A year-end wrap-up from LWN

2019 is coming to a close. It has been another busy year with a lot going on in the Linux and free-software communities. Here at LWN, we have a longstanding tradition of looking back at the predictions made in January to see just how badly we did; it's not good to go against tradition no matter how embarrassing the results might be, so we might as well get right into it.

Visionary?

The 50th anniversary of Unix happened just as predicted; your editor is looking like a true visionary so far. That prediction also suggested that we might see "interesting work in alternative operating-system models" this year. Whether the development of systems like Fuchsia or seL4 qualifies is a matter of perspective. One could also observe, as Toke Høiland-Jørgensen recently did, that "the Linux kernel continues its march towards becoming a BPF runtime-powered microkernel" and conclude that the most viable alternative to the Unix model is developing right under our noses.

The prediction that there would be more hardware vulnerabilities was no less obvious back in January. Holes like MDS, SWAPGS, and TSX async abort duly put in an appearance. It seems unlikely that we are done at this point. A minor consolation might be found in the fact that, by most accounts, communications between the kernel community and hardware vendors regarding these vulnerabilities have improved as predicted.

Did kernel development become more formalized, as we thought might happen in January? Certainly there have been discussions around workflow issues and Change IDs that would point in that direction, as does the increased emphasis on automated testing. One might argue that the kernel community grows up far too slowly, but things do change over time. The suggestion that projects would continue to transition away from the patches-over-email model ties into this as well; even the kernel community is talking about it, though any such change still seems distant at the end of 2019.

Issues with the supportability of BPF APIs did arise as predicted, but the statement that "more kernel APIs will be created for BPF programs rather than exported as traditional system calls" has not been fully borne out. That doesn't mean that we aren't seeing interesting APIs being created for BPF; for example, it may soon be possible to write TCP congestion-control algorithms as BPF programs.

Did somebody try to test out the kernel's code-of-conduct as predicted? As of November 30, there had been no code-of-conduct events in the last three months, and only minor events before. That prediction, happily, has not worked out. Thus far, it seems that the code of conduct may actually have succeeded in making the kernel community a nicer place without the need for any serious enforcement efforts.

Whether we are seeing an increase in differentiation between distributions as predicted is unclear. There is clearly a growing divide between those that support systemd and those that do not; the Debian project is trying to decide where it fits on that divide as this is written. Fedora is working to prepare for the future with initiatives like Silverblue and Modularity. All distributors are trying to figure out how they fit in with the increasing popularity of language-specific package repositories — a concern that probably feels more pressing than differentiation from other distributions.

Your editor predicted that there would be more high-profile acquisitions of Linux companies in 2019; inspired by the purchase of Red Hat by IBM, the article also suggested that Canonical might finally be sold. Not looking quite so visionary now.

The Python community did indeed complete its transition to a post-Guido governance model, and recently held its second steering-council election. Is the Python 3 transition a memory as predicted? Perhaps so; certainly there are fewer discussions on the topic than there once were. For many, though, it remains a relatively vivid and unpleasant memory.

There are definitely groups out there trying to come up with new licensing models, as predicted. In January it seemed like these efforts would mostly be driven by companies trying to monetize their projects, but the emphasis appears to have shifted to attempts to drive other agendas. Thus we saw the Twente License that requires observance of human rights, the Cryptographic Autonomy License with its prohibition against locking up user data, the Vaccine License making rights available only to the vaccinated, and the CasperLabs Open Source License, which adds all kinds of complexity for unclear reasons. As a general rule, these are not truly open-source licenses and they are thus not going far, regardless of whether one agrees with their objectives.

The crypto wars have not yet returned as predicted, but there are a number of chilly breezes suggesting that the right to use strong encryption may yet come under serious threat.

The web-browser monopoly may not have gotten much worse over the past year, but that is mostly because there isn't room for things to get much worse. Chrome dominates the market, with other browsers relegated to single-digit-percentage usage shares. That causes site developers to not care about making sites work with anything but Chrome (and maybe Safari), forcing even dedicated users of other browsers to launch Chrome to make specific sites work. To those of us who lived through the period of Internet Explorer dominance, much of this looks discouragingly familiar.

The final prediction worried that free software has increasingly become a way for companies to develop software efficiently while depriving competitors of license revenues. Certainly there are companies that see free software that way. That is not an entirely bad thing; a lot of new software and a lot of development jobs result from companies working from this viewpoint. But there is more to free software than that, and many members of our community continue to work toward a vision of a world that is more free and more secure. We can only try to support those efforts, and that vision, as well as we can.

Events not foreseen

The other side of evaluating predictions is looking at what was missed. One obvious omission was the leadership transition at the Free Software Foundation as the result of Richard Stallman being forced out. In retrospect, it seems clear that a change had to happen at some point, but saying when it might occur is always hard ahead of the actual event.

On the kernel front, the many-year effort to get lockdown capability into the kernel finally came to fruition; that is another event that had to happen sometime, but your editor didn't expect it in 2019. Even more unpredictable was the pidfd API, which seemingly came into existence, fully formed, after the beginning of the year, though rumblings were certainly evident before then.

Another surprise was the openSUSE project's decision to separate from SUSE and form a separate foundation. This move is being made partly to make it easier for openSUSE to seek support from multiple sources, but the project still is likely to be dependent on SUSE for some time — perhaps indefinitely. The process of negotiating the project's future relationship — and the use of the openSUSE name, which the project elected to retain — with its former corporate owner is likely to be complex. One can only wish openSUSE well as it charts its course going forward.

Closing another year

In 2019, the LWN crew produced 50 Weekly Editions containing 266 feature articles and 56 articles from 16 guest authors. We reported from 26 conferences hosted on four continents — something that was made possible by our ongoing travel support from the Linux Foundation. It has been a satisfying year, but it's fair to say that we are ready for a break.

We cannot sign off, though, without acknowledging the force that keeps this whole operation going: you, our readers. Advertising revenue hit a new low this year but, thanks to you, we have not been overly dependent on advertising for many years. Subscriber numbers are down a bit since last year, which can mostly be attributed to some of the people who came in for the Meltdown/Spectre coverage opting not to renew. With a slightly longer perspective, it is clear that our base of support remains solid, and for that we are extremely grateful. We wouldn't be here without you.

On that note, we'll sign off for the year; we look forward to resuming next year with the January 2 Weekly Edition. As always, the lights will not go completely out between now and then; be sure to check in for the occasional article and update. Meanwhile, we wish a great holiday season for all of our readers.

---

to post comments