I have a #snac server running under NetBSD on a 486, which also runs X. It's indeed painfully slow but it does work. And surprisingly well, too. An OS from 2025 on hardware from 1994.

And if it wasn't for crypto being too slow to actually work I'd be doing the same on the 386SX-class machine that I also have running NetBSD. But with a hyper-optimized SSH handshake taking over a minute, I have no hopes for 2k RSA signatures or any kind of TLS handshakes with remote instances happening in anywhere near the timeframe they would need to..

I'd like to say that is an achievement no matter what your first language is ;)

CC: @linuxenjoyer@blahaj.zone

Reminds me of all those times we were hiring sysadmins for our "hosting organization". And the majority of applicants came from the hospitality industry, despite our ads very clearly using words like FreeBSD and MySQL and titles like Command Line Warrior and stuff..

My jaw literally dropped here when I read this. It's so, so familiar.

I don't know what your native language is, but imagine how this exact same thing feels for someone whose first language is not English .. All the same scariness plus we may not even know what the words mean! And looking them up in a dictionary leads to a whole other kind of rabbit hole, and even if you understand the definitions and use in normal human language, it does very nearly jack shit to help understand wtf it means in the programming context.

I know this, I've tried to learn programming since I was, what, 8? In a vacuum too, since I lived in the middle of fucking nowhere in Norway for the first 17 years of my life. Imagine only having the MS-DOS or PC-DOS handbooks and some GWBASIC code written by Bill Gates to start out with. And the vocabulary of a 8 year old kid whose grasp of the English language is limited to what he learned during 6 months of school in Australia when he was 5...

I'm almost creeping up on 48 now and still can't code for shit.

I'm having the weirdest problem on this machine .. Trying to use #DTrace on #NetBSD to profile #snac2, sometimes it works sometimes it will not actually give me any data (just print the headers). I'm using the hotuser script from the OpenDTrace toolkit.

Another issue is that the only way to stop dtrace is to kill -9 it, which takes the watched process with it in the fall..

Halp? :)

It's so sad when it's over ... I envy anyone who's able to see it for the first time in this day and age.

#Babylon5

Larry Laffer approves this message!

I haven't obsessed this much with my #X11 desktop since I was .. much younger. Trying to make it functional and purrty on 30 year old hardware is challenging but fun. Thanks to #NetBSD and the fact that most old X and #Xorg tools are still around and do all the things they always did - and haven't bloated much in those 30 years - helps a lot.

I'll take Wayland when it comes my way without me having to lift a finger, but until then I'm glad the X Window System is still around. Keeps this old hardware useful.

Trying to optimise http://floppy.museum for (even) older browsers. Some of the issues I'm trying to solve include utf8-to-latin1 translation (the original HTML has some silly double- and triple-byte characters), and variations of JPEG that simply aren't understood.

Turns out Netscape 2.02 is too easy, so in this picture is IBM WebExplorer v1.1h running on OS/2 Warp Connect. Using the magic "work area" feature of folders (mark a folder as a work area to have the OS manage objects within it as a kind of unit), I can open several windows at once. True multi-process browsing 😉

#retrocomputing #browsers #floppy #museum #html #BrowserWars

The sound of a 486SLC doing a SSH handshake.

Well .. that went sideways. Despite many attempts, I have yet to find a 286 - or a 386sx for that matter - that will boot the NetBSD floppies without failing in some way or other. I'm not yet certain (perhaps someone here knows?), but there may be instructions missing from the various 486SLC and 486DLC CPU variants that my ugprade modules have. Or there are other bugs that I have not been able to figure out.

Anyway, I've reduced my ambitions ever so slightly, and am now in the process of installing NetBSD (-CURRENT) on what is essentially a 386SX-class machine: 16-bit bus, 24-bit addressing, 16MB RAM, and nearly as unpleasantly slow as the 286 I had planned to use. It is however equipped with an IBM-branded 486SLC, which is from the Blue Lightning series. This one definitely has a full 486 instruction set. More hardware details will follow when I've completed the build (and installation).

Meanwhile, the obligatory screenshot from the installer. Note the ETA for simply unpacking base.tgz ..
#RunBSD #Retrocomputing #Slowcomputing

Herewith, the actual machine. Some specs ..
- Motherboard: Zida TD60C
- CPU: Harris 286 28MHz
- RAM: 16MB "RainbowRAM", 55ns (check out the LEDs!)
- Graphics: Cirrus Logic CL-GD5422, 1MB
- Network: 3com EtherLink III (3c509)
- Floppy: 2-in-1 90mm (3.5") and 1.2" floppy drive
- Storage: Promise DC200M caching IDE controller, 2.5MB cache
- SCSI: Sound Blaster 16 SCSI :D

#sfw #hardware

I had #Minix on another 286 for a while, and there's some 16-bit "port" of Linux that is not as old as it should be. And of course the venerable OS/2 1.x, and a few actual #Unix variants as you point out. I didn't know about V6on286, that's a beautiful little nugget, thank you!

