23 - Enhanced Hard Fork Mechanism#

  DIP: 0023
  Title: Enhanced Hard Fork Mechanism
  Author(s): Pasta
  Comments-Summary: No comments yet.
  Status: Proposed
  Type: Standard
  Created: 2021-07-29
  License: MIT License


This DIP aims to introduce an improved hard fork activation mechanism that includes both masternodes and miners while also ensuring sufficient upgrade time is available to partners.


Currently, we utilize a heavily modified version of BIP-9 for hard fork activation in Dash Core. In the previous version of Dash Core, this calculation took into account the percentage of both miners and masternodes that had upgraded. However, since the introduction of the Deterministic Masternode List, the activations have purely been based on miner signalling.

Previous Work#

Current System#

The current system declares a start timestamp. The first block of height % 4032 == 0 with a timestamp greater than the start timestamp begins the first signalling cycle. Each signalling cycle has a threshold that must be met in order for the hard fork to lock-in. In our modified BIP-9 style activation, this threshold decreases at an exponential rate (slow at first then accelerating) from 80% down to 60% after 10 cycles have passed. In any cycle, if the final percentage of blocks signalling is greater than the threshold, the activation is locked in. Then after another cycle completes, the hard fork activates.

This system has a number of problems. First, it is actually problematic if the miners upgrade faster than the masternodes. This could lead to a situation where the miners fork, yet masternodes create ChainLocks on the “legacy” fork. In practice this has never been an issue, but it could be. Second, it is possible that this system results in very quick upgrades and forks (as quick as only 2 weeks) which are disruptive for services utilizing Dash. In practice this has resulted in the developers setting the start timestamp to a month or two after the release.

New System#

Stage 0: Miners signal without any effect on consensus. It is impossible to lock-in at this point. The purpose of this signalling is to enable monitoring of which miners have upgraded their software and get an idea of the adoption rate. Additionally, it is important to note that this signalling does not take up any space, and as such there is no incentive to minimize the time where signalling is done.

Stage 1: Any valid LLMQ_400_85 produces a mnhfsignal message to signal that the masternode network has been upgraded. Since this is done by the LLMQ_400_85 quorum, it will in effect only occur once 85% of the masternode network has upgraded to this version. Once this message has been created, it should be included in a block as a zero fee special transaction, similar to quorum commitments. This can be done by any miner in any block, but it should only be included once. For testnet the LLMQ_50_60 quorum should be used instead.

The mnhfsignal message should look like:








The version bits associated with the hard fork




Hash of the quorum signing this message




BLS signature on version by a public key associated with the quorum referenced by quorumHash

The payload of the Masternode Hard Fork Signalling Transaction (the new special transaction with type=7) should have the following format:








Commitment special transaction version number. Currently set to 1. Please note that this is not the same as the version field of the mnhfsignal message




This equals the payload of the mnhfsignal P2P message

Stage 2: Once this mnhfsignal message has been mined, two parameters are set: masternode_activation_height and min_activation_height. The cycle beginning at masternode_activation_height is the first cycle where miner signalling matters. Its starting height is calculated based on the mnhfsignal message height as shown here:

masternode_activation_height = mined_height - (mined_height % 4032) + 4032

This cycle’s threshold is nThresholdStart (the maximum threshold value). Each subsequent cycle will see its threshold decreased until it reaches nThresholdMin (the minimum threshold value).

Regardless of miner signalling, activation only occurs once the height defined by the min_activation_height parameter is met. This is set to be 6 signalling windows (~6 weeks) after the masternode_activation_height. It is calculated as shown here:

min_activation_height = masternode_activation_height + 4032 * 6

If the threshold is met at the end of a cycle, activation_height is calculated by each node as shown here:

activation_height = max((height + 4032), min_activation_height)


A certain timeout block height will be specified at the implementation of the hard fork to indicate the height at which the deployment is considered failed. If a mnhfsignal message is not included in a block with a height less than the timeout block height, the hard fork is deemed to have timed out and cannot be activated.

Similar to BIP-9, a later deployment using the same bit is possible as long as its timeout block height is higher than the previous one’s timeout or activation height. This is discouraged until necessary, and even then it is recommended to have a pause in between to detect buggy software.