Bitcoin P2P network - Hard Fork

[?QuestionMark?]

As far as I know, when a hard fork happens some nodes run another version (for example higher block size) and original nodes won't accept such blocks. Is the P2P network splitting up by itself, by for example the original nodes start not accepting connections anymore from a peer after some time which sends invalid blocks (new version node), or does a hard fork project set up a whole new p2p network with domain name servers etc?. One of them is dnsseed.bluematt.me, and some are hardcoded. Is a hard fork like bitcoin cash running on the same name servers?

[BrewMaster]

things like bitcoin cash have nothing to do with bitcoin, there are hundreds of them that were created by creating a copy of bitcoin and creating a similar duplicate of everything. for example the DNS seeds which are ONLY used when the nodes run for the very first time are created and hosted by those altcoin creators and their network is completely separate from bitcoin's network.

[NotATether]

A hard fork is when someone modifies a bitcoin client to alter or remove consensus rules that are in the bitcoin protocol.

So if someone does that, they are running a version of the bitcoin protocol that is incompatible with what everyone else is running.

What this means is that depending what you've changed, some messages/blocks/transactions/etc that you send to nodes running the original protocol will be rejected by them, because according to their (original) rules, those messages contain unrecognized or invalid content. That is what is called a hard fork.


Basically if we have a bunch of nodes that together form "the network" such as:

    B
A     D
    C     E

And all of these nodes are running an authentic Bitcoin Core, but some time later, the guy who runs node E decides to change some lines about consensus rules in the code and compiled and runs it, then he has just made a hard fork of bitcoin. This is how the network will react so such a hard fork:

    B     ||
A     D  ||
    C     || E

In other words, node E isolates itself from the bitcoin network. If more nodes start running E's code then they become isolated too but they form a group with E.

None of the nodes A,B,C,D are actually blocking E from connecting, but they are forced to reject some of the messages it sends that violates their own consensus rules, and accept the messages that do follow them.

    || B     ||
A  ||    D  ||
    || C     || E

This is node A running the original Bitcoin Core, B,C and D running some forked Core, and E running yet another forked Core. All messages that pass through the double bars may or may not be considered invalid by the nodes on the other side.

[ranochigo]

for example the DNS seeds which are ONLY used when the nodes run for the very first time are created and hosted by those altcoin creators and their network is completely separate from bitcoin's network.

The node will always query DNS seeds when they are unable to get any peers. A common scenario is if the peers.dat is not backwards compatible or if the user deletes the peers.dat and anchor.dat completely.
Depending on the difference in protocol rules, the transactions that is made on Bitcoin could very well also be valid on a fork of Bitcoin and thus exposing the user to the risk of replay attacks and that is also one of the main concerns when dealing with Bitcoin forks.