Owen Le Blanc: creator of the first Linux distribution | Hacker News Reader

Owen Le Blanc: creator of the first Linux distribution

lwn.net

161 pointssohkamyung34 comments

Some more context from a former colleague: https://techrights.org/n/2025/05/02/Manchester_Computing_Cen...

MCC Interim Linux wikipedia page notes it started out with Linux kernel 0.12 https://en.wikipedia.org/wiki/MCC_Interim_Linux

https://www.kernel.org/pub/linux/kernel/Historic/old-version...

It makes me want to play, configure, compile, tidy and optimize! https://github.com/ESP32DE/Boot-Linux-ESP32S3-Playground

Owen used to organise the Manchester Linux User Group at the MCC as well, I fondly remember those early days when I was learning Linux. Looking back it was an amazing privilege to connect with some extremely knowledgeable people in the Linux ecosystem.

Yeh a few ManLUGers still get together for a Jitsi call about once a month; not many these days.

I still remember Owen showing me Linux (I was a Ph.D. student in the graphics lab at MCC, so this was probably around 92-93). He's such a nice guy.

I had no idea he had such a claim to fame....though I suspect he didn't either!

What a glorious piece of history. I wonder what other "scratching my itch" solutions became so mainstream that people forgot about the original authors.

I think all of todays popular Linux distros, Debian, Gentoo, Fedora, Arch, SUSE and so on, are all very much "scratching my itch" projects that somehow managed to outlive the original authors engagement with the project.

It's not like any of them where planning to be used by millions of people.

Fedora wasn't like that, it was spun out of Red Hat when they went enterprise only with RHEL.

Yes and no. I realise that to younger members of the Linux community they're all from long ago, but they're not the same age.

There aren't really clear generations in Linux distros, but as an approximation:

Debian is pretty old, but it's a 2nd gen distro, borne from dissatisfaction with the very early SLS.

So was Slackware, but it took SLS and improved it. Slackware is arguably the oldest surviving distro.

SuSE has roots as a German version of Slackware. Red Hat's package manager was bolted on later.

Gentoo and Arch are relatively modern, being 21st century projects. Arguably, they're 3rd gen.

Fedora is a 4th gen distro, younger than any of the others here. Its ancestor was Red Hat Linux, which was contemporaneous with Debian -- but was left behind by Debian's technical encancements: in 1996 or so, Debian introduced `apt`, a package manager with automatic recursive dependency resolution. This put it far in the lead of Red Hat, which still only had RPM and no dependency resolution.

Red Hat went in another direction. Red Hat Linux 7 became RHEL, a commercial, paid-for, supported distro.

The free RHL went on for 2 more versions, reaching Red Hat Linux 9, which then became Fedora Core, version 1 of the free unsupported community distro.

RHL was killed off after v9.

As I understood the story as an analyst at the time, Red Hat’s intention was to just kill RHL after a decent interval but there was sufficient outcry that they came out with Fedora.

But I’m sure there are many different recollections and variants of the Fedora was planned all along story told over the years that the “truth” is probably pretty elusive at this point.

> Debian is pretty old, but it's a 2nd gen distro, borne from dissatisfaction with the very early SLS.

Scratches their own itch, check.

> So was Slackware, but it took SLS and improved it. Slackware is arguably the oldest surviving distro.

Itch scratching, check.

>SuSE has roots as a German version of Slackware. Red Hat's package manager was bolted on later.

Pretty sure this was itch scratching as well.

> Gentoo and Arch are relatively modern, being 21st century projects. Arguably, they're 3rd gen.

Both are itch scratching projects!

> Fedora is a 4th gen distro, younger than any of the others here. Its ancestor was Red Hat Linux, which was contemporaneous with Debian -- but was left behind by Debian's technical encancements: in 1996 or so, Debian introduced `apt`, a package manager with automatic recursive dependency resolution. This put it far in the lead of Red Hat, which still only had RPM and no dependency resolution.

Arch and Gentoo are from 2002, and Fedora from 2003.

