How does Catapult actually determine a unique network?

Mark_Price · June 6, 2018, 4:37pm

Does Catapult use a unique identifier for a network or is there some other unique manner of identifying a network?

As far as I can tell NetworkIdentifier is used to do this.

namespace {
	const std::array<std::pair<const char*, NetworkIdentifier>, 4> String_To_Network_Identifier_Pairs{{
		{ "mijin", NetworkIdentifier::Mijin },
		{ "mijin-test", NetworkIdentifier::Mijin_Test },
		{ "public", NetworkIdentifier::Public },
		{ "public-test", NetworkIdentifier::Public_Test }
	}};
}

Which seems to map to this:

#define NETWORK_IDENTIFIER_LIST \
/* A default (zero) identifier that does not identify any known network. */ \
ENUM_VALUE(Zero, 0) \
\
/* Mijin network identifier. */ \
ENUM_VALUE(Mijin, 0x60) \
\
/* Mijin test network identifier. */ \
ENUM_VALUE(Mijin_Test, 0x90) \
\
/* Public main network identifier. */ \
ENUM_VALUE(Public, 0x68) \
\
/* Public test network identifier. */ \
ENUM_VALUE(Public_Test, 0x98)

But I feel like this is not right. On Tendermint for instance, it generates a unique network hash, then you set at least one root peer - then all new nodes can find each other using that hash and listed peer(s).

Also I’m aware that there is a generationHash but it seems to be dependent on that NetworkIdentifier - if this is used, can we guarantee the unique network using network ID mijin-test?

/// Creates a network info around a network \a identifier, a nemesis public key (\a publicKey)
	/// and a nemesis generation hash (\a generationHash).
	constexpr NetworkInfo(NetworkIdentifier identifier, const Key& publicKey, const Hash256& generationHash)
			: Identifier(identifier)
			, PublicKey(publicKey)
			, GenerationHash(generationHash)
	{}

What is the process for Catapult?

gevs · June 6, 2018, 4:47pm

As you can see in NEM’s first version, the network identifier is also used in the address generation process :

github.com

evias/nem-php/blob/master/src/Models/Address.php#L119


}


// step 1: keccak-256 hash of the public key
$pubKeyHash = Encryption::hash("keccak-256", $pubKeyBuf->getBinary(), true); // raw=true


// step 2: ripemd160 hash of (1)
$pubKeyRiped = new Buffer(hash("ripemd160", $pubKeyHash, true), 20);


// step 3: add version byte in front of (2)
$networkPrefix = Network::getPrefixFromId($networkId);
$versionPrefixedPubKey = Buffer::fromHex($networkPrefix . $pubKeyRiped->getHex());


// step 4: get the checksum of (3)
$checksum = Encryption::checksum("keccak-256", $versionPrefixedPubKey, 4); // checksumLen=4


// step 5: concatenate (3) and (4)
$addressHash = $versionPrefixedPubKey->getHex() . $checksum->getHex();
$hashBuf = Buffer::fromHex($addressHash);
$encodedAddress = hex2bin($addressHash);


// step 6: base32 encode (5)

For Mijin Test, it is the network ID 90 which will result in S (hex ‘90’) starting addresses. Unsure if this answers your question.

Mark_Price · June 6, 2018, 5:04pm

So perhaps the prefix is used in the peer discovery then.

So I can probably safely conclude that when people want to launch their own networks they should change the string identifier and the prefix, compile the code, then deploy.

gevs · June 6, 2018, 5:24pm

correct, when you call /node/extended-info you can see the networkId in there.

I am unsure if it is really so important to change the networkId. When you are running your own network, you also provide with a peers list (or at least the tip of it, as you mentioned) and those peers would simply be configured with your networkId.

Network Name doesn’t matter the slightest I think, but the identifier yes. If you wish a more custom experience (addresses starting with X,… whatever) you would always change the identifier in the software yep.