@bpl @jaypatelani @corvus @cmccullough
Well, that's kind of damning: https://github.com/NetBSD/src/commit/f3b327fa6f8b4ff6f758a8b9e8190aa20109e222
@netbsd ? 馃ズ
Re: tmux in OpenBSD, false positives are possible, because the tool is just looking at filenames and metadata, not actual file contents. If they're copying source code from a slop project, but not the metadata/commits, then it will give you a false negative.
Not as damning as one might think.
That's a commit to external/ where Christos Zoulas is importing into NetBSD's mirror all of the changes made to the original ISC BIND in version 9.20.24, en bloc. M. Zoulas didn't use the LLM, and the LLM-generated code is only one of a whole slew of changes made to the BIND source tree in that version.
The person who uses the LLMs is Michal Nowak who works for ISC in QA.
https://gitlab.isc.org/isc-projects/bind9/-/commit/e13302a6bc9b196564b5e7afe703fae24311ceeb
#ISC has an explicitly pro-LLM stance on #BIND, and I wouldn't be surprised if #Kea and DHCP follow suit.
https://gitlab.isc.org/isc-projects/bind9/-/merge_requests/12363
https://gitlab.isc.org/isc-projects/bind9/-/blob/main/CONTRIBUTING.md#contributing-code
There have been good reasons not to use ISC BIND for over a quarter of a century (monolithic software, runs with superuser privileges, multiple hats, lots of parsing). Alas, it is an expenditure of time and effort on the #NetBSD people's parts to switch to a non-LLM-generated DNS server software. But it is well past due.