08.09.20

Features Considered Harmful (Revised)

Posted in GNU/Linux at 7:52 am by Dr. Roy Schestowitz

2020 figosdev

• Chapter 1: Start With Perfection
• Chapter 2: Controlling Your Computing
• Chapter 3: Miseducation
• Chapter 4: Computing vs. Marketing
• Chapter 5: The Psychology of Developers
• Chapter 6: Freedom is Personal
• Chapter 7: Computing Fundamentals
• Chapter 8: (Don't Let's) Throw Caution to the Wind
• Chapter 9: For the Want of a Pixel
• Chapter 10: The Computer Anybody Can Edit
• Chapter 11: Rebuilding Communities
• Chapter 12: The THRIVE Guidelines
• Chapter 13: I Would Have Supported the Coup (Under Very Different Circumstances)
• Chapter 14: If We Weren't Silencing Founders, Critics and People We Just Don't Like
• You are here ☞ Chapter 15: Features Considered Harmful

Chapter 15: Features Considered Harmful

Summary: “But the benefits of Free software, free candy and new features are all meaningless, if the user isn’t in control.”

Edsgar Dijkstra is famous for hating on BASIC programming. But it was his editor (Wirth, I believe) who incited decades of clickbait titles by working the infamous line attributed to Dijkstra into the top of his article.

“Basic itself isn’t harmful. Some of its features may lead to worse programming, but I think Torvalds (as a coder) proves that using it doesn’t necessarily prevent you from gaining good programming skills.”His stance was reasonable enough, so long as we are talking about the actual arguments that were made — which applied more to the original line-numbered versions of Basic than today’s versions, which look like Pascal by comparison. It was good enough as a first language for Linus Torvalds. Dijkstra actually had a number of great ideas that BASIC once lacked. And in more ways than one, it doesn’t bother me that software evolves.

Whether it was his intention to save Basic or get people to use a better language instead, Basic itself isn’t harmful. Some of its features may lead to worse programming, but I think Torvalds (as a coder) proves that using it doesn’t necessarily prevent you from gaining good programming skills. Some languages indeed will teach better discipline, but if you’re determined to be lazy, you’ll probably find (or create) tools that suit your preferences.

I am not against features, per se. I’m also against prohibition and the drug war. But the great harm done by heroin cannot be dismissed — it kills people, and although there is no law that says you must try heroin, getting away from it isn’t always as simple as “just don’t use it”. Fortunately, while I did find Basic somewhat addictive in practice, I have not tried heroin. Though I’ve certainly lived in places where it was a problem. In fact, it’s a problem that tends to increase when the alternatives are fought harder against.

GitHub could arguably be the heroin of the Free software world. We know the harm it does, we’ve been warned about it for years, there’s absolutely no mandate to use it at all — yet people keep finding themselves addicted to it. GitHub isn’t known so much for killing people, but it poses a great threat to projects that use it. What GitHub actually kills, is software freedom.

I’ve written a lot about GitHub lately, but in this article it is just one example of a larger problem. Like with Basic, it is not “GitHub” itself, but some of its features that we should worry about. And the warnings against it have come from Torvalds and Stallman alike.

The complaints from Torvalds against GitHub are closer to Dijkstra’s complaints about Basic — GitHub encourages bad practices in Git management, and breaks existing features. It trains you to be a worse Git user. I think this is a minor problem next to the others. But just as Dijkstra is a pioneer of structured programming, Torvalds is the original author of Git. That makes the critique much more notable.

The fact that GitHub breaks Git the way that it does, fits in with a larger complaint of my own — even if Torvalds decides (or is paid to) change his mind about it. It was developed by Chris Wanstrath, but it was developed along lines that are not entirely different from Microsoft’s EEE tactics — which today I will offer a new acronym and description for:

1. Steal
2. Add Bloat
3. Original Trashed

It’s difficult conceptually to “steal” Free software, because it (sort of, effectively) belongs to everyone. It’s not always Public Domain — copyleft is meant to prevent that. The only way you can “steal” free software is by taking it from everyone and restricting it again. That’s like “stealing” the ocean or the sky, and putting it somewhere that people can’t get to it. But this is what non-free software does. (You could also simply go against the license terms, but I doubt Stallman would go for the word “stealing” or “theft” as a first choice to describe non-compliance).

