First time I’ve heard about libadapta. Very happy that folks are doing something about GNOME’s anti-theming dogma and I’m excited to see what libadapta or a derivative might bring in the future.

I’m very glad to see projects like libadapta as themable alternatives to the libadwaita dogma. I’ve painstakingly themed my desktop to look and feel like a cohesive, modernized NT 4 workstation and should seriously consider contributing to libadapta in anticipation of libadwaita coming to more and more programs.

I am very stubborn about my computer’s GUI, but also hopeful the community can bring back theming where GNOME is dead set against it. If they can make WindowBlinds for modern Windows, the equivalent in Linux is definitely achievable.

The lightest Windows 95-esque setup I’ve achieved was IceWM on Debian. Manually install the GUI to avoid unnecessary packages. Around 200 MB RAM usage from cold boot and very snappy on an Atom netbook with 2 GB RAM. With zram swap set to 50% of total RAM (swapping to the tiny, slow eMMC proved frustrating), I could comfortably browse most websites and work in LibreOffice. If you use a no-frills distro (like Debian), performance shouldn’t change too much with updates.

It should come with a Windows 95 theme, but some settings are available only in the config files. Adding a theme like Raleigh for GTK3 will make it look more cohseive without consuming much extra resources.

As for Wayland, I think the only performant options would be labwc or a modification of Weston. I’ve no experience with XFCE on Wayland, but that would open up the option of the Chicago95 theme.

Seeing where desktop Linux was just less than 10 years ago and where it is now gives me optimism for mobile Linux. But I suspect the overlap between developers and users of those 3 features is pretty small, so they might be a ways out.

I was about to suggest getting a head unit that isn’t tied down to CarPlay or Android Auto, but then I realized I drive a really old car from the days you’d easily take out the faceplate or the whole unit to deter theft.

The readers in consumer-grade computers have been hit-or-miss for me. The contacts seem to either lack gold plating or develop a layer of tarnish.

Someone a while back put a set of very cheap SSDs through a torture test, and after exhausting many of the write cycles, left them alone for months. When powered back on for reads, the drives were slow as the error correction hardware was working overtime to compensate for the loss of trapped charge over time, but mostly recovered their performance after a while.

That said, I have a similar anecdote where one of my very worn test bench SSDs kept complaining about the same bad sectors despite OS reinstalls until I just overwrote it with zeros using dd. Was fine for many months thereafter.

I’ve no idea either if SpinRite has some secret sauce that FOSS utilities have yet to replicate, but it sounds like a non-destructive read-write test with badblocks ought to do the same.

Also, my CF cards, SD cards, or USB drives from the early 2000s and early 2010s almost never give me trouble despite spending years unplugged. More recent flash memory is a different story though and I suspect the shrinking gate sizes and advent of TLC/QLC/PLC haven’t helped. I’ll usually splurge a little these days to get the industrial or high endurance MLC flavors and hopefully avoid the issue.

Absolutely. And in the case of laptops, there might be no way to upgrade the GPU. I’m keeping an eye on this since I have an otherwise perfectly good laptop that also needs an unsupported legacy nvidia driver.

Not flawless, but that’s on me for insisting on a very particular look and workflow that involves lots of manual config editing.

I’ve noticed the same thing, if a laptop has a dGPU, it’s almost certainly Nvidia. My pet theory is that the integrated GPU of AMD APU chips are decent enough that laptop manufacturers who work with AMD don’t bother tacking on a dGPU.

Doing anything with partitions safely means having a current backup of your disk.

What you want to do might be possible if your disk was set up with LVM, but I don’t have too much experience with it. If you have a more traditional partitioning scheme, the problem is that you can’t really shave off the front of the next partition if you want to expand /home to the right, and you can’t tack free space to the front of a partition, if say, you shrink the partition to the left of /home.

If, for example, you want to expand /home into a partition on the right, you’d have to shrink that partition, back it up, delete the original, expand /home to the desired size, and copy the backed-up partition into its new, smaller place. If that other partition is not mounted, you could do the procedure without a live USB, if you insist.

What desktop environment are you using?

Among the people I know in real life, some post (non-tech stuff) to Reddit, some write reviews on Yelp, and some have called customer support hotlines for tech products. But none have ever posted online to ask for tech help, at least not to my awareness. Neither did I back when I used Windows, and not for a couple years even after switching to Linux.

