WingNews logo WingNews
top | item 20074447

(no title)

unlogic | 6 years ago

To me, this makes total sense as the project moved to Apache. Obviously, much more people will be able to consider contributing when it's in Java. Apache goal is sustainability and long-term viability, and Java would work better for that.

I also consider this a success story for Clojure. It gives Clojure another usecase: a "production-ready prototype" language where the resulting "prototype" can last for eight years and benefit thousands of developers until it gets rewritten to something else when all the hard questions are answered, and most experimentation/wandering is over.

discuss

order

neilv|6 years ago

(I think the HN title was edited since your comment; I think an earlier title highlighted the move from Clojure to Java.)

> It gives Clojure another usecase: a "production-ready prototype" language where the resulting "prototype" can last for eight years [...] until it gets rewritten to something else when all the hard questions are answered, and most experimentation/wandering is over.

That general story, first heard from Paul Graham, about projects starting with Lisp and being successful that way, before eventually being rewritten, works for me.

I've been hoping at least a couple college startups would be inspired by this to use Racket initially, but if they have, I haven't heard of it. (I speculate that the current FAANG feeder emphasis among CS students hasn't helped. Who has time to play with a Lisp, when the now-all-important whiteboard interviews won't use it.)

> Obviously, much more people will be able to consider contributing when it's in Java.

If you have a company, and you want to hire Lisp people (whether it's Clojure, CL, or Scheme/Racket), I think you can probably hire people, because Lisp people like getting paid to use Lisp.

If you're looking for unpaid/volunteer contributions to an open source project, there's all sorts of things that affect that, and it's not unusual to get little-to-no contributions.

stingraycharles|6 years ago

> I think the HN title was edited since your comment; I think an earlier title highlighted the move from Clojure to Java.

This is correct, it was “Storm ditches Clojure in favor for Java”.

amortize|6 years ago

> I also consider this a success story for Clojure.

It is a success story for Clojure, but this move is a big negative feedback for the language. A team starting out on an open-source project will be mighty reluctant to start it with Clojure; because it might get rewritten not so much into the future. That is not good news for Clojure

fnordsensei|6 years ago

If, five years (or what-have-you) down the line—when you've pretty much enumerated all use-cases for the project, solidified what it's supposed to do and how, learned from your mistakes, and so on—you can't come up with a better design, something is off. With all that knowledge, there should be plenty of space for performance optimizations and pruning of vestigial stuff for a non-trivial project.

As for switching to Java, that makes perfect sense to me if that is what the contributors prefer. If you're going to make a big rewrite, you might as well change the language to one you prefer when you have the opportunity.

As someone who spends 100% of dev time in Clojure, I'm quite happy to see this development. Dipping into Java isn't uncommon for optimizations in Clojure, so this is like someone taking the time to do a massive under-the-hood optimization from the point of view of a Clojure user. I doubt it changes the project's status much in the Clojure community. Perhaps it'll even see an uptake in use.

What it doesn't say much about is Java vs. Clojure in my opinion. Different contributors, with different amounts of pertinent knowledge, and years apart. I'm not sure how you would control for those variables in a comparison. It should be read more as, "we put in a heck of an effort to make this thing faster and more useful", and kudos for that.

cutler|6 years ago

Pegging Clojure as a prototyping language relative to Java is hardly encouraging.

unlogic|6 years ago

Oh no, I'm saying it's another usecase, not the only one. And yet, it is a pattern I follow quite often in my mixed Clojure/Java projects: Clojure code is much easier to tinker with, take to new directions, rapidly modify and see the results. But once I settle with an approach/algorithm, I rewrite some things to Java for performance, or for easier interop with other Java/Scala teams, or because the imperative mutable code looks much better in Java.

So, you can say that Clojure among everything else is my IDE to write Java code. Is that a bad suit for a language? Depends on your perspective, but I personally am very happy that I have Clojure by my side.

usgroup|6 years ago

I think it’s noteworthy that whatever the benefits of clojure it didn’t outweigh the adoption issue.

It’s quite a mature language so I’m not sure that bodes we’ll for its prospects this late in the game.

unlogic|6 years ago

Whatever the benefits of Clojure are, they were enough for Nathan Marz and first contributors to write Storm in it and succeed. Note that it's not an argument about impossibility of writing something like that in Java (see Spark, Flink, Heron, etc). But at that exact moment in time, a Clojure Storm was created and became useful to many people for years. There were many more Java programmers than Clojure programmers then and now, yet the Java coomunity didn't produce their own Storm first.

I think it speaks of something.

lvh|6 years ago

If you can have the foresight to know how your project will be used many years down the line, sure, make technical decisions based on that :-)

(Personally, I think the change is rather evidence of the primary drivers behind Storm development in 2019, and... that's not a good thing.)

yogthos|6 years ago

A Java shop inherited a Clojure project and rewrote it in their preferred language. It's silly to read more than that into this.

stingraycharles|6 years ago

What if Clojure is a language that bodes well especially early in a project’s lifecycle?