Open collaboration groups tend to strive for consensus in their decisions. Consensus as an ideal is a worthwhile goal because it allows for all participants to have a direct influence on the decisions and therefore fosters participation and values the participants. However seeking consensus can also lead to a lack of progress because individual voices have the ability to block all progress.

When we created CAcert, a community CA, we realized that this was not a sustainable model. So we decided to borrow the concept of “rough consensus” from the IETF and adapt it to our needs. This has served CAcert well and has been the single best model for collaborative group management I have yet encountered.

To quote the CAcert Wiki:

Rough consensus is somewhere between a unanimous acceptance, and a clear super-majority. A simple majority vote of 51% is not sufficient, and a 100% vote is not required.

In practice this means that all criticisms must be addressed sufficiently to be satisfied that all have had a fair say. But it also recognises that there will always be some who disagree. Perfect is the enemy of the good.

This method allows for votes as an indication mechanism to determine the level of agreement that has been reached without the presumtion that a majority of any type will be sufficient to pass the decision. Even a single voice may suffice to indicate a lack of rough consensus. Which leads to the question of how a group can determine that rough consensus has been reached.

The interesting answer is that this can be decided using rough consensus. The group can take indicative votes at any time both on the matter at hand itself or on whether rough consensus has been reached. Presuming honest dealing among participants, the vote indication on the matter could be significantly different from the vote indication whether rough consensus has been reached.

As a practical matter if the answer to the question what would need to change about the proposal to reach unanimous consensus is:

  1. nothing
  2. any changes made would not increase the level of consent

The first would be an indication that there are simply some whose objections are systemic rather than matter specific; this means there is rough consensus and the decision should be passed.

The second case could be an indication that there are simply factual incompatibilities leading to two mutually exclusive choices that can never be reconciled. In this case an indicative vote can be taken whether rough consensus has been achieved or whether an entirely new approach needs to be discovered and proposed.

In short this methodology allows for effective consensus driven decision making without the detrimental elements of requiring unanimity. While this model may not fit every possible situation, it served very well so far and can only be recommended as a general modus operandi.