I suspect most Ubuntu users are among that common crowd. They might look up an issue on the internet, but expect to ask for help from a dedicated support center. Or can’t be bothered to sign up for an account and post to the places that can answer their questions, which are usually very “techy” and possibly even intimidating to beginners.

As for my setup, the upgrade from Debian 12 to 13 went very smoothly. I had to fix a few obscure config files, but nobody else really touches them, and it didn’t stop it from booting. Replaced a deprecated package with its Flatpak equivalent as well. Only unsolved issue is the xfce4-panel consuming all of one core on occasion for no apparent reason.

glim is like Ventoy, but built on GRUB with much cleaner code and no blobs. Doesn’t work with as many ISOs though.

Any way you boot it, gparted should work just fine from a live ISO.

If you’re just testing things out, Ventoy should be fine. If you want full assurance nothing fishy is going on when installing, do it the old way and dd the image to the USB.

Linux Mint, as many have suggested, but Fedora would also be a good choice if there’s any bleeding-edge hardware not supported otherwise.

I got away with a 380 MB /boot during upgrade, though that assumes you aren’t dual booting another distro that also needs some room. Have you tried deleting old kernel versions?

But if you want to future-proof, the issue is that shrinking a partition from its “top” is not a supported function. For ease of explanation, suppose we want a 1.5 GB /boot partition:

1. Shrink nvme0n1p3 by 1.5 GB
2. Create a new partition and format
3. dd the old boot partition to the new partition
4. Resize the new /boot partition to the full 1.5 GB
5. Delete the old boot partition
6. For good measure, reinstall GRUB to make sure it is aware of the new partition: https://wiki.debian.org/GrubEFIReinstall

This assumes your fstab file mounts by UUID (default in recent versions of Debian). If not, update /etc/fstab to match the new partition. It’s been a while since I last did this, so definitely have your backup on hand and perhaps double check with other resources in case I left out any steps.

More precisely, shrinking relies on the presence of empty blocks. A filesystem usually fills from “top” to “bottom”, so there would be no empty blocks to shave off the top of your nvme0n1p3, you can only shave off at the end. If you really don’t want /boot at the end, you’ll have to shrink nvme0n1p3, back it up, delete nvme0n1p3, expand /boot, re-create nvme0n1p3, and dd the backup back into its place.

The OS bundled with TI-84 Plus CE Python Edition graphing calculators is a wacky one. It runs primarily on an eZ80 core for backwards-compatibility with earlier graphing calculators while also handling an ARM core for Python functions. Parts of the assembly code can be traced back to the TI-82 calculator ROM from 1993.

Most people wouldn’t think a calculator would ever need a security solution, but it even verifies application signatures so students are less likely to load cheating utilities on them.

Up until 95, Windows was mostly a desktop environment for DOS. From 95 to ME, Windows was an OS that used DOS as its bootloader and compatibility layer. Not sure how to put it, but it was simplistic and fundamentally different from Linux.

The thing with NT-based Windows (including modern editions) is that the underlying system is joined at the hip with the GUI. Whereas Linux with your choice of coreutils is a perfectly capable OS without the GUI, many features of Windows are only accessible through the GUI.

Given enough time and resources, pretty much anything exclusive to Windows could be ported to Linux and vice versa. A lot of the difference just comes down to history and the ensuing conventions, workflows, and file hierarchies.

Even if we stripped out all the cruft and spaghetti code from Windows, there would be lots of nasty idiosyncrasies in its design, informed by its OS/2 and VMS (see Dave Cutler) heritage, profit maximization, revolving door of devs and interns, and years of bending over backwards to accommodate legacy programs.

Despite the major version jump, the update just works. One hour to download and install and one hour of housekeeping, but that’s on me for messing with configs most people wouldn’t ever touch and finding a replacement for a deprecated package. New features for me to check out at my leisure, all well-tested with no disruption to my established workflow.

I hacked together a similar setup back when, for logistical reasons, I had to squeeze everything on one machine.

It’s doable, but be prepared for some challenges.

You will probably have a much better time than I did given the abundance of RAM. Getting graphics to work properly is the most arduous part. I never really figured that part out. Having to forward USB accessories also got really tedious.

Cool kids run Red Star OS