Commons license.

Your support will help MIT OpenCourseWare

continue to offer high-quality educational resources for free.

To make a donation or to view additional materials

from hundreds of MIT courses, visit MIT OpenCourseWare

at ocw.mit.edu.

TADGE DRYJA: OK, so today I will talk about mostly fees, which

is an interesting and somewhat contentious

issue in Bitcoin and all these other systems.

So schedule for today, problem set 3 description,

we'll talk about fees, these unknowable acronyms--

Child Pays for Parent, Replaced By Fee--

and then talk a little bit about long-term incentives

in Bitcoin, and how that relates to fees.

OK, so the problem set 3 is to grab UTXOs.

Hopefully it's fun.

You're going to be making transactions

on the real Bitcoin test network.

So one of the parts of it that hopefully is not too hard

is downloading and install Bitcoin D. So I'll put a link.

The current, most up-to-date version is 0.16.0.

I believe the next version will just

be called 17, because a lot of the developers are like,

this is annoying, it's always starting with a zero,

it's like 0.1, 0.2, 0.3.

And they're like, well, at some point,

should we get to version 1?

And they're like, well, everyone's

going to think version 1 means it's better.

Let's just call the next version 17 instead of 0.17.

So download that, and then sync up to testnet.

If you just run it by default, it

will sync up to the main network, which will download

something like 170 gigabytes.

You don't want that.

And the homework assignment cannot be completed on the main

network.

There's no free money--

I haven't put any free money on the main network.

However, on the test network, it'll work.

And it's about 11 gigabytes of download.

So I think most people, on your laptop,

probably have 11 gigs free.

It's doable.

And it's kind of fun to see how it works.

So if you don't have access to 11 gigs,

you can probably get an Athena node to do it.

It's not too big.

You can run a tiny little node with not much RAM, not

much space, and it'll work.

You can also sync up on a Raspberry Pi.

It just takes a little longer.

So I don't think it'll be too hard to do.

You can probably get around--

sort of the goal of the assignment

is to check out the actual Bitcoin D and Bitcoin CLI.

You can use a block explorer if you want.

I'm not going to say that's cheating,

but it's also not as fun.

So the block explorers websites will

help you find a lot of things, because they

have different databases and different indexes

so you can sort of search through things.

It's more fun if you write it yourself.

But if you try this,

OK, that works too.

Yeah, so hopefully you get familiar with the Bitcoin CLI.

If you want to do your own scripting with Bitcoin,

you can use anything you want.

So Bitcoin CLI is a command-line utility

that will let you interact with the Bitcoin node.

So for example-- is this readable?

Kind of, right?

So if I want to get peer info.

I want to see-- this is a computer down

on the first floor, running Bitcoin Mainnet,

I want to see, OK, who's connected to me?

And then it'll tell me.

It returns JSON, and it tells me all about the people

I'm connected to.

If I want to say, get wallet info, it'll tell me

I don't have any money--

things like that.

And then there's, like, Help.

It's weird and it's not super well documented.

It's fairly well documented in this stuff.

There's also tons of undocumented hidden RPC calls

that are deprecated, or new, or programmers put in

and they don't want people to use them.

They only want to use them for themselves.

So it's kind of a mess, but it's fun

to see exactly how this software works.

So hopefully you can try that out.

And I will add more UTXOs to grab in the next few days.

So that might be fun.

You could, if you wanted to be a jerk, just steal all the money,

and then no one else could do it.

But please don't do that.

I mean, this is not as adversarial a setting

as actual Bitcoin.

If someone does it, I'll be like, OK, fine,

and I'll just like reseed the place.

So hopefully we don't have too many trolls

trying to prevent everyone from having fun

with this assignment.

That said, there is at-- aspect some of them are--

the one I put in the assignment last night,

there's five outputs that you can grab.

One of the sort of hints-- there's only five outputs.

So the first five people to grab it get the coins.

So there is a little bit of timing stuff there.

Any questions?

Cool?

Yeah, so if you have questions about it,

office hours tomorrow, also IRC, stuff like that.

OK, so today's main topic--

transaction fees.

So we described earlier that the transaction fee is just

the difference between the sum of the input amounts and output

amounts.

What's interesting is it's implicit.

So if you actually look at a transaction,

you can't tell what the fees are.

You have to look at its ancestors

to tell what the fees are.