I came up with this SABOTage acronym when I was going to sleep, and originally it was Steal, Add Bloat, Attack — I guess spelling isn’t a strong point when I’m tired. But this is what people do even in the Free software world today; they take away compatibility (as GitHub did with some Git features Torvalds thinks should work properly), they add stuff that is easier for a large corporation to host (Gitlab too, is terribly bloated I’m afraid — but it can be self-hosted at least) and they attack the original — by dragging everyone into GitHub (as it’s “better”).

I really do understand the appeal of GitHub — I’m a former user as well. While the complaints of Torvalds are relevant to this discussion, the complaints by Richard Stallman are more important to me. In 2015, he said to GNUstep developers:

“GitHub does things that are quite bad for free software and is not interested in changing them. If you want to move off Savannah, please pick some other place.”

This led to a shallow debate on the merits of GitHub vs. alternatives, and Stallman argued that GitHub negatively affects the license choices people make. One person replied that he was having an unrealistic expectation of GitHub, but this was the thing — we can make it about the design of GitHub, or we can look at the effects. In effect, GitHub successfully gets people away from making choices that are good for free software.

Expecting a platform for free software hosting (if that’s what you intend to use it for) to not be designed in a way that harms free software is a completely reasonable expectation! Particularly when the response to it not being suited to free software’s goals is “Don’t use this, it’s bad for what we do”.

It also includes non-free Javascript, which many people are willing to forgive or overlook sometimes. But this discussion was about code hosting for the GNU Project itself! If any project should not rely on GitHub and non-free Javascript, it’s the flagship project of Free Software itself.

“But it still works if you turn Javascript off” they said… again, this is one situation where many of us are willing to overlook non-free Javascript: if the website still works when you disable it. This is still wildly inappropriate for the GNU Project to endorse, because they’re still encouraging users to run it.

Other than the fact that running and promoting only Free software (yes, I’m familiar with Stallman’s latest article on the topic) is one of the goals of the GNU Project, I think there are worse things about GitHub. And I think that GNU projects that continue to use it unapologetically, such as GNU Radio and GNUstep, are proving that their developers DON’T care about your freedom, and do not represent (nor achieve) the goals of the GNU Project.

And the fact that Microsoft has spent decades trying to co-opt and control Free software? Who honestly cares about that?

But I am aware that such Microsoft-neutral or Pro-Microsoft developers are not necessarily agreed with by every developer on these projects. In the instance of GNUstep, we are talking about the leader of the project who doesn’t care about your freedom.

I did say that GitHub was just an example; it’s a very big example, though not the only one. Microsoft is taking over Python as well. And the way it’s taking over Python does have the aim (and the success) in dragging it into the GitHub trap, because GitHub is perfect for that sort of thing — but the tactics being used would hurt Python with or without GitHub as well. It conquers projects the same way as empires conquer nations — by planting flags in whatever they want to own:

This Techrights article is from 2010, and though they have done this farther back than that, and continue to do the same, Microsoft is still planting flags all over the place. It wants to run your conferences. It wants to host your code. It wants you to agree to its terms. It wants you to adjust your development to its heavily contrived, self-serving “standards” from OOXML all the way back to Rich Text Format.

Again and again, Microsoft “Steals” or “Steers” the development process itself so it can gain control (pronounced: “ownership”) of the software. It is a gradual process, where Microsoft has more and more influence until they dominate the project and with it, the user. This is similar to the process where cults (or drug addiction) take over people’s lives, and similar to the process where narcissists interfere in the lives of others — by staking a claim and gradually dominating the person or project.

Then they Add Bloat — more features. GitHub is friendly to use, you don’t have to care about how Git works to use it (this is true of many GitHub clones as well, as even I do not really care how Git works very much. It took a long time for someone to even drag me towards GitHub for code hosting, until they were acquired and I stopped using it).

Due to its GLOBAL size, nobody can (or ought to) reproduce Microsoft’s network effects.

