2017-03-16 21:47 GMT+01:00 Niall Douglas via Boost
This implies that the problem is a lack of review managers. I see my other email wasn't replied to. Perhaps because it is easier to talk about review managers. So let me sum up my other response to see if we can get some discussion here:
I did intend to reply to your earlier email to say you made a lot of good points I hadn't considered. Unfortunately sick children intervened, I was stuck in Hospital A&E until 4am last night due to a fever we couldn't break (all ended up well, fever broke this morning).
In your earlier email you made the extremely valuable point that one third of the queue is in the process of being cleared, which I hadn't considered. Let's assume my estimate that 25% of the queue would never pass a review due to a glaring deficiency. That means that the "real queue problem" is just 9 libraries. I also am pretty sure at least two of those libraries the author has given up on getting a review and so are stale, so it could be that the true backlog is just 7 libraries, or about 30% . That's manageable.
It's a really a problem of *optics*. It *looks bad* to the wider public if we have a really big review queue well exceeding *15%* of all libraries already in Boost. As Edward says, it puts people off submitting, creates morale problems etc. Part of the solution I think needs the queue to be constructed and presented differently, and that is what the rest of this email is about.
* Not having a review manager might be an indicator of not enough interest in a library. It is the job of the author to ensure there is enough interest by the community. Perhaps the author hasn't done enough promotion. Maybe more solicitation on the ML is required or perhaps people just don't find the solution interesting. One person saying, "that sounds like a neat library!" shouldn't constitute interest.
That last sentence stood out for me i.e. "One person saying, "that sounds like a neat library!" shouldn't constitute interest."
I began thinking that that could help a lot with preventing unready and uninteresting libraries entering the review queue in the first place.
Here is my proposal:
1. All libraries in the review queue without managers attached are removed (including my own!) and the authors emailed to say the following new policy applies. The review queue is therefore emptied.
2. For a library to enter the review queue in future, it requires at least one (and preferably more) named members of the Boost community to publicly endorse the library to enter the review queue. Their names will be listed alongside the library in the review queue page at http://www.boost.org/community/review_schedule.html.
3. Endorsing a library has NO RELATION to review managing a library. Indeed if only one person endorses a library for review, they are not permitted to act as review manager.
4. To find someone to endorse a new library for review, the library author ought to ideally canvas for a library's motivation before they ever begin writing or designing it, but failing that they need to approach boost-dev and publicise their library seeking someone to publicly endorse it for review. Other forums work too e.g. reddit/r/cpp, the Incubator or anywhere else. Ideally I'd prefer if the Incubator *was* the place where people endorsed a library for review and their name automatically was added to the review queue page, but I appreciate that's a lot of scripting.
I am personally highly unsure of Robert's suggestion (he claims it was mine, it was not) that every author of a library entering the queue needs to review manage a library first. Speaking for myself as someone who has managed a review three times now with a fourth time starting tomorrow, I would be highly unsuitable to make a recommendation on a domain far away from anything I've ever used, I'd be likely to recommend the wrong thing through ignorance.
The above proposed policy effectively pushes the bottleneck higher up the chain, but I think that's no bad thing. Library authors, myself included, like to build cathedrals irrespective of whether anyone will ever use them nor appreciate them. Currently it's too easy to build a library nobody will ever use and get it into the review queue where it will languish for many years because no review manager will touch it. That part needs to change.
I like the idea. Of course, regarding the endorsements, you now have to define who qualifies as "Boost member". Is it anyone who signed up for boost-dev mailing list? Another, similar suggestion. When we were planing for review with Robert, we were already aware of two people having informally committed to submitting a review. I liked the idea a lot. Maybe it can be formalized. One of the criteria for review-readiness could be to have at least N (where N = 2 or similar) people who declare that they would submit a review. This declaration is not binding. This prevents the situations where a review ends in the rejection due to lack of reviewers. I am not sure if it is the same as endorsement. Regards, &rzej;