In this Orientation Module you will learn about Decision Making within the project. As with the previous Level 1 Module, if you have prior experience with an open source software project, especially one at Apache, then much of this material will already be familiar to you.
Please send an email to firstname.lastname@example.org to let us know you are working on this module.
In the previous Module we read about collaboration on the mailings lists, how to do it efficiently and how to avoid the most common pitfalls. We use the mailing lists for many things, for asking questions, for sharing information or the like. But one of the most important uses of the mailing list is for decision making. It is important to understand how decisions are made in an Apache community. Here are a few general principles that you should keep in mind:
There are two primary ways of managing product changes, which go by the names Commit-Then-Review (CTR) and Review-Then-Commit (RTC). For most cases we operate in a CTR mode, meaning that our Committers are able to check in changes as they desire, with no advance approval or review.
We trust our Committers to do the right thing. By default Committers don't ask permission before acting. They avoid unnecessary discussion and email traffic. This is not because they are anti-social. This is because they realize that in a project of this size it is impossible to discuss every small change in advance. Discussing too much is both unnecessary and unproductive. We have a "time machine" called Subversion that allows us to undo any changes to the product or website. So if a Committer believes that a change would be uncontroversial, and the change is reversible, then the default approach is to go ahead make the change.
Terms that you might related to the above are: JFDI and "assuming lazy consensus".
However, there are times where CTR is not appropriate and RTC is used. This happens, for example, at the end of a release cycle, when we want to carefully review changes made to the product, in order to control the final quality before the release. When we're in RTC mode, no changes are made to the code unless first discussed and approved on the mailing list.
A Committer can also voluntarily invocate a RTC on their own changes. For example the Committer might be considering a change that of greater significance. Maybe he would
value additional input, or technical review. Maybe the change is controversial or require coordination. If CTR is not appropriate then one can make a proposal and send it to
the dev mailing list.
The convention is to send all proposals in their own new thread. (Don't hide a proposal 10 posts deep in an existing thread). Put "[PROPOSAL]" in the subject line or make it obvious that you are making a proposal.
Because the Volunteers are spread out all across the globe, in various time zones, and many have day jobs or other committments, the convention is to wait at least 72 hours for feedback on a proposal.
In cases where the proposer wants to act on their proposal, if there are no objections, they should state this in the proposal. For example, "If there are no objections voiced within 72 hours, I'll go ahead and make these changes". This is called "stating lazy consensus". You can read more about lazy consensus here.
In Apache projects it is common to use a shorthand way of responding to proposals, where +1 indicates approval, 0 indicates indifference and -1 indicates disapproval.
In most cases proposals are decided by consensus, based on community discussions. Only in rare cases, and in a small number of pre-defined administrative questions, do we resort to a formal counting of votes. The places where we require voting are: voting to release, voting in a new Committer or PMC Member, Voting in a new PMC Chair. That's it. Generally speaking, voting on any other topic is avoided in favor of consensus building. With voting there are winners and losers. With consensus everyone wins.
Another aspect of decision making in an Apache project is the "veto". Every Committer has the ability to "veto" a change, for technical reasons, provided he explains the technical reasons for the veto, describes an alternative approach, and offers to help implement the alternative approach. Vetos are quite rare.
There is one disorder of community decision making that is common enough to warrant a colorful name: bikeshedding. Follow the link and read more about this topic.
To apply the above skills, be sure you are subscribed to the project's main mailing list. Keep your eye out for terms like "proposal", "lazy consensus", "vote" or "veto" and see how they are used in practice. Compare actual practice to the above descriptions. No, we're not perfect. But we work best and have the most fun when we follow the above guidelines. And so will you when you apply then in your own list communications!
Congratulations! You have completed this Module Please send a note to email@example.com so we all know. This is also a good opportunity to send along any feedback or questions you might have on this Orientation Module.