I understand the draw of network effects. That’s why larger federated instances of code hosts are going to be more popular than smaller instances. We really need a mix — smaller instances to be easy to host and autonomous, larger instances to draw people away from even more gigantic code silos. We can’t get away from network effects (just like the War on Drugs will never work) but we can make them easier and less troublesome (or safer) to deal with.

Finally, the Original is trashed, and the SABOTage is complete. This has happened with Python against Python 2, despite protests from seasoned and professional developers, it was deliberately attempted with Systemd against not just sysvinit but ALL alternatives — Free software acts like proprietary software when it treats the existence of alternatives as a problem to be solved. I personally never trust a project with developers as arrogant as that.

It’s difficult for me to get excited about these “next generation” tools, when I spent several years working to GET AWAY from Microsoft, and they want me to get all my software from GitHub. If I wanted to get all my software from Microsoft and the rest of GIAFAM, I’d just use Windows.

And speaking of, the coup continues this week, with the new COO at Microzilla: Adam Seligman — “formerly of Google, Salesforce, and Microsoft.”

GREAT! That’s also how they gradually took over Nokia, Apache Software Foundation and became the boss of Linus. [Roy added to this: Even the COO of GitHub now bosses Linus]

Corporate takeover, disguised as an upgrade.

Here’s something else to consider — the way that websites subtly (and sometimes innocently) add to the problem with handy Share icons (which I’m not entirely against). A colleague informs me that one of the things that draws people to GitHub is the way that other websites make it easier to integrate with it. I can’t fault his logic, he’s right. But here’s the reality of that. Such tie-in features will always be implemented for the largest option first, and typically the largest option only.

They’re not going to bother reinforcing smaller choices usually, they’re going to reinforce the largest one. So this practice itself — while technically and theoretically neutral (as it could offer several options for code repos) actually encourages monopoly in practice most of the time. These are network effects, compounded.

There’s a meme about creepy vans with “FREE CANDY” painted on the side, which I took one of the photos from and edited it so that it said “FEATURES” instead. This is more or less how I feel about new features in general, given my experience with their abuse in development, marketing and the takeover of formerly good software projects.

People then accuse me of being against features, of course. As with the Dijkstra article, the real problem isn’t Basic itself. The problem isn’t features per se (though they do play a very key role in this problem) and I’m not really against features — or candy, for that matter.

I’m against these things being used as bait, to entrap people in an unpleasant situation that makes escape difficult. You know, “lock-in”. Don’t get in the van — don’t even go NEAR the van.

Candy is nice, and some features are nice too. But we would all be better off if we could get the candy safely, and delete the creepy horrible van that comes with it. That’s true whether the creepy van is GitHub, or surveillance by GIAFAM, or a Leviathan “init” system, or just breaking decades of perfectly good Python code, to try to force people to develop differently because Google or Microsoft (who both have had heavy influence over newer Python development) want to try to force you to — all while using “free” software.

If all that makes free software “free” is the license — (yes, it’s the primary and key part, it’s a necessary ingredient) then putting “free” software on GitHub shouldn’t be a problem, right? Not if you’re running LibreJS, at least.

In practice, software that is “Free in license only” sidesteps the implication of effectively free software, which is that its users are also effectively free. If free software development gets dragged into doing the bidding of non-free software companies and starts creating lock-in for the user, even if its external or peripheral, then they simply found an effective way around the true goal of the license. They did it with Tivoisation, so we know that its possible. They’ve done this in a number of ways, and they’re doing it now.

If people are trying to make the user less free, and they’re effectively making the user less free, maybe the license isn’t an effective monolithic solution. The cost of freedom is eternal vigilance. They never said “The cost of freedom is slapping a free license on things”, as far as I know. This really isn’t a straw man, so much as a rebuttal to the extremely glib take on software freedom in general that permeates development communities these days.

But the benefits of Free software, free candy and new features are all meaningless, if the user isn’t in control.

Don’t get in the van.

“The freedom to NOT run the software, to be free to avoid vendor lock-in through appropriate modularization/encapsulation and minimized dependencies; meaning any free software can be replaced with a user’s preferred alternatives (freedom 4).” – Peter Boughton █

Licence: Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0)

Permalink  Send this to a friend