Fedora was based on someone starting to package FOSS software for RHEL, more itch scratching!

what about nixOS ? third gen as well ?

In this kind of hierarchy, I'd personally say fedora is third gen (due to it being so similar to Redhat) and that nixOS is fourth gen. Both came out around the same time, but took such vastly different routes with different kinds of itch scratching.

Yeah, that sounds about right.

Gentoo and Arch are different takes on the same ways to build a distro.

Gentoo took the FreeBSD ports tree model and applied it to Linux: still relatively conventional packages, but they're source and you compile the whole thing each time. Arch, still conventional packages, but no fixed release cycle.

Nix throws all that, and the directory tree, out.

Slackware: tarballs are good enough, they're all my dad and grandad ever needed.

RHL: we'll have a package format, where packages can depend on others.

Debian: we'll take the RH idea, but make a tool that can go fetch and install what's needed by what you asked for.

Nix: you don't need to worry about stuff like packages or where stuff is. Tell us what you want and we will make it happen. (But you won't like the disk layout, so don't look.)

Guix: we'll do Nix but with Scheme.

AppImage: hey, you know Acorn did that apps-as-bundles thing first? And it's on Linux as ROX? What if we just zip up the bundles and mount them on demand?

Flatpak: that sounds too hard, dude. But we all agree Git's cool, right? So, what if we could, like, distribute apps over Git?

Gobo: all this packaging and dependency stuff is BS because you're still using a disk layout you improvised on the fly on some 1960s minicomputer with like 20 tiny little hard disks. Here, let's do a clean modern layout like NeXTstep did, but for the whole OS, then you don't really need a packaging tool any more. It's all just bundles, all the way down. But they're versioned. Just copy what you need.

(Entire rest of Linux world) Waah! But mah FHS! If I don't have my FHS then I won't compile!

Gobo: OK, OK, I'll fake it with symlinks for you, then hide it.

Snap: hey, app bundles sounds good, but let's compress them into single files and mount them when needed. Bunch of symlinks and you'll hardly be able to see the joins.

don't list the mount points with mount in any system using snap. You would not like how many garbage are there hiding the real stuff.

findmnt --real

Fair point. It's annoying but it's a small price to pay.

I have machines around here running umpteen distros and OSes, but recently, I switched all my Ubuntu machines to more or less all-snap. I've purged and removed a load of 3rd party repos and replaced them with snap-packaged versions. Even on 12-13 year old machines, it's quick these days, and it makes all kinds of problems on distro upgrades and things just go away.

> Red Hat went in another direction. Red Hat Linux 7 became RHEL, a commercial, paid-for, supported distro.

This was the second time we had a Red Hat 7, though.

Going to have to disagree with the chronology. Fedora and RHL isn't a substantial divergence and it belongs with SuSE based on its 1995 roots.

Fedora DID diverge substantially but not when they slapped a new label on the same tin.

Arguably Ubuntu/Mint Make up a 4th generation and Nix/Qubes/and Silverblue are 5th

Slackware 2.0 in 1995's Summer was my first distro, it came with the first edition of Linux Unleashed book.

For the younger in the audience that is how we distributed software back in the day, floppies, CDs and DVDs, alongside books and computer magazines.

Very few people had the means to go online, let alone connections able to download a complete distribution.

Has the distribution model been good for Linux? It led to different approaches to things like desktop environments, packaging, and a variety of platforms, but 30+ years later, there are several sane choices for server distros, desktop distros are even more fragmented, and the most popular user distros are Android and ChromeOS.

I think so, because the users seem to like having different options. For commercial software, it makes sense to count how many devices use a particular distribution as the measure of “success”, but for projects like most Linux distributions , I don’t know that number of users makes sense. Why should we care how many users a particular distribution has, when almost all of them aren’t paying or contributing? Having more users doesn’t make the software any better inherently, and nobody is making money from those users. Instead, I would argue that user enthusiasm and dev interest are better measures of success for open source projects like this, and arch, Debian, Linux mint, etc are all doing fine in those regards.

