Really good read on language design and community management. Sometimes the principled design approach takes a toll on everyone when it's iterative https://lukeplant.me.uk/blog/posts/why-im-leaving-elm/

I don't follow Elm, so it's the first I heard of this controversy. Nevertheless, I understand why Elm 0.19 needed to impose these restrictions. Elm is a pure functional language. Functions cannot have side effects. It's the same for my language, Curv, so I'm dealing with similar issues. I don't see any discussion on the internet explaining the Elm 0.19 restrictions from the language maintainers perspective, so I'll give my take.

This post focuses on what he sees as hypocrisy: many non-essential modules owned by the maintainers and their friends are impure (eg Evan's markdown module). So the whole purity transition is inviable if the maintainers themselves can't adapt to their own pure standards.

Some people can't handle the benevolent dictator for life control that language authors can have. It certainly didn't hurt Python, which reached up into the top 3 languages, with benevolent dictator Van Rossum just stepping down recently. Evan Czaplicki is a smart guy, and he is doing a first class job on Elm. However, with the limited resources available, the features people want (and Mr. Plant wanted native modules and a localization system) may not happen in a timely fashion, and in that case you have to bail. It is after all still before version 1.0, so one has to give them some slack. Mr. Plant has a 7000 line program he doesn't want to rewrite, so i understand the pain, but luckily 7000 is a small program. Elm doesn't have magic dust so it can be converted, the effort can be measured in days of work. There is a lesson here for all language designers; it is much less painful for the users if you only add to the feature set; subtracting a previously available feature may cause agony among the user base. But there is a weakness in the additive language evolution process, because if you avoid pain in the user base by only adding (C++ are you listening?), if you keep up that process, you end up with a kitchen sink type of language mess after a few decades, and the whole thing should be be thrown out. If your original design is clean, hopefully you won't need to add much. I used Modula-2 and it only needed adding a preprocessor for conditional compilation, and some unicode string markers in its evolution. It was so simple it was complete from the start.

@Edward de Jong / Beads Project agree with all of that -- one thing I'll add though is that the "1.0=production" distinction is far more complicated than you make it. Plenty of companies use Elm in production (which the team does evangelize), and plenty of other projects haven't hit 1.0 and it's been years (eg React Native)

In my first job after college i worked for a consulting firm that made a decision support system for Fortune-50 companies. It was a terrific product, but when a user wanted a feature, they would throw serious money at the company and they got it. If they put enough money on the project it got bumped up to #1. It is an honest system, money is voting; in open source you have an economy of feelings, where people think their devotion to a product and evangelism somehow give them more votes. In many ways i prefer the clarity of the commercial world. A language is a tool, and its value is created by users accomplishing something else. It is a crucial part of a project, but just like a hammer or saw, it is not the final product. The one area where software tools have a flaw is that when you construct a house the hammer and saws used to build the house are not left inside the house; while in software, there is usually some runtime system that is permanently embedded in the product and if that runtime gets broken the whole thing stops functioning. This is a key reason the software industry has been like building sandcastles.

"I'm well aware how entitled people can be in OSS".. followed by a gigantic wall of text that doesn't manage to disguise just how entitled he is..

One thing that made Modula-2's simplicity survivable was that there was a way to call external functions which might be in assembler or C. When you encounter a very high cost area of computation, like bitmap masking and rotation, you occasionally need to dip into lower level languages, and having an "escape hatch" can be very helpful. However in modern high-runtime environment languages like Elm or Beads this isn't really feasible for the most part. The more dramatic the improvements the language offers, the less it will connect with the existing systems. This is an intrinsic tradeoff i believe. In

@Kartik Agaram care to explain your 👎? All of his complaints eventually boil down to him demanding to use the time of people he's not paying. That is never acceptable in my book.

He's a better writer than me, so if that's all you got from this post I can't do better, sorry. It's vehemently not what I got from it.

That's a "no", then.

Yep. You're trying harder than usual to 'win' this argument. Think about why that is. And maybe reread the post. It's pretty well-written so I'm not sure where the disconnect is in your mind.

Having been on both sides of this, I agree with the author. @Wouter I'm with you on the entitlement and the ridiculous expectations people have of open source maintainers - I still remember the first time I had someone screaming at me because LightTable didn't work on their crazy custom version of Linux. But, if you openly encourage people to take a dependency on you, if you go out of your way to build a community and promote the use of your software, you are signing up for an implicit social contract. You made the choice to place yourself in that position and so you are now held to the behavioral standards that come with it. You can't have it both ways: if you want to just sit at your desk and make a thing you want, that's fine. If you put that up on github and some other people take an interest, that's fine too. You have no responsibility. But once you actively promote your system and tell people they can and should build on top of it, you've explicitly chosen to be a steward. There are significant benefits that come with that choice (Fame! Fortune! 😉 ), but it isn't free. It comes with the burden of supporting those you encouraged. You can still of course choose not to do that, but you're now breaking a societal contract and that eventually causes posts like these.

@ibdknox I don't buy that.. even if you generally want your software to be popular, and you want to be accommodating, your time never grows beyond 24hrs a day, yet the amount of people, and the amount and diversity of demands will always dwarf your time, as it gets popular. And every decision you take, no matter how reasonable, will be someone's most most evil move ever, and they will spend all their energy fighting you over it. There is no "social contract" that can fix this. In the end, the amount of time the original author is going to spend on your demands is "best effort" and "within reason", at the discretion of the original author.

Yeah, I certainly appreciate the scar tissue maintainers have, and the level of bad behavior they often have to put up with. Certainly I'm willing to believe the ratio of bad behavior is 10:1 or 100:1 against maintainers. But it seems excessive to claim there's no social contract at all. In this case what we're talking about is sins of commission, not omission, actions that have taken lots of energy to take. So I don't think "best effort" holds water as a justification.