On 04/07/2014 09:30 AM, Rich Freeman wrote: > On Mon, Apr 7, 2014 at 8:36 AM, Tom Wijsman wrote: >> 2. No, "In the case of disagreement among QA members the majority of >> established QA members must agree with the action. [...]". > > My question is whether they need to seek approval, and when. I can > read the GLEP just fine. I'm not sure what you mean by, "No." Are > you saying they don't need to seek approval before taking action? > Honestly, this kind of ambiguity is exactly what I want us to avoid > when we get into the operations of QA. It is better to say, "no, > pre-approval of individual QA actions is not needed because..." or > "no, pre-approval of individual actions IS needed because..." Short > answers to complex questions lead to interpretation. > > I think it is an important thing to clarify. Are you suggesting a > workflow where any individual in QA can take an action they feel is > necessary, and then the rest of QA only comes into it if somebody > notices and disagrees, at which point the action gets undone? Or is > the workflow that when somebody wants to take an action they first run > it past the team? > >> >> 3. A mail to qa@gentoo.org and/or an agenda item, wait until a vote >> or lead decision has been made; we already do things like this. > > Well, clearly nobody waited for a vote on this one. At least, I can > find no record of a vote approving the mask on the new virtuals. If > this is only done after a QA action is taken if there is disagreement, > then we should clarify what happens in the meantime (does the mask/etc > stay in place for a few weeks?). I think the importance of clarifying > when individuals should act on their own becomes more important if any > team discussion only comes after the fact. If team discussion happens > before actions are taken then there is less risk of inconsistent > actions, but of course more latency before action can be taken. > Most of this already has policy. Quoting from our notification rules, agreed upon in one of our earliest meetings: "We fix and send a friendly reminder to the maintainer(s) for trivial fixes; open bug, wait 2 weeks, make the change for larger but non-critical fixes; make the change and send a notification for critical fixes." This does not actually cover the QA team taking non-emergency action without filing a bug, nor does it specify whether these decisions need to be discussed among the team. Those will need to be sorted out at a future QA team meeting, and we will also be discussing how to deal with internal disagreements and writing down our procedures more explicitly. However, I think that it is reasonably to infer from this existing policy that unilateral action should only be taken in an emergency (and by emergency I mean something like an ebuild potentially breaking systems). This was not an emergency, and from my understanding it could have waited a bit longer for team discussion. Zero_Chaos's actions were taken too hastily, and he and I will be discussing this. I was not there at the time, so I do not have all of the facts, but this is what it looks like from a cold read of the situation. I know that the virtuals were introduced without discussion, which was not good, but the quick action was not the right action, and I am not entirely convinced that this was a QA problem. This was a people issue (not notifying about a new thing), not a technical issue, and so I don't think we should have been involved in the first place. ssuominen's revert of a mask was also inappropriate. I also am disappointed by the constant bickering and sniping going on both in IRC and on-list. I think all of us should be polite and professional, even when disagreeing with somebody. Pretty much all parties involved in this argument have been disrespectful to each other, and I am not okay with that. Finally, the QA team will be discussing what we should and should not be tackling, and when it is okay to call an action a QA action. Like I said earlier, I'm not sure that this was something we should have been involved in, and so we will need to think that part through. We will also be covering potential consequences for willy-nilly flashing of the QA "badge" (a problem several people on this thread have pointed out). tl;dr: introducing the new virtuals without discussion: bad. Unilateral masking in a non-emergency: bad. Reverting QA stuff: bad. QA action workflow: will be discussed, clarified, and written down. People bickering uselessly: bad. What qualifies as a QA action and when it's okay to call it such, and consequences of doing so improperly: to be discussed. Chris Reffett Gentoo QA Lead