How and why Linux has thrived after three decades in Kernelland

'Just a hobby, won't be big and professional like GNU...'

Open Source Summit At OSS EU, LWN editor and long-time kernel developer Jonathan Corbet shared a long-term perspective on how and why Linux has thrived for a third of a century.

Corbet's talk offered a rare, nearly unique, insight into the rise and rise of Linux over 30 years. There aren't that many developers who've been involved for so much of that time, and of those, few are good communicators who are both able to talk about and as well as actually wanting to do so. It's quite well known that Linus Torvalds himself doesn't enjoy public speaking.

Youtube Video

We've written about Corbet's annual kernel talks at the Open Source Summit a few times before on The Register, including the 2022 installment as Rust was allowed in, and the 2023 one when he talked about developer burnout and reducing the long term support lifespans for the kernel. This year's talk was a little different, being more of a retrospective.

Torvalds' original 1991 announcement said that his kernel project was "just a hobby, won't be big and professional like GNU" and Corbet opened by contradicting that:

So I think that we can conclude that for perhaps the only time ever, Linus Torvalds was wrong, and Linux is big and professional.

As he laid out, it really did not look like that in the early days. This vulture was there, watching with interest, although I didn't try to install the thing myself until about 1995.

The beginning of the Linux era was really the end of the Unix era, where it seemed like Unix systems proprietary Unix in particular, was going to be the operating system in the future. Unix had a good run, but the Unix vendors each made their own version and kind of made it buggy and awful in their own way, and really tore that market to pieces.

Meanwhile, Microsoft was on the rise. It was very common to hear in the early 1990s that everything was gonna be Windows, right, not just desktops but servers, mobile systems, whatever: it was all going to be Windows. That was just kind of the accepted wisdom.

He credits this to the open development model, as legitimized by Netscape open sourcing its browser.

Now, Eric Raymond, for all of his faults, put his finger on something here, which is that the way that Linux was developed was a bit different than anything that came before it — either proprietary software, or the free software projects at the time. Look at the GNU project, the X Consortium, the BSDs, all these projects were run in a very centralized sort of way, with a small group of people who could really work with the software and make decisions and who would every now and then deign to toss out a release that everybody else could use.

Linux instead was open to everybody. Linux would take patches from anybody. There were no boundaries.

He proceeded to point out that despite many significant world events over this time, none of them noticeably impacted the pace of development of the Linux kernel. He called out the launch of VA Linux and its meteoric early success as early signs of the dotcom boom. Just a couple of years later in 2001, El Reg regularly reported on dotcoms dying: the Industry Standard, Daily Radar, The Times Interface, The Net, and so on…

Most of these early Linux companies either went under, or emerged transformed into much more modest businesses… But the kernel kept growing.

Later in the noughties, there was the famous SCO versus IBM lawsuit, which he summarized splendidly:

If we're gonna talk about the fight in this era, of course, we have to talk about the big one, which was the SCO lawsuit. This guy named Darl McBride put on his cowboy boots and stood up in front of the world and said that Linux is full of millions of lines of code that were stolen, stolen from him and his company in particular, and distributed illegally and that as a result of that, IBM owed him $5 billion. IBM disagreed.

In fact, in the longer run, the lawsuit helped Linux, by legitimizing it. As Corbet put it:

It was scary for a while. We really did not know how this was going to play out, and of course, it played out very well – in fact… so well that it made us a whole lot stronger than we were when we found it for a few reasons. As I like to say, we had the good luck of being attacked by idiots.

Harsh, but fair.

It wasn't a technological crisis, but by 2008, things were looking bad once again, thanks to the Credit Crunch. Another dozen years later, it was COVID-19, and yet kernel 5.6 came out like normal.

Corbet talked about the role of corporate investment, and the possibly unfortunate necessity of having lawyers, as highlighted by some early Linux hardware, such as the Axis network camera and the WRT54g router, which The Reg covered in 2003. The WRTD54g used Linux, but Linksys didn't release the source code until pressured. The result was not only a victory, but better Linux networking.

He discusses the adoption of Larry McVoy's Bitkeeper source-code management system, the struggles with its proprietary terms, and the controversial moves that led to the creation of the aptly named Git.

The talk covers the importance of real-life in-person meetups for the developer community's health, and how US politics are making that more difficult. It covers the importance of the GPL over other FOSS licenses, and that of the kernel-versus-userland split, and the distribution model. In a section that will probably rub some people up the wrong way, he questions whether C is still the best tool, and mentions the importance of Rust.

The talk covers a lot of ground, and we don't want to write our précis of it. It's worth a watch, and we find Corbet's presentation style a welcome contrast to the smooth, polished, corporate keynotes that are a common feature of many technical conferences these days. ®