> sweet careers are made of this, so who am I to disagree? Compile the world; Java Python C. Everybody’s looking for some bug. Some of them want to maintain you. Some of them want to be maintained.

For those missing the reference:

https://youtu.be/qeMFqkcPYcg?si=at-YtggekbPdv7sN

The desire of the HN community to pull a random person's uninformed opinion about a topic that they, justifiably, wrote for their own interests and amusement and then pontificate about how either stupid or amazing it is will never ceise to confuse me.

Effects on their own are a very active area of research and I would laugh behind a PL researchers back if they claimed it was a solved issue. Between Monads, call-by-push-value, and algebraic effects there is really no clear "how do people actually use this" answer.

But that's not the job of a PL researcher anyway, or a random software engineer for that matter. Sorry to say, the software engineer knows next to nothing about "the right way" to design language features that people want to use or enjoy using. If anything this should be an HCI person with a penchant for PL or vice versa.

There’s a certain amount of hubris to say, “I don’t know anything about this and you’re making a mistake.” It’s off putting and kills the whole rant.

I’ve heard opinions from smart people with lots of experience who say algebraic effects are not worth the squeeze. I’ve also heard some say that we should all be pushing the boundaries: they are the future.

So the matter doesn’t seem to be decided. Now isn’t the time for maxims.

Every time I read stuff like this it just makes me laugh, I honestly never know who to listen to in these debates.

Research doesn't work like that. I like the idea of separating contract and implementation in algebraic effects. It might pave the way to bring back some sanity to imperative languages and help us write better code, since it's pretty clear that the "real world" doesn't care much about pure functional languages no matter what they bring to the table. Or algebraic effects could be like monads, many like to talk about them while people building real stuff have no clue about it, nor do they care. But we'll never know unless we explore.

To which implementation is the author referring, I wonder?

I can’t say I recognise any of these issues from freer, polysemy, nor bluefin.

The author says the approach they advocate (just using function parameters) is similar to "dependency injection". It looks like in FP/objects-are-a-poor-man's-closures terminology they're talking about Continuation Passing Style (CPS).