to see the current algorithm
as the goal in and of itself, instead of an imperfect but workable means of
achieving the actual goal.

To distinguish this definition from your own "honest miners are those
> who decide on double-spends by mining the transaction they saw first"
>

This definition of honesty is not my own, the one Bitcoin has always used.
Obviously if Satoshi had wanted transactions to be double spendable by fee
in the mempool he would have made Bitcoin work that way, instead of coming
up with the nSequence based replacement scheme instead.

First-seen *is* a protocol rule, as much as Set-Cookie storing data in a
browser is an HTTP protocol rule. The fact that auditing compliance with it
is harder to do than some others does not make it less of a rule.

I completely disagree.
> Miner's proof of work makes transactions irreversible.


Again you are hopelessly confused. Miners that are trying to double spend
are *by definition* not making transactions irreversible, they are trying
to make transactions reversible.

Look at it this way. There is no inherent reason BitUndo has to undo only
Finney attacks. If it gets sufficient hash power it could offer undoing of
1-confirm transactions too, right? Sure it'll mostly fail but that's
already a part of its business model. Sometimes it'll get two blocks in a
row and succeed. It's a very minor tweak to what they're doing. Would you
argue these miners are still useful? After all, it's impossible to be
certain after the fact that miners built on top of the "wrong" block
because forks occur naturally.


> Even if you always had to wait for transactions to be confirmed with
> some irreversible proof of work (as described in Satoshi's
> whitepaper), it doesn't follow that "automatically resolves the
> Bitcoin experiment as a failure". I don't understand how can you
> conclude that.
>

What I said is, if you believe all miners are willing to double spend for a
fee then this resolves the experiment as a failure. This is also obvious -
if you can pay miners to go back and rewrite the chain at will, Bitcoin
doesn't work.

Consider the incentives. Let's say all miners are "smart" in your
estimation and are willing to double spend transactions for higher fees.
Because all miners follow this ridiculous policy, they should be willing to
fork the chain at any point to claim the higher fee on the new tx. After
all, although they will throw away the work they did on the previous chain,
if the fee on the new tx is high enough to balance this then it can be
profitable for them to do it.

Because a double spender can afford to give nearly all of his new tx away
in fees, this means even txns well buried in the chain can be profitably
double spent: even if the double spender gets back only 10% of the
transferred amount, if it was a big transfer for some expensive object,
they still win! They got object + 10%

Do you see now why your definition of honesty is completely broken?

[-- Attachment #2: Type: text/html, Size: 5148 bytes --]