The three most important conclusions from the ARB sessions at UDS this week are:
- With over 200 apps in the ARB queue (and growing), and only 3 ARB members, we have hit the point where it's impossible for the ARB alone to manage the queue. This isn't really a surprise, just "ah, yes, we've reached that bridge we saw coming, time to cross it". - It's time to start making the move to more automated publishing. Fortunately, we've had commitments from the Foundations and Security teams to invest resources this cycle in the tools to make automatic sandboxing and package checks possible. (And, thanks to David Planella and Michael Hall for all the work they've done getting those resources allocated to us.) Automated tools won't cover all cases, and some packages will still need to be manually reviewed, but automating as much as possible will put the application process in a much, much better position. - But, developing those tools is a process, so we can't just switch overnight. In the meantime we need to figure out how to scale manual reviews to match the rising demand. A lot of the changes we discussed are minor process refinements, more efficient workflows or tools that are available now to help out with small chunks. One is a policy change, and the ARB would appreciate input from the TB on whether this makes sense. Currently, apps are approved by three +1s from ARB members. The ARB would like to change this to a model that more closely matches the Ubuntu archives, and require two reviews: - One technical review, made by any Ubuntu Developer. This covers packaging, code review, security review, and any other technical details that are the same between Extras packages and main/universe packages.) - One compliance review, made by an ARB member. This may cover technical aspects of the package, but more importantly covers the details that are unique to Extras packages. This is similar in many ways to an archive-admin role. Opening up app approval reviews to all Ubuntu Developers increases the potential capacity for app reviews. Approval reviews can be a lightweight "I have a couple hours to help this week", instead of an intense long-term commitment. Widening the pool will also help with scaling for bursty demand, like the App Showdown contest this summer, where the ARB got ~150 app submissions in ~3 weeks. When announcing this change, we'd also make it publicly known that we're equally happy if Ubuntu Developers want to contribute by adopting apps from the ARB queue for universe+backports or debian-mentors. (Using their judgment picking out the apps that are substantial enough to make sense in universe or Debian.) The idea here is that a) it's most important to get the best apps in *somewhere*, b) the app developers mostly only care about "getting into Ubuntu", and don't care about (or often even know about) the significance of different archives, and c) many Ubuntu or Debian Developers would be more motivated to help get apps into debian-mentors/universe+backports, than into extras. Does this policy change around manual reviews seem sensible? Any thoughts or concerns? Thanks, Allison -- technical-board mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/technical-board
