i have a 32 gigs usb drive and, given that most isos are around 2-5 gigs, it's such a shame to see nearly 30 gigs of perfectly good free space go to waste whenever i need to use it as a bootable drive. is there any way i can burn an iso image to the usb drive while still being able to use its free space?

i run debian 13 on my laptop. it runs on a 5200rpm hard disk, so some bootup slowdown is to be expected, but it got really bad for some reason. booting up could take up to 3 minutes just to get to the display manager

after running systemd-analyze blame i found the two main culprits: docker and snapd. i had snapd and flatpak installed so that i could have access to as many applications as i could, but it seems that snaps have a huge amount of overhead. i knew about the one million mountpoints caused by snaps, but the amount of services they have to start on boot surprised me. snapd alone took 30 seconds to start and then there were its dependencies

my boot time is now down to 1min 50s. i recommend anyone who still has snapd installed on a non-ubuntu distro to uninstall it

i know this sounds silly and it should be obvious, but i've been using x forwarding at work for a few days now, but it just dawned on me that i'm running wayland on my plasma machine and the x forwarded window is display through xwayland. it works so well that i didn't even notice a difference and in fact it seems to perform better than on x

this is not even the first time xwayland works better than pure x at work. i also need to use horizon client every once in a while and it got so much more stable after i moved to wayland -- even though the application claims wayland is unsupported

the context is: the 470 legacy driver doesn't compile on the linux 6.12 kernel. because of that, debian decided to officially drop support to that driver. i tried installing the driver myself using nvidia's official installer, but the installation indeed fails during the module compilation stage.

this means i am stuck with nouveau. it got better since i last tested it on bookworm, but one major pain in the ass is that nouveau has no support for performance levels for my card and it runs at the lowest clock bc of that (~400 megahertz instead of its max ~900 mhz).

this causes a noticeable performance hit, even for desktop usage, but it's good enough for work. waching full hd 60 fps video is a bit painful, but it's possible. but gaming, which was possible, got way worse. even a lightweight game like celeste got frustrating to play due to stuttering.

i guess i'll have to deal with it and maybe this is the cue to buy another graphics card and never buy nvidia again, but i'm thinking about what my options would be here:

1. downgrade to bookworm. not easy to do, would only delay the problem.
2. install an older kernel and use only that. not sure how, the official repos only have the 6.12 kernel. i could get the older kernel from the bookworm backports and pin it to prevent any updates, but mixing repos from different versions makes me uneasy.
3. patch the driver. there are a few patches floating around that make nvidia's driver compile on the 6.12 kernel. applying the patch by hand is annoying and i would have to re-apply it at every kernel update.
4. cope.

any ideas?

edit

and it runs at the lowest clock bc of that (~400 megahertz instead of its max ~900 mhz).

that was a mistake. i was reading the clock off of my onboard video chip, which also happens to be nvidia. the onboard chip is at .../dri/0; my graphics card is at .../dri/1. nouveau seems to support reclocking for my card, but i'm trying to change the clock and the video signal goes crazy when i do it

trixie (aka debian 13) is about to get released with plasma 6.3. it seems that finally x11 is being left behind, which is good, but it worried me a little bit because

1. my nvidia graphics card is old: the 470 driver is the latest version that supports it (so no wayland support from nvidia proprietary drivers ever)

2. on bookworm (debian 12, the current stable version), nouveau works pretty well, but it crashed more or less daily when i tried to daily drive it at work

x11 is still very well supported by plasma 6, but the near future has no place to it and i worry i would eventually get stuck without updates to my system as the newer versions lose x11 support. i decided to try wayland+nouveau again on trixie to see if i had better luck this time

it all worked way better than i expected. performance is seemingly on par with the proprietary driver, i've had no crashes so far and i've been using it for a week and even screensharing, one of the most problematic aspects of the experience last time i tried, worked well. the one problem i had was with the slack flatpak, which didn't support wayland for some reason, so it had to run on xwayland. screen sharing wayland applications from x11 apps is possible through the xwaylandvideobridge, which kinda works, but it crashed xwayland entirely at one point, killing both x11 applications i had running. i won't blame that on the system itself and installing the slack deb package fixed the problem anyway

all in all, it seems like i can safely switch to plasma 6+wayland+nouveau at work

i've been playing with cppfront for a few minutes now and it's been a surprisingly pleasant experience so far. i'm tempted to try it out at work to see what happens, but i wanna know if anyone tried to use it in production and what your experiences are

for those who haven't heard of it, cppfront is a cpp2 to c++ compiler, a bit like coffeescript for js. cpp2 is herb sutter's proposal of a new and cleaner c++ syntax with better ergonomics, better orthogonality, and better defaults