A lot of the boilerplate is largely for new contributors. Leaving or removing the boilerplate isn't going to make a difference to the regular contributors but could add some friction to new ones. That being said, it's pretty annoying as a code reviewer to get a pull request with a title and no other context. I'm not asking for the overwritten, fully formatted LLM PR summaries, but a quick explanation of what intent of the PR is would be helpful. It would be great if Beam contributors as a whole could be more consistent in that effort... I worry that ditching the template will further encourage the pattern of PRs only coming with titles.
So my opinion is "leave it and try to be more active as a community in writing PR descriptions" without enforcing it via infrastructure. On Tue, Sep 29, 2026 at 2:25 PM Danny McCormick via dev <[email protected]> wrote: > I find the template useless for my own work, and I don't think it helps me > review changes from committers. But I do wonder if it is helpful for new > contributors (IMO this is who the feature is for). > > My (unsubstantiated) opinion is that: > > > 1. It is useless for most people, but also harmless > 2. It probably sometimes helps new contributors > > > Given that, I'd vote to leave it, but I don't care very much. > > Thanks, > Danny > > On Tue, Sep 29, 2026 at 2:06 PM Kenneth Knowles <[email protected]> wrote: > >> Hi all, >> >> For discussion: https://github.com/apache/beam/pull/40339 >> >> I just reviewed yet another pull request that has the description >> "PLEASE add a meaningful description here" from the template, just left >> there. >> >> >> - Since 2026-01-01 it looks like 370+ PRs have left the template >> without modifying it [unmodified]. >> - There are a total of about 1800 non-dependabot PRs in the same >> time. [total] >> - In fact, of the 1430 others, quite a few are just blank or just say >> "Fixes #<issue>" so I would round up to 25% "blank" descriptions. >> >> So my thoughts on this: >> >> - A rate of 75% of people actually describing their change is >> probably satisfied by the null hypothesis: this many people would do it >> without any template begging them to do it) >> - A fully blank PR field might even be *more* likely to be edited. >> - The other stuff in the template is really mostly noise, or should >> be automation. (like choosing a reviewer or the CI status badges) >> - A lot of PRs now have their description written by an LLM, or taken >> directly from the commit text (also written by LLM) so we don't need the >> template. >> - We could have an LLM check if they actually added a description and >> fail the check, (with a shortcut for totally unmodified description, >> or tiny edit distance, that uses zero tokens) >> >> WDYT? >> >> I'm just in a "cleaning up extra / old stuff" mood lately. >> >> Kenn >> >> [unmodified] >> https://github.com/apache/beam/pulls?q=is%3Apr+state%3Aclosed++created%3A%3E%3D2026-01-01+%22add+a+meaningful+description%22 >> >> [total] >> https://github.com/apache/beam/pulls?q=is%3Apr+state%3Aclosed++created%3A%3E%3D2026-01-01+-author%3Aapp%2Fdependabot >> >