So for example, if you want to go through here and-- oh,

how am I going to do this?

Jeez.

If I want to say, OK, what's the most current block?

getblockchaininfo.

OK, so the most current block is this.

And then I can say, OK, getblock--

can I just say getblock?

Yeah, I want to get that block.

OK, there's all the transactions in this block.

I can do getblock verbose true, and it'll give me

all the transaction data.

But instead, I can just pick one of these transactions

and say getrawtransaction, get that thing.

Oh, here's another weird thing.

So it'll just give you hex, which

is not very useful, unless you somehow can read hexadecimal.

And then in the actual thing, it says, like, examples.

Wait, true?

Did they change it?

You used to have to do 1 at the end.

Yeah.

But true works too?

It used to not work if you said true,

and you had to put the number 1 instead of true.

There is a lot of weird bugs in Bitcoin

that have been there for years, and I

guess they fixed that one.

The most egregious is probably the multisig bug,

where when you're spending from a multisig output,

you have to push a zero-byte onto the stack first,

or else it doesn't work.

It was just a bug Satoshi had.

And now it's consensus critical.

And people joke about it, like, oh, we also

have to put a zero-byte in whatever new

RPC we create, because there's just, like, bugs.

Anyway, so for example, this transaction that I just brought

up, this is how it'll look.

Here's the TXID, it's 370 bytes, it's

got this lock time, which-- ooh, I should mention that.

I'm guessing this transaction was created by Bitcoin Core

Wallet, not a different wallet.

That's a hint.

It's got one input--

no, two inputs.

OK, so inputs are specified-- this TXID input,

this TXID input--

two inputs and one output of value, 0004500.

We can't tell what the fee is from this.

No, sorry, two outputs.

Here are the two outputs.

So we can tell these two outputs, however,

we don't know what the fees are.

We know the total output amounts,

but the input amounts are not encoded

in the transaction itself.

In order to figure that out, we'd

have to go to this transaction and see what vout 0's output

amount is, which we can do.

And it's 0052-- you know, so we can add it up that way,

but it's not written directly, which is a little weird.

And the idea of the fee is, whoever mines

that block that includes that transaction gets the fee.

So the sum of all the inputs and outputs within a block

do add up, modulo the new coins that are created.

But long-term, when there's no coins being created,

when they've all been created, you can say,

OK, we can look at a block, add up

all the inputs spent in all the transactions in the block,

add up all the outputs created in all the transactions

in the block, and those should equal.

Otherwise, it's invalid.

Well, they don't have to equal.

You're always allowed to destroy coins.

So if the miner says, hey, there's

all these fees that can be paid to me,

and I just don't grab them, you can do that.

So it's a less-than-or-equals operator that-- it's the check.

And people have destroyed coins, usually inadvertently.

There was one in January where there

was a block that just had a zero output for the miner reward,

so all whatever, 13, 14 coins were just destroyed.

Yeah, it seemed like a bug.

So that's how fees work.

What's interesting is, you don't know who you're

paying when you create the fee.

You sign sort of, OK, here's my transaction,

I sign it off, whoever confirms this transaction gets $5

or whatever.

And you throw it out there.

So it's kind of interesting.

There could be legal--

I've heard there's legal questions, like, well, what

if one of the miners is in Iran, now

you're paying someone in Iran, are you in trouble for that?

You don't know.

You can't tell who going to mine it beforehand.

And generally, we talk about fee rates, or sort

of a specific fee-per-byte.

And it's expressed in Satoshis per byte, and one Satoshi

being the minimum possible bitcoin

amount, which is 1, 2, 3, 4, 1, 2, 3 zeros, so seven years.

There's eight digits in bitcoin.

And you usually keep that.

So you pad it out.

So even in something where it's 0.0134,

they put the extra four zeros.

In the actual code, this is just a 64-bit signed integer,

and there is no decimal point.

So it's sort of a UI thing to put a decimal point in so

that people can think of, oh, that's two bitcoins,

or that's half a bitcoin.

But really, the code only deals with the Satoshis,

and the decimal point is purely UI layer,

which can be confusing sometimes,

when you're dealing with JSON, and when

you're dealing with different programs,

tossing numbers around.

You can get a factor of 100 million off, sometimes.

It's a big enough factor that, usually, it's pretty obvious.

So from a miner's point of view, they

want to prioritize, based on the TX size and the fee rate