| opendevreview | OpenStack Proposal Bot proposed openstack/openstack-manuals master: Imported Translations from Zanata https://review.opendev.org/c/openstack/openstack-manuals/+/1008640 | 03:36 |
|---|---|---|
| opendevreview | Grzegorz Grasza proposed openstack/security-doc master: Add OSSN-0111: Keystone MFA enforcement gaps and replay weaknesses https://review.opendev.org/c/openstack/security-doc/+/1006742 | 12:20 |
| frickler | tmazur: maybe you want to look at https://docs.openstack.org/kolla/latest/contributor/contributing.html#ai-generated-content for some inspiration in dealing with overmotivated AI users | 17:28 |
| tmazur | frickler: Yes, thank you! That's exactly what I was looking for. We will prioritize adding the similar guidelines to our contributor's guide. | 17:32 |
| fungi | looks basically like a copy of https://opendev.org/zuul/zuul#ai-generated-content | 17:49 |
| fungi | ah, the review which introduced it said as much | 18:24 |
| dansmith | fungi: sucks when you thought you caught someone red-handed, only to find out they've already confessed :) | 19:22 |
| fungi | touché | 19:24 |
| fungi | (not that i was accusing anyone of anything, merely observing a striking similarity that turned out to be explained elsewhere) | 19:25 |
| dansmith | I know of course :) | 19:25 |
| dansmith | also, considering the license(s) it'd be hard to accuse someone of stealing something that has "free, take one" printed right on the front | 19:26 |
| fungi | "steal this book" | 19:26 |
| dansmith | hah | 19:26 |
| -opendevstatus- NOTICE: The Etherpad service at etherpad.opendev.org (also used by meetpad.opendev.org) is being restarted now to apply a security update; any open pads will display an error message until refreshed, and author colors may get reset. | 19:53 | |
| enricokern | o/ | 20:31 |
| fungi | howdy enricokern! | 20:31 |
| enricokern | i noticed there are some discussions about my "ai" assisted patches i commited | 20:31 |
| fungi | which ones? | 20:32 |
| enricokern | the horizon caching ones. its like 13 in total | 20:32 |
| fungi | i don't think they were discussed in here | 20:33 |
| enricokern | yeah maybe i got it wrong and it was in the horizon channel. | 20:33 |
| fungi | from the horizon team's ptg etherpad, it looks like they may be planning to discuss ai-generated contributions more broadly in a session tomorrow | 20:33 |
| enricokern | gotta check. can i join the PTG tomorrow for horizon without signing up before to explain ? | 20:33 |
| fungi | yes, no need to register for anything, just connect to whatever room they're using (should be able to click on the session in the schedule at https://ptg.opendev.org/ptg.html and it'll connect you) | 20:34 |
| enricokern | ok cool. will be there. I understand the fuzz with AI and i dislike stupid commits unviewed myself, but my changes are running since months in production large scale environments and were properly tested | 20:35 |
| enricokern | yes its AI code, but its not like its hundreds of lines its all pretty minor | 20:35 |
| fungi | tmazur is the horizon project team lead though and is also in here, so may be able to better direct you to where and when the topic will be discussed | 20:35 |
| enricokern | i also never commited anything myself, i do openstack since 14 years now. im not a coding guy, but im the guy who put this in production and made tons of customers happy with openstack as an engineer :) So i just want to explain the reason behind it | 20:36 |
| enricokern | i never intendted to push them anyway, i just did because i was asked. i never did so maybe i also did some errors or missed somethin, but i tried to correct it | 20:36 |
| fungi | yeah, i've also been involved in openstack for 14 years, thought your name looked familiar | 20:37 |
| enricokern | but for the past discussion, i think a proper AI guidelines really is necessary | 20:38 |
| gouthamr | o/ enricokern: glad you're asking here. people talking is what kept this software progressing over those years! | 20:39 |
| gouthamr | thank you for choosing to contribute! if you're willing, understand the horizon team's problems with AI-assisted commits | 20:39 |
| gouthamr | and the PTG session would help in that regard | 20:39 |
| enricokern | im super willed. and i will join the PTG session tomorrow | 20:39 |
| fungi | openstack's parent foundation has some ai contribution guidelines (recently refreshed) at https://openinfra.org/legal/ai-policy | 20:40 |
| enricokern | yeah i also updated my commits to reflect that | 20:40 |
