Kasper König

In this post we follow up from my previous one. Check that out first. Until now, we have a node which mines blocks, and allows us to send data to be included in said blocks. It’s all nice and well, but something is missing to make it useful. Since blocks are guaranteed by the difficulty of producing a valid chain of the same length as the one we already have, our chain is as secure as the total computation power we’ve used to produce it. That is not very secure if the entire chain is generated on a single computer. Many computers could generate a longer chain in a short time. We need to start mining on multiple computers.

For this we need to introduce an important concept in crypto-economics: consensus mechanisms. A consensus mechanism is a way in which multiple miners can come to an agreement about what the current chain looks like.

For instance, if there are two miners in the network, and they both mine for 20 minutes without talking to each other, upon synchronizing, they need to agree on which chain is the right one. In this sense, only one chain is ever really valid, but some miners may not know about it yet.

Bitcoin uses a fairly simple consensus mechanism - the longest (most difficult) chain is always right.

If two miners talk to each other, and one sees that the other has a block more than itself, it will immediately copy the state of the other miner, and start mining from the new last block.

To do this in our program, we need a few things. First, we need a way for miners to find each other and communicate.

Node discovery

In bitcoin, there are several layers of service discovery fallbacks and mechanisms, and a lot of concerns in which notes to pick in order to avoid certain types of attacks and censorship.

We’ll go a bit simpler, and use basically the last fallback that bitcoin has: some locally hardcoded addresses. To do so, we need a way for two nodes to act differently. We can start out by simple running them on two different ports - we can use a commandline argument through ARGV - let’s default to 3000 if we don’t supply one.

my_port = ARGV.any? ? ARGV.first.to_i : 3000
Kemal.run(my_port)

now we can spin up our server with a port:

 crystal src/crystal-blockchain.cr 1111 

And in a different session spin up another:

 crystal src/crystal-blockchain.cr 1112 

Alas! We have two selfish miners! Let’s make them curious.

Let’s hardcode a list of discoverable node ports, and select one that’s not our own:

PORTS = [1111, 1112]
my_port = ARGV.any? ? ARGV.first.to_i : 3000
other_port = PORTS.find { |p| p != my_port }

We want to occasionally check other nodes to see if their chain is longer than ours. We can start with a new fiber which checks every 10 seconds:

spawn do
  loop do
    puts
    puts "Checking for new blocks"
    begin
      response = HTTP::Client.get "localhost:#{other_port}"
      body = JSON.parse(response.body)
      puts "Other chain is #{body.as_a.size} long"
      puts "My chain is #{blockchain.size} long"
    rescue
      puts "Can't connect to node with port #{other_port}"
    end
    sleep 10
  end
end

In order to better track the output, I deleted the line that prints every mining iteration:

 puts "Mining! Not solution: #{block.render}" 

Let the two nodes run for a little while, and we can observe in the output that the nodes are indeed finding eachother, and comparing chain sizes. When we run this for a while, a node is sure to get ahead.

Synchronizing chains

The next step is to import the chain from the other block when it gets longer than ours. We ideally want to do this as soon as a block is mined by someone else. If we don’t, we are working on a shorter chain, and while we might get on par by finding a solution - the other node is already mining on the new block, and might just as well find a solution as us - meaning they would keep having the longest chain. To make sure we don’t end up with any invalid blocks and therefore wasted energy mining, we import the other nodes chain when it gets longer, and mine at the end of it.

Of course, in the real world only the blocks which are newer than the ones we already have would need to be imported, but it’s a bit simpler to just import the chain for now.

puts "My chain is #{blockchain.size} long"
if body.as_a.size > blockchain.size
  puts "Importing..."
  new_blockchain = [] of Block
  body.as_a.each do |json_block|
    data = json_block["data"].as_a.map(&.as_s)
    new_blockchain << Block.new(
      json_block["index"].as_i,
      json_block["timestamp"].as_s,
      data,
      json_block["prev_hash"].as_s
    )
  end
  blockchain = new_blockchain

As soon as the other node has a longer node, we create a fresh array of blocks, and parse each element of the json endpoint into a fresh block. In the end we have a copy of the remote chain, which we overwrite our own with.

This is great, but it puts us at risk - if a node would start to cheat, we wouldn’t catch it. It could generate a block that is actually not valid, and we’d happily copy the chain anyway. Let’s fix that. First, let’s move the json stuff into a method on the block to clear up our code a bit:

Adding in block.cr:

def self.from_json(json_block : JSON::Any)
  data = json_block["data"].as_a.map(&.as_s)
  new_blockchain << Block.new(
    json_block["index"].as_i,
    json_block["timestamp"].as_s,
    data,
    json_block["prev_hash"].as_s
  )
end

And in crystal-blockchain.cr:

puts "Importing..."
new_blockchain = [] of Block
body.as_a.each do |json_block|
  new_blockchain << Block.from_json(json_block)
end
blockchain = new_blockchain
puts "Imported!"

This gives us space to validate as we go:

new_blockchain << Block.from_json(json_block)
unless new_blockchain.size == 1 || new_blockchain.last.valid?(new_blockchain.last(2).first)
  puts "cheats!"
  puts new_blockchain.last.render
  puts new_blockchain.last(2).first.render
  raise "Other node is cheating!"
end

However, as it turns out we’re not importing nonces, meaning we’re getting unsolved blocks. Whops. Let’s add it as an initialization argument:

def initialize(@index : Int32, @timestamp : String, @data : Array(String), @prev_hash : String, @nonce : Int32 = 0)
  @hash = calculate_hash
end

And then add it to our .from_json method:

def self.from_json(json_block : JSON::Any)
  data = json_block["data"].as_a.map(&.as_s)
  Block.new(
    json_block["index"].as_i,
    json_block["timestamp"].as_s,
    data,
    json_block["prev_hash"].as_s,
    json_block["nonce"].as_i
  )
end

Now when you spin two nodes up, they will start synchronizing whenever one gets ahead. We know that only blockchains that are valid are getting imported, and herein lies the core of a blockchain: It’s quite fast to validate that it took as much computation power to generate it as expected. What we have now is a system that is very hard to overwrite. If two people runs these two notes for a while, one could not easily change the state. The bad actor would have to use more computing power than the other node, and the longer the system runs, the harder to manipulate.

Adding more nodes

We could extend the algorithm to work with a few more nodes if we liked. One more might be good, to illustrate how:

First, let’s add the new port, and select all the foreign ports:

PORTS = [1111, 1112, 1113]
my_port = ARGV.any? ?  ARGV.first.to_i : 3000
other_ports = PORTS.select { |p| p != my_port }

Then we loop around the ports:

other_ports.each do |other_port|
  begin
    response = HTTP::Client.get "localhost:#{other_port}"
    body = JSON.parse(response.body)
    puts "Other chain is #{body.as_a.size} long"
    puts "My chain is #{blockchain.size} long"
    if body.as_a.size > blockchain.size
      puts "Importing..."
      new_blockchain = [] of Block
      body.as_a.each do |json_block|
        new_blockchain << Block.from_json(json_block)
        unless new_blockchain.size == 1 || new_blockchain.last.valid?(new_blockchain.last(2).first)
          puts "cheats!"
          puts new_blockchain.last.render
          puts new_blockchain.last(2).first.render
          raise "Other node is cheating!"
        end
      end
      blockchain = new_blockchain
      puts "Imported!"
    end
  rescue
    puts "Can't connect to node with port #{other_port}"
  end
end

That’s all it takes. We can now spin up 1113 and they all start working together.

This concludes the basics of mining. Next step is to bring a real monetary system into those blocks of data. The current state of the code can be found here. To be continued…