A release candidate is not news, and a forum thread is not a decision
Serguey Asael Shinder2 min readThis desk is fed mostly by the projects themselves: improvement-proposal trees, client repositories, foundation blogs, governance forums. That is deliberate — a merged proposal or a tagged release is the event, and trade coverage is a pointer to it. It also produces a specific problem, which was visible on this site's first day: the feeds deliver a great deal of motion and comparatively little event.
Four tests separate them, and they are cheap enough to apply every time.
A release candidate is not a release. On the day this desk opened, the newest items on two chains were v1.15.0-rc.3 and v1.15.0-rc.4, published hours apart. An rc is an invitation to test. Writing it up as news produces a story that is obsolete when rc.5 lands, and it teaches a reader that our headlines do not mean anything. The desk stayed empty instead, which is the honest outcome.
A forum thread is not a decision. Governance forums are where proposals are argued about, and the argument is often more interesting than the outcome — but a post is a position taken by one participant. The decision lives in the proposal document and in the client release that implements it. Where we quote a forum, the piece says it is a forum.
A status field beats a headline. Every proposal system carries one: Draft, Review, Accepted, Final. It is the single most reliable thing in the document and the thing coverage most often drops. A change described as arriving, whose proposal reads Draft with an empty feature key, is a plan — and the piece should say so in the sentence where the plan is described, not in a caveat at the end.
A commit is an event only if it changes what an implementer must do. Repository feeds deliver every merge. Most are editorial. The ones worth a note are those that change a rule: a scalar that must now be reduced, a limit raised from one number to another, a proposal moved from one list to another. The test is whether somebody has to change code or a plan because of it.
None of this is a counsel of silence, and it is not a licence to leave a desk empty either. Applied to a single day's queue, these tests rejected a dozen release candidates, a daily social digest and several threads. What follows a rejection is not a blank page: it is going to the project itself. Two desks had nothing in the queue and a real, dated release on the project's own release page — which is where they were filled from. The rule this desk works to is both halves at once: every desk gets something, and nothing gets filled with rc.4.