jacky has quit [Remote host closed the connection]
jacky has joined #spectrum
edwtjo has quit [Ping timeout: 246 seconds]
edwtjo has joined #spectrum
edwtjo has quit [Ping timeout: 272 seconds]
edwtjo has joined #spectrum
cole-h has quit [Quit: Goodbye]
vilhalmer has quit [Ping timeout: 244 seconds]
vilhalmer has joined #spectrum
tazjin has joined #spectrum
<pie_>
i suppose this might be relevant here, hopefully what i say/ask will make some sense to someone;
<pie_>
-*- pie_ noob
<pie_>
<pie_> runtime structures arent purely functional and compositional. it may be illustrative to take a look at how capabilities based OSs deal with this (which i havent done), but at the end of the day if youre doing some sort of integration, you have to pull all your threads into one place, and so its also a question how that should be done (windows registry? dbus?) (not sure if dbus belongs in that enumeration)
<pie_>
stuff) structures distinction is a problem that i havent seen anyone writing about, and which i think needs consideration - or likely im looking at it wrong
<pie_>
<pie_> energizer: polezaivsani: how do you handle integration while keeping the benefits of the nix approach, by which i mean things like task switchers not being able to access the necessary informations (like icons) from running programs to work properly, because they dont know how to find the information. arguably this is due to bad infrmation discovery APIs in things unrelated to nix, but i think the dead(data at rest)/live (runtime state
<pie_>
<pie_> if you start something like a kde shell, you expect things to just work?
<pie_>
(xpost from nixos-chat)
<qyliss>
Yes, there's a reason I'm not going to provide GNOME in Spectrum :P
<qyliss>
Overall I still don't have a great idea of how all this will work, though.
nicoo has quit [Ping timeout: 240 seconds]
nicoo has joined #spectrum
cole-h has joined #spectrum
<qyliss>
I'm trying to figure out what to do next, now that I've got compositor VM working
<qyliss>
(At least, working well enough as a proof of concept)
<qyliss>
The most achieveable NLnet milestone is called "Implement hardware isolation", and involves investigating hardware isolation methods, and publishing a report about them
<qyliss>
I think I've proven that the Wayland compositor doesn't need to run on the host, which was one worry I had about isolation
<qyliss>
Even though technically I haven't implemented stuff like display passthrough -- it doesn't really need any research to show that that is possible, it just needs to be implemented.
<qyliss>
So probably what I should do is look at other hardware isolation.
<qyliss>
Like network devices
<Profpatsch>
qyliss: oh lol, that plays into the weechat config discussion we just started in $otherchannel
<Profpatsch>
GNOME has the same, dynamic config you can change from UIs at runtime
<qyliss>
Yeah
<qyliss>
I think to some extent we have to just try to move past that
<qyliss>
Another thing I should probably do soon is start on host service supervision
<qyliss>
Because starting everything by hand is starting to become infeasible
mx08 has quit [Quit: WeeChat 2.8]
mx08 has joined #spectrum
<qyliss>
The other thing I should be starting to do is invite people to suggest how they'd like to configure a Spectrum system in Nix
jb55 has quit [Remote host closed the connection]
<qyliss>
That feels a bit scary though, which is why I haven't done it yet
<IdleBot_2e4f9b4b>
Apriori it is not even clear what should be the ration of Nix configuration and start-time configuration…
<qyliss>
Yeah...
<qyliss>
cole-h: just checking -- do you plan to review the crosvm read -> recv patch? (you reviewed 2/2 but not 1/2). No pressure ofc -- just want to know if I should wait :)
jb55 has joined #spectrum
<cole-h>
I figured there was no need to, since it's already on interguest :P
<qyliss>
Ah, I see
<qyliss>
interguest is really my WIP branch
<qyliss>
i.e. things that I would like to be reviewed at some point will show up there before review, because they're required to get something to work
<qyliss>
The doc patch didn't go to interguest because, well, it's a doc patch :P
<cole-h>
Ack. OK, sending along a reviewed-by for 1/2 :^)
<qyliss>
If something goes to the list it's because I think a review would be useful if somebody wants to give one :)
<qyliss>
Thanks so much!
<qyliss>
Hope you know how much I appreciate all the reviews!