IRC channel logs

2023-05-08.log

back to list of logs

<RandyFarmer[m]><Zarutian_iPad> "Randy Farmer: re https://..." <- The blog post announcing the debugger is ASCII, and I included the full text are the end. I'm calling that done.
<sneek>wb Zarutian_iPad :D
<RandyFarmer[m]>We are now hosting the debugging video.
<RandyFarmer[m]> https://spritely.institute/files/videos/Causeway-Live-Demo.webm
<drakonis>noice
<dthompson>Goblins 0.11.0 has been released! https://spritely.institute/news/spritely-goblins-v0110-released-time-travel-distributed-debugging-and-more.html
<vvvvmvr[m]>Congratulations! Excellent work on this release
<civodul>dthompson: wo0t! congrats everyone! 🎉
<civodul>the debugging story is awesome
<dthompson>civodul: thanks! still early days but pretty cool so far!
<civodul>i was browsing the manual and everything looks so lean and clean
<civodul>re debugger, how much of the interactions and past state is kept?
<dthompson>civodul: it's customizable. we store data in a ring buffer and we picked an arbitrary default of 256 "events"
<civodul>neat
<civodul>woow, really cool
<dthompson>the ring buffer can be resized during a vat's lifetime, too.
<civodul>nice
<dthompson>our history tracking is not yet optimal. future work there to golf down the size of the data we store.
<civodul>ok
<dthompson>but it's not too bad if your vat doesn't have a ton of actors
<dthompson>as we dogfood more we'll see what makes things fall apart :)
<civodul>heh :-)
<dthompson>of course to really dogfood this thing properly we'll need to add network support. we'll be starting on that problem real soon.
<civodul>do you currently assume that vats are on the same machine?
<civodul>(newbie question)
<dthompson>civodul: yes that's the current limitation for "phase 1" of the debugger
<dthompson>there's some prior art for debugging over the network that we'll be looking at.
<civodul>oh that's a limitation of the debugger
<civodul>ok
<dthompson>ohhh sorry I misunderstood the context of your question
<civodul>or does the whole history tracking feature lack network support?
<civodul>trying to make sure i got it right :-)
<dthompson>the history tracking is per vat, so you can't query the past state of a remote actor.
<dthompson>due to the aforementioned limitation
<civodul>ah ok, now i got it, thanks :-)
<dthompson>what's interesting is that the notion of a "vat" is not encoded in CapTP, the network protocol
<dthompson>which is good, because it means interoperability is easier. a system need not subscribe to the vat model to participate.
<dthompson>but it certainly poses an interesting question for network debugging. :)
<dthompson>I have a lot of research to do, but maybe there will end up being an actor that is a proxy to a remote vat that can be used to retrieve debug data. we'll see :)
<Gnuxie>why do spritely folk not buy into ocap all the way down though and it's just game that having a secure environment doesn't include the implementation itself
<Zarutian_iPad>hmm?
<Zarutian_iPad>all the way down to hardware CHERI style or?