- I'd enjoy hearing from someone knowledgeable about Linux kernel evolution what they consider the big contributors to modern Linux taking up many times more resources than it did back when I ran early Slackware on a similar 486 with 12 MB of RAM. NOT so I can rant about how "modern software is bLoATeD because programmers are LaZy" (ugh), but I'm truly curious what has changed, because we usually tend to think of software growing over time for several reasons, some of which don't seem very applicable here:
1. Using a higher-level language/runtime that's more efficient for programmer productivity but that outputs less efficient code — nope, still using C.
2. Decreased effort on optimization. I'm sure this is a factor, but I also wonder how applicable it is here. In my day job, I may carelessly use an exponential algorithm or an inefficient data structure because it's quicker to just get the job done, but my impression is that the kernel is still going to pay a lot of attention to such things. Not quite as much as 30 years ago, probably, but kernel engineers aren't just using bubble sort all over the place now.
3. Increased functionality. The modularity of Linux, especially the kernel, mitigates this somewhat: sure, a 2025 kernel supports thousands more devices, file systems, etc than a 1995 one did, but you can compile most of those out. Still, my impression is that we end up with a kernel not just somewhat larger but many times the size of the old ones.
4. Tunables (or things that potentially could be tunable). This would actually be my best guess for where big easy wins could be had: code paths that allocate a 1 MB buffer when they could get by—and would have, years ago—with a few KB. On a system with GBs of RAM, larger allocations like that are probably advantageous but they can really crush an old system.
Likewise in userland: `ls` surely has more options now, but not _that_ many more, right?
- As the last line says: Here there's a bill there's a key!
- I was surprised to read this part: "For now, FreeDOS 1.4 can't run Windows for Workgroups in enhanced mode, but can run Windows 3.1 in standard mode."
I know very little about the project but I'd have guessed that ~100% DOS compatibility would've been achieved early on. There's just not very much to DOS! I'm sure there are reasons, of course, would be interested if anyone knows.
- Nice. Thank you for what you do making this site a fun place to visit.
- What was special about its interrupt handling?
- The PDP-11 was one of the first systems to use vectorized interrupts and interrupt chaining. This meant that when an interrupt occurred, the CPU could jump directly to a specific handler for that device, execute its code, and even pass control to other handlers if needed. Once completed, execution would resume exactly where it left off.
For example, a sensor connected via a bus card could trigger an interrupt when new data was available, allowing code to store the value in memory or write it to disk automatically, without the main program needing to check for updates. This eliminated the need for wasteful polling, making the system more efficient and responsive by handling events asynchronously in the background.
- I had one of these! (Probably still do, somewhere.) Got it when they were on clearance.
It was kind of a glimpse of the future in some ways. A small, light, fanless machine with a usable keyboard. And it ran for so many hours that you didn't have to think about batteries, you just used it. We're (largely) used to that today, but that's not what laptops were like back then. For doing simple tasks (writing a journal in Pocket Word, etc), it blew away my big ThinkPad.
- In 2024, “8-bit” in the lexicon is a signifier of “old graphics”. When most people say “8-bit”, they don’t mean anything technical (like “2^8 = 256 colors”), they mean “looks like an old Nintendo or whatever.” Put a slightly higher-res version next to it and they’ll say that’s 16-bit. Same way old timey cars are all Model Ts to me.
- I'm having so much fun with my Playdate. If you like old retro games on amazing hardware, it's a gold mine.
- >Sneaky Apple marketing considering (afaict) Windows was already outselling OS/2 even before the 1990 launch of Windows 3.0.
Yeah, but OS/2 was the heir apparent and what serious people used (well, they were all planning to, one day). Comparing themselves to OS/2 (and thus IBM, the object of all of Apple’s ire) would have been perceived as giving themselves an air of legitimacy in 1990.
Sure, a bunch of people are running that toy “not even an OS” Windows. Half of ‘em are probably on clones!
That attitude changed really rapidly.
- IMO this is a good example of why Microsoft deserved a lot of their success with Windows 95.
I was a big OS/2 fan at the time, but this is the sort of thing IBM was institutionally incapable of coming up with. Sure, OS/2 supported long file names… if you were on an HPFS file system that (practically) only OS/2 could read. But I'd guess the significant majority of non-corporate OS/2 users dual-booted with DOS. And, IIRC, files with long names were totally invisible to the otherwise superb DOS emulation. And I feel like I even remember some OS/2 programs having trouble with long names because their authors were more likely than not on FAT file systems. So it always seemed more like a "theoretical feature".
- More
Kernel has really only gotten bigger for security, scheduler, file system and network.
Command line, server only installs on raspberry pi are surprisingly light on disk and ram consumption. We expect ipv6, a good file system that doesn’t blow over in a power outage. I don’t think twice about pulling USBC power when things go sideways. 486 Pc era. No chance. That was a moment to reacquaint myself with a diety.
Userland wants more. Some standards are followed. Some code is shared (good and bad). I want 4K hidpi and fonts and colours on a raspberry freaking pi. That convenience costs in terms of storage, compute, and ram. Compare to Mac or windows and Linux looks anorexic.