It is tempting to consider the replacement for a lively evolving market of good options is just the best option only better for all the labor focused on a singular end but in fact the likely result of a singular option is most of that labor that literally only exists because of the interest in doing ones own thing just ceases to exists and is lost.

What persists never had the privilege of benefiting from ideas taken from all those other now non-existent projects and is on the whole mediocre.

The packaging model of distribution is ubiquitous. Every distribution does the same thing just using different control files and tools. The differentiation between distributions is all in their packaging policies and platform decisions. In a loose way Unix (SysV, HP, AIX) started packaging before Linux but Linux said "Every project is its own package" and ran with it. The de-facto "Download release; apply patches; configure; make; make install; collect files" is present in every distribution package. Everything up through deployment is the same pattern across all distributions.

The amount of software available for linux vs the BSDs tells me that the distro model has not hurt linux. If a homogenous software stack from a single centralised set of software was beneficial, it would be more likely that porting to linux from a BSD codebase would be the norm, rather than the other way around.

This really brings back memories of how painful installing any software in the early 90s was. The small company I worked for got us a Yggdrasil CD to try but we were unable to get it installed on any of the PCs we had at the time. MCC might have done better, but we hadn't heard of it.

Was Yggdrasil that bad? My first distro was Slackware and, with the help of the book accompanying the CD, it was doable. Sure you had to define modelines for X11 (the Xorg name didn't exist bad then) to support your monitor and supporting GPUs was quite the endeavour, but in the end we'd make it work. We'd even compile and run Emacs (in 45 minutes or so).

It was called XFree86

I don't remember what the first Linux distribution I used was, but it a set of floppy disks I downloaded from a local BBS.

I somehow got it to boot up but didn't really know what to do with it after that.

Could very well have been Slackware. Slackware was my first Linux distribution, it came as a set of like at least 20 floppies. All of mine were repurposed AOL disks. After spending about a solid week or so downloading the whole set of disk images over a slow and intermittent dialup connection, the next most painful thing was the fact that floppies were notoriously unreliable. Some disks would throw I/O errors when writing. Some would get caught immediately after when verifying. Many others showed no problems until install time. Getting two dozen floppies to actually read 100% of their contents successfully took a week or two on its own because I only had one computer to work with.

I very clearly remember my very first version of Slackware -- pre 3.0.0 (which I actually bought on cd for a few bucks). I don't remember that first version, just that I downloaded the floppy disk sets over zmodem at 14.4kbps (thankfully saving to hard disk, not to floppy).

That first version of Slackware I used had the Linux kernel 1.2.8; IIRC that series went to 1.2.13 before going through the a.out->ELF transition.

Anyway, original point, that Slackware distro of 1.2.8 had a bug where every single time I had to reinstall the bootloader for a newly-compiled Linux kernel (which I had to do regularly), LILO was broken and hung at the `LI` prompt... those who were there may remember, the number of letters of LILO: that were output gave a sign to the source of the error.

But every single time, I had to rescue boot, and try to remember what I had to fix to make LILO work again.

The comments section on the article is nice, lots of people's memory's of MCC Interim Linux and Owen.

So first linux distribution was this one Feb 1992.

And first linux distribution with a GUI was "TAMU linux", 3 months later: https://lwn.net/Articles/91371/

Both were released by universities

FWIW, I think I was the first person to ever produce a unix-like distribution for the Atari ST/TT line in 1992. It installed MiNT (MiNT is Not Tos) on a free partition, with a minix filesystem and various optional disks (including make, ash, gcc, etc.)

The install-help docs were written using Calligrapher, an application I still think was way ahead of its time on the ST. There are postscript docs as well as ASCII ones at the link below.

[1] https://websites.umich.edu/~archive/atari/Mint/Distrib_kit/D...

Ooh great to see this pop up on the HN front page - I have great memories of working with Owen at UoM :)

Previously: https://news.ycombinator.com/item?id=43782975 (no comments)

Reply on news.ycombinator.com