Linux kernel source (well actually build object code) ("ksrc") Anime episodes ("tt") General archive, mostly home photos ("arc"), source tarballs, scholarly publications, etc. Photos from my phone yet unsorted and uncompressed ("k") Git clones that I needed at one time or another ("prog") Installation images ("sw") One oddly large issue, 02488269 - logs downloaded from Jira for analysis Everything 3D printing related VM disk images and XMLs My mp3 collection - oddly enough mostly bought at Amazon if you can believe that
I was sure that videos were by far the largest consumer of disk space, kinda surprising to see Linux up there.
@sun@anonicus It can't be real. What kind of hardware supports 1TB? The biggest I've seen so far has been 192GB. It is within one order of magnitude, but still.
@sun@lanodan@Zergling_man Sadly I have to print every day. I mail a couple of parcels every day for my eBay side hustle. Every one of them needs a short letter and a documentation cum insert leaflet. My printer is a Brother 2740. It's been solid for many years. Full support under Linux. Picks paper no matter how thick the stack is, down to the last sheet. HP is generally trash in my experience. The last good printer they did was LaserJet 4.
@lanodan That's the point: there will always be smart kits who are attracted to the good ratio of usefulness to the downsides of C. But C++ is basically COBOL of 2030s.
@kaia I saw a breakdown for either MS Word of LibreOffice, in which the vast majority of space used by its packages was fonts for all the world's languages that you'll never use. The 2nd worst waste of space was the subpackages and plugins that supported compatibility with other file formats. Cannot quite find it now, but this is how I remember it.
@sun@allison@meso Remember that X11R5 and earlier relied on server-side rendering heavily. But all the text is done by client now, with e.g. Xft2. So you see these bitmaps pushed even in terminal emulators.
@sun@allison@meso The only thing that makes remote X workable is ssh -Y. But even with that, I would go to significant lengths in order to avoid it (like running VMs or port-forward for Vinagre SPICE on 5390).
@sun@mia@shibao Meanwhile the management says that complexity and enormous YAMLs are not a big deal, we'll have AI write all the scripts and manage all the pods very soon.
@sun@shibao@mia Still, I'm watching a k8s demo right now and it's absolutely horrifying, a sea of YAML that a guy is deftly navigating to change one variable and "editing this causes a restart" and magic happens, now your whole cloud is upgraded from OSP 16.2 to OSP 18.01. Meanwhile someone did something like that by accident, IP renumbering happened, and then DHT cannot find any data, users are unhappy. They call our 1st layer support, they call 2nd and the guys mostly put out all the fires, get all the workloads restarted, but data is still not there. Eventually someone pinged me, we found that someone did everything but forgot to actually write out the rings, so everything looked good in builder, but services uses old IPs that they consumed from old compressed rings. The k8s only makes it even more opaque and impossible to fix, even find what's wrong. I hear people say I'd rather support JS in IE6, or program Android, or all sorts of grass is greener stuff.
P.S. I understand that the contents is what matters. I used to fix a 1,000 line code in FORTRAN IV, which was a little difficult. But still, boring Python boilerplate can run for much longer than 1kloc.