| fungi | openstack as a project doesn't have its own separate from openinfra foundation, but individual openstack project teams may have specific policies (the kolla team for example), and it sounds like the horizon team may be looking at doing so as well | 20:41 |
| enricokern | i also read all the code changes, its not a blind submit. obviously i lack the deep insight in the code myself, i can just report what i see from 6+ months in operation of the code | 20:41 |
| enricokern | i dont say anyone should push ai slop into a project, thats not productive | 20:41 |
| enricokern | but for people who know what they do and have the proper test environments, it changes the scope | 20:42 |
| enricokern | i highly welcome rules regarding to AI, nobody needs slop | 20:42 |
| JayF | enricokern: the difficult path to navigate is defining what "slop" is ... it's a different bar for everyone, and digging through it is tough | 20:43 |
| * JayF recalls the invalid ironic security bug that took multiple hours to disprove | 20:43 | |
| enricokern | it just looks kind of strange that i pushed 15 different commits today as a first commiter using ai. as i said i did it because i was asked by the people who use the patches, they could just have happily kept using the custom images | 20:43 |
| enricokern | i agree | 20:43 |
| enricokern | gladly the openstack people like you who do this for ages have the propper board and means to discuss this and make the right decisions | 20:44 |
| gouthamr | enricokern: don't interpret written text as canon, unless it reads like red lines :D | 20:44 |
| fungi | also, the llm craze has more generally just resulted in a sudden influx of proposed changes, overwhelming the maintainers and forcing them to make more decisions about what they can effectively prioritize reviewing | 20:45 |
| enricokern | absolutly | 20:45 |
| gouthamr | our PTGs are judgement-free zones really, we discuss quite a lot, freeform... so don't get offended or assume the community's rules are in etherpads | 20:45 |
| fungi | not that they ever had enough bandwidth to review every proposed change as it was, but it has gotten a lot harder lately | 20:45 |
| gouthamr | ++ | 20:46 |
| enricokern | i dont get offended dont worry :) as i said i dont care if the patches get in or not. would be nice if people who know it really check it up, but if not , we just keep using it for our clients as custom images | 20:46 |
| enricokern | and obviously it would make sense to get our company also more integrated and do more contributions (humans) | 20:47 |
| gouthamr | we're struggling with "sloppy" contributions, so i can resonate with the team's language on the etherpad. so far, my mitigation is 1x1 conversations with each contributor in their gerrit changes.. and have had success doing so.. this doesn't scale though.. so, in parallel, we're trying to share some skills to align maintainer expectations with contributors | 20:47 |
| enricokern | we also work usually with partners that we also pay who in turn contribute to kolla as example. so indirect we so far didnt contribute ourselfs, but via our partners that are contributors that we pay for changes that we found | 20:48 |
| gouthamr | here: https://review.opendev.org/q/openstack/agentic-workflows | 20:48 |
| gouthamr | its work-in-progress though, but you see our intent there | 20:48 |
| enricokern | it absolutly makes sense. and we are just at the beginning and its overwhelming already | 20:49 |
| enricokern | need to put the guidelines and stops right now in or its to late | 20:49 |
| enricokern | all that maters for us so far atm is, our horizon loads in 4 seconds instead of 90 or 120s for the same task. If others can benefit from that fine, i also commited some side bugs found in actual use. e.g its not like "check it" no its actually really used by 6 different clouds daily in germany with a total of hundreds of hypervisors and 50+ operatos | 20:51 |
| enricokern | and skyline anyway should get rid of horizon at some point anyway | 20:52 |
| enricokern | we have a mature product with openstack, and all this vmware buzz and even the AI stuff that we run ourself on openstack and alot of others lately push openstack further. i welcome that | 20:53 |
| enricokern | thank you all for the warm words and welcome. i will join the PTG tomorrow and then lets see :) | 20:54 |
| fungi | of course, all the best and enjoy the ptg! | 20:54 |
| gouthamr | great, ty for caring about horizon! | 20:55 |
| enricokern | i missed tmazur if he comes back wink me for him. i will join later again. thank you all! | 20:56 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!