IRC channel logs
2026-02-04.log
back to list of logs
<ridley>Can someone post in here when the FOSDEM recordings are up? I managed to catch Dave's first presentation but wasn't able to watch any of the others live. <PuercoPop>How expensive are vats? Are they in the general vicinity of a thread or more of a green thread? f/e would it make sense to model an ActivityPub server with one vat per resource (Actor/Inbox/Outbox/Followers/etc) or is more like if I have a web server I spin one vat per CPU and have each vat handle multiple requests? <PuercoPop>Yeah, I saw that. I took it to mean like it can only one thing/task/actor at a time <dthompson>hey everyone, online for just a moment before my flight back from FOSDEM <dthompson>ridley isn't here right now but regarding FOSDEM recordings: 3 out of 4 of my talks are now viewable and the 4th should be coming soon. I believe the other talks need review by christine and jessica when they get back home. <dthompson>PuercoPop: vats are green threads aka fibers <dthompson>so they're cheap in that regard. it's *not* one pthread per vat. <dthompson>ArneBab: that example in the manual is showing how to make a custom vat that uses a pthread. just for demonstration purposes. you're right that the default vat implementation uses fibers. <ArneBab>I had completely missed that the CRDT room existed ⇒ more talks to watch <theesm1>hey everyone o/ re our goblins shepherd discussion at guix days: replaced my guix home shepherd with the goblins shepherd (based on the spritely/shepherd repo) as promised on my train trip back home & it didn't even cause chaos & is running stable so far :o <dthompson>*I* haven't even done that and I also haven't run the goblins shepherd build in awhile <dthompson>the newest update is just that ludo gave me commit access to shepherd today <dthompson>so we are going to *slowly* proceed with bleeding edge sheperd experiments <theesm1>dthompson: yup, was surprised how flawless it worked! got a bit of train delay right now and two hours to spare, so maybe i'll even feel experimental enough to drop it in as PID1 even though I'd expect breakage then <theesm1>always looking forward to make my weird computer setup even weirder on the pocket reform :p (if there are metrics you're interested in let me know! like ram usage over time compared to non-goblins shepherd etc.) <dthompson>we've been so focused on making it work at all that we haven't done any measurements ourselves yet <dthompson>I expect it to have higher baseline memory usage which is probably fine but what I don't want is unbounded growth due to leaks <dthompson>our internal dogfood chat application has a server process that has had many months of uptime without crashing or exhausting server resources so that's encouraging to me <theesm1>yes, i'll try to share stats after a week of usage (think my atop records those anyways, so maybe that'll even allow a comparison to before... kinda curious about that) <ArneBab>I’l have to switch to guix home first, I think -- still using a plain manifest. <ridley>Ooh, glad to see the shepherd experiments are progressing <sneek>Welcome back ridley, you have 1 message! <ridley>Guess I have lots of things to watch now! <ArneBab>sneek: later tell dthompson: shepherd on goblins sounds like goblins is becoming more critcal (fleet management), so it could be time for pentests. Sadly the security testing capacity of https://nlnet.nl/NGI0/review/ has been exhausted … <ridley>Ah... I was just thinking yesterday I should make sure to get in the NLnet security testing queue. <theesm1>hope the general nlnet/horizon grants budget will increase within the next years (as it seems that there's been increased political interest in doing so during the last few months) <ridley>Looking at the FOSDEM talk list is always overwhelming. Any recommendations outside of the Spritely ones? <ArneBab>sorry that it’s just the video links: I downloaded them to watch them offline <theesm1>ridley: the what's new in git talk on the main track day 1 was interesting, can't remember the exact title though <ridley>theesm1: "Evolving Git for the next decade"? <ArneBab>that’s also a pretty good overview of a wide view to performance programming.