Bitcoin's consensus rules define what is valid, but this isn't helpful when we're looking at changing the rules themselves. The trend in Bitcoin has been to make such changes in an increasingly inclusive and conservative manner, but we are still feeling our way through this, and appreciating more nuance each time we do so.

To use Bitcoin, you need to remain in the supermajority of consensus on what the rules are. But you can never truly know if you are. Everyone can signal, but everyone can lie. You can't know what software other nodes or miners are running: even expensive testing of miners by creating an invalid block only tests one possible difference, may still give a false negative, and doesn't mean they can't change a moment later.

This risk of being left out is heightened greatly when the rules change. This is why we need to rely on multiple mechanisms to reassure ourselves that consensus will be maintained:

  1. Developers assure themselves that the change is technically valid, positive and has broad support. The main tools for this are open communication, and time. Developers signal support by implementing the change.
  2. Users signal their support by upgrading their nodes.
  3. Miners signal their support by actually tagging their blocks.

We need actual consensus, not simply the appearance of consensus. Thus it is vital that all groups know they can express their approval or rejection, in a way they know will be heard by others. In the end, the economic supermajority of Bitcoin users can set the rules, but no other group or subgroup should have inordinate influence, nor should they appear to have such control.

The Goodwill Dividend

A Bitcoin community which has consensus and knows it is not only safest from a technical perspective: the goodwill and confidence gives us all assurance that we can make (or resist!) changes in future.

It will also help us defend against the inevitable attacks and challenges we are going to face, which may be a more important effect than any particular soft-fork feature.