All of those things are absolutely wonderful and make many of todays software developers look ... spoiled? What I want, however - and what I love doing - is making this old hardware do stuff its makers never dreamt of, things that are as far removed from their time as possible. That's why I will, if #NetBSD permits, run bleeding edge BSD on a 286-on-486steroids, and why I run web+ftp+irc servers (yes, multitaskign) on one 286 and multiple BBS nodes on a 386 - like one used to do, of course.

I cannot state often enough how amazing it is that there's still software developed today that will work under such constraints.

#RunBSD #EndOfRant

And it lives! Apparently I'm officially operating a Motherboard Bakery! :)
Now I need to get it properly configured and tested with DOS, then I can move on to the next steps - which involve the CPU upgrade, and assuming that works, creating actual, physical floppies.

(In other news, the #snac instance on this poor Pentium Pro server is sweating hard whenever I post something. So let me know at @ltning@anduin.net if you have problems receiving/reading my posts. I've made some tweaks but it will be unavoidably detained for a while following each post, my apologies for that..)

#BIOS #POST #Retrocomputing

And I know one should not apply regular water to electronics, but I'm trusting Oslo Waterworks to supply me with nice, clean water from the tap. It comes at high pressure to boot, so I can quickly get rid of all the vinegar to prevent it from slowly eating through everything. As long as I dry it properly and quickly, it should be fine, I'm told.

I said I was baking - It was not a randomly chosen term.

I'm a bit of a n00b at this, but I have salvaged a couple of boards before using vinegar and a toothbrush, so I'm trying my luck with that. After applying the vinegar and watching, hearing and smelling it sizzling, I dunk the affected part thoroughly for a few minutes...

So as usual, its eBay to the rescue. Found a board just like it, except there's good news and bad news: the good news is the CPU is socketed on this one, making the upgrade job a lot easier.

The bad news? See picture.

Baking my next #NetBSD project, a thread...
[Edit - fixed attachment]

I've said somewhere I want to run NetBSD on a #286. Now obviously that's not actually possible, but I should be able to do the next best thing - run it on a 286 upgraded to a 486SLC!

But wait, most 286es only support 4MB RAM, although the ol' chum of a chip supports a whopping 16MB. So I have to find a motherboard that can do this.

Thing is, I already have one. See picture. But it's currently occupied doing very important Enterprisy stuff - it runs IBM OS/2 1.3 Extended Edition..but at least I know what I need!

For the server hosting this instance (Pentium Pro machine):
- https://dmesgd.nycbug.org/index.cgi?do=view&id=8170
- https://bsd-hardware.info/?probe=f6feb27f87

For the 486 running larry.weirdr.net:
- https://dmesgd.nycbug.org/index.cgi?do=view&id=8169
- https://bsd-hardware.info/?probe=2ec3f61bc7

Enjoy. :)

CC: @jaypatelani@bsd.network

Yes, been doing some hardware work. :)

Nah, it's the same size (roughly) as the regular kernel, the main difference is it allows ISA/VLB VGA for the console driver, afaik.

CC: @dch@bsd.network

See attached. This is a different motherboard but same CPU, with a PCI NIC, a second ISA NIC, and a completely unrelated Gravis Ultrasound PnP :D

CC: @grunfink@comam.es @stefano@bsd.cafe

I had to build my own kernel since the default kernel simply needs too much RAM. But once USB and a couple of other niceties were removed, it ran fine with the binaries from the i386 release.

Most of them can do more than they announce. It's like OS/2 2.11.. IBM said it can SMP across 64 CPUs but they could only test with 32 as that was the best hardware they had at the time :D

Yesterday I found a 486 board in my collection that boots and happily deals with 256MB of EDO RAM! The speed is hare-raising :D

You must let me know when you do. All modern Unixes I know dropped 386 support a while ago :(

CC: @grunfink@comam.es

Is the telnet emulation issue related to "cls" from some BBSes just giving a weird character instead of actually clearing the screen? :)

I did it. Be gentle. https://larry.weirdr.net/

I did a thing so I don't have to wait 20+ seconds to SSH into (or out of) my #486 running #NetBSD.

https://github.com/ltning/sshbench

#unix #ssh #benchmark #shellscript #bourne #runbsd

Ah. Makefile.NetBSD is missing sandbox.o. Consider this a bug report.. ;)

CC: @grunfink@comam.es