On my Intel 486-33MHz box Slackware 1.01 for some reason refused to boot. Happily installed 1.1.2 though and even managed to run X and networking.

Could that be because of the virtual resolution 1024x1024 in the Xconfig?

No, it is the XF86_Mono server that I needed to use at first due to lack of S3 video chip support. The clocks it uses are hard-coded I think, and that it is a true mono-mode, maximum 64k of video memory. That effectively will limit it to 640x480. However, the example suggests that 800x600 is possible, but I don't know what display standard as a mono that was supported by? I guess I'd have to read the code to know what it is.

That reminds me. When my boy was reading through the install instructions, it said something about if you had any problems with Linux, send email to some Linus person.

It does seem like a hardware issue on kernel 0.99p12. It would be cool if you could figure it out! The kernel that comes in 1.1.2, v0.99p15, does seem like it is a good leap in hardware support.

BTW I did get dosemu running. It's ok for very basic console applications. I'm not sure I'd try it for games. Also, I seem to get some odd problems with the keyboard all over the place. Things don't work like they do now, so maybe just a little rudimentary.

This definitely has the feel of a necro thread, even though it isn't.

I remember the early years and Dosemu. It always looked like it was going to work, said encouraging things, then died.

Dosbox, OTOH, was a piece of sh** originally. But that changed soon enough.

The trouble was that everything lied to Dos from day 2 onwards. But the lies were all a bit different. Emulating that on linux sure was tricky.

That was what I was going for. Writing as if had been done it back then.

Despite the flaws it has, I guess the thing is I still knew exactly what to do to get things running, as not a whole lot has changed. All the basic tools worked for me as I can still do with them today. And that's a good thing. It's only gotten better. The simplicity of just a basic environment in such an early version is not a big deal and nicer than DOS. I have built and used LFS for a year before I switched to slackware, so I'm used to small shell environments and minimal X. So for a 486 it's decent. However, I think the subsequent releases will be a bit better choice covering more 486 and even pentium era hardware.

People were doing exactly what they are doing now. Stuff is compiled & tested on machinery of the day. When they got fed up debugging it, they released it. Then they finished debugging it.

But back then there was a much greater variety of crappy hardware. M$ wrote emm386.exe which was loaded in config.sys. It's trick was to make extended memory pretend to be expanded memory. Expanded memory was an ISA card which needed I/O between 640K and 1024k and a precious interrupt. But it worked. Extended memory was up there above 1024k and couldn't be accessed, because config.sys also loaded an A20 handler. So extended memory didn't work without emm386.exe.

In those days I was maintaining pcs for Crown, Berger and a few others. I did it for one month, put up thousands of miles, and wore myself to a thread. So I made utility floppies for the mechanical guys and let them do the miles. There were no decent roads here back then. They would turn up, ring me, I'd tell them what to do.

I guess I keep thinking more about what could be done with my old machine!

About when was this? I wish it was that easy! The setserial and slattach does not come in Slackware 1.01. I can compile in SLIP and it will acknowledge the interface, but I can't get it to communicate IP with a modern system this way just yet. I tried dip and kermit to try to coax it. I didn't have to worry about making a cable. I bought a cross-over adapter years ago to play games locally before I had a NIC for this, so that part is done. I can connect the old machine to the new one and get a terminal through the direct connection, so I know I should be close on getting IP packets if that is working.

So I was poking around the kernel sources, and noticed my NIC, 3c509, is in 0.99pl12! But doesn't Slackware 1.01 come with 0.99pl12? Yes, but it's an alpha version, where it didn't have the source file yet for it. So it's just a matter of getting a copy the release version of 0.99pl12 copied over. I could boot to DOS which has the NIC driver and do it that way, but that's too easy. Let's do it the old Linux way! I decided to do PLIP, since that seemed easier. (btw I tried PLIP before SLIP when came time of this writing) This version of kernel came out Aug 14, 1993; so 30 years ago today. What difference a few days made! I might have a full ethernet working given how fast things went back then, so not bad for Slackware 1.01. To get started, I gave my boy a bunch of really different but all cables with db25 ends (no adapters to be used) and asked him to find and provably show which for sure is a laplink (parallel crossover). Funny to think about, but it was something to learn some multimeter use and how things really work.

Unfortunately, the PLIP in 0.99pl12alpha is broken. Some person decided to change how the protocol worked, and didn't really get it stable. Reading through the history of the plip.c source on kernel.org explains it. Basically anything before about kernel version 1.1.4 won't work with anything later. That's when someone starting fixing it back to what it was supposed to be. Since I've written code for the parallel port for communicating before, I thought might as well just fix it on the machine. So having a recent copy of plip.c on a screen to where I wanted to connect, I walking through the other fixing it. I got it mostly stable. I did have to insert a sleep and slow things down because of some kind of "collision" issue, I don't know why that ever was a thing.

I got the newer kernel source, and found the 3c509.c was hidden away, not publicly visible in make config. I guess they considered this driver alpha quality in 0.99pl12. Well guess what? It doesn't work. I guess it was too much to ask things to be perfect, but I guess that is how it was, you tried, and fixed things if you could. 0.99pl13 has some more changes for the driver, so I copied that over. It does work on that! But a bit unstable per my initial testing. I was about to give up, since 0.99pl14 comes on the next version of Slackware 1.1.1 I have, so no sense to just try the kernel, when I could upgrade. But I noticed on the git repo at kernel.org there is a 0.99pl13k, and it shows that the 3c509 driver was made public and can be selected in make config. Well there it works pretty stable! Awesome!

So the network code is kind of quirky, unless maybe you used the older, more common NE2000, or did dialup like many people did then, as they could be the more tested options by that point. I guess getting the network code right must have caused some of those first flamings I'm not sure if I care about doing any more serial, as I figure using kermit is enough if I had to do something related to that interface.

Oh, and there isn't really a memory leak in this version of Slackware. I was wrong about that. I misread the 'free' command, as the command output was missing the line that subtracts out the 'buffer' usage. I also see that the sound driver does detect my sound card! I can now get some programs over and use this OS more. This was pretty much as close to a complete OS to get up and running as it were for the time. I don't have the IDE CD driver though. I think that was still around a year out maybe??? So maybe not the perfect OS for this machine, but it is still impressive.

An experimental OS! Sorry for the missing SLIP/PPP support (didn't remember well), for networking an NE2000 nic is the way to go. There's also a special device that you connect to the // port to get connected (forgot the name) but maybe unsupported
My last test of an old Slack (kernel 0.99pl..15?) was on a 486 mobo (SiS chipset) but install on a flash disk connected directly on the first ide port You can buy such drives on the Bay. Install is faster than on an old Conner drive

We did plenty with them back in the day.It would hack it as a Space Quest arcade station, if it met the min requirements. i586 was a cutoff point for a lot of stuff, because the instruction set gave extra legs to the code.