On 6/16/26 18:51, Neil C Smith wrote: > Hi, > > On Fri, 12 Jun 2026 at 23:19, Michael Bien <[email protected]> wrote: >> Although 2) would be tempting, 1) aligns more with the ASF recommendations. >> I do also think that it can be useful during review to know if LLMs helped >> to produce a commit. (Its sometimes easy to spot, but it would remove having >> to guess). > There's also the tempting 2b) - remove the advertisement fields by > disallowing their usage entirely. Well, there hasn't been enough > discussion in this thread yet! :-)
IMO: whether a change comes from an IDE/refactoring tool, a LLM, or a specially trained cat doesn't matter much to me as long a human understood the change, checked it for correctness and is happy about it. During review (or when problems show up) however, it might be useful to know if some of it was generated by a refactoring script, a cat or written by hand. That is also why I like the "Assisted-by" tag as used by the linux kernel, since I think it is more descriptive than "Generated-by" and wouldn't overload "Co-authored-by:" further etc. Regarding forbidding LLM use for code generation entirely: Probably not necessary I think. But no strong opinions there. btw I just saw that the graalvm project uses a ruleset which was also inspired by the linux kernel https://github.com/oracle/graal/commit/2cdf99e580aa433675c191aae70ae515b27aabfe > I think doing something like 1), with "Assisted-by" in both the commit > message and PR description, might be desirable. Yeah it would be good to mention it there too. also: I see very few reasons why someone would need generate the PR text, given that it is read by humans you likely interacted with before. 1-3 sentences or bullet points in your own words should carry more value than a page which reads like a product commercial ;). Similar for contributions where using your own words would at least indicate that the change was understood. There is also the risk that something is submitted which wasn't read at all. Given that anyone can generate summaries after the fact the intrinsic value is likely also low. > It's interesting the > ASF guidelines mention contributors but not committers. Certainly > committers need to be aware of the ASF guidelines when producing their > own contributions. But committers merging third-party pull requests, > particularly from people without an ICLA, may need to know this > information to be clear about the provenance, meet the guidelines and > fulfill their own ICLA requirements? yes that is what I wondered too (first paragraph of the original mail). > I must remember that footnotes should be zero-indexed, but options > should not! :-P sorry about that - will use letters next time. errata: Square brackets were for links, round closing bracket for the points. The first paragraph mentions "point 3", this refers to the point *inside* ASF LLM guidelines. -mbien --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected] For further information about the NetBeans mailing lists, visit: https://cwiki.apache.org/confluence/display/NETBEANS/Mailing+lists
