[thesz]

https://www.youtube.com/watch?v=06x8Wf2r2Mc - at 10:52 Simon Peyton-Jones talks about life of programming languages.

Ruby, having at least million users around the world, is in immortal category.

What fascinated me in the original post is "...data model on the principles of the Value Object, introduced by Martin Fowler."

I mentioned Haskell because there is a long, long history of the use of immutable objects in computing science, dating back to Church and Turing's communication. The immutability is what differentiates lambda calculus and Turing machine.

That is why I remembered Haskell and the video above.

Yet, it is "introduced by Martin Fowler," for some reasons.

This is very sad.

[revskill]

Those Java Enterprisey guys as Martin Folwer is the reason i stop caring about Enterprise. Shitty stuffs get complicated for no reason.

[pjmlp]

People love to bash Java for what has been typical enterprise architectures since large corporations have used computers.

Or do people think that before Java we didn't have any architecture astronauts on large corporations?!?

[revskill]

Hacking on Java frameworks is mostly a pain. Both from IDE tooling (too much RAM, configuration fighting,...) to the documentation. Basically it made simple problems harder for just "JAVA" reason.

For comparison, i've spent many months to hack on Java framework in a company before to deliver a hackable feature.

By the same time, i also released 3 production applications from scratch with Rails.

The issue is not Java is impossible to have such productivity, it's the mindset, the culture issues from Enterprisey people.

[pjmlp]

Again, that culture was already there decades before Java came to be, COM, SOM, CORBA, Taligent, ER model, Booch method, Rational Unified Process, Structured systems analysis and design method. Yes, I am an old dog at this game.

Had Ruby and Rails been embraced at the same scale, you can bet there would exist a comparable flavour of Ruby EE.

[revskill]

As i said, it's about the mindset that matters (rather than throwing me a bunch of design patterns).

If you're at work and want to get shit done fast for the salary reason: There's no reason NOT to use Rails.

If you want to go enterprise, use something lightweight and modular, composable instead. Throwing a bunch of inheritable BeanClass serves no purpose on the job.

UNIX philosophy came long ago, before JAVA.

[pjmlp]

UNIX philosophy is cargo cult from a book worshiped by FOSS folks.

CORBA was born on UNIX, original implementation targeted C, C++ and Smalltalk.

Java was years away to be created, again by a UNIX company.

Java EE, was reborn from an canceled Objective-C project created by Sun and NeXTSTEP, again two UNIX companies.

[KronisLV]

As someone who has done a lot of "enterprise" Java, I feel your pain.

> The issue is not Java is impossible to have such productivity, it's the mindset, the culture issues from Enterprisey people.

I do think that a lot of it will depend on what sorts of projects people work with, as well as how old the frameworks are. Some design choices made in the development of those will certainly show their age. Consider this article, for some nice arguments in that regard: https://earthly.dev/blog/brown-green-language/ (I'd say this also applies to frameworks, not just languages themselves)

I actually recently used Dropwizard to create a proof of concept system for storing and querying millions of files by their metadata and using Java wasn't too bad in the end: https://blog.kronis.dev/tutorials/1-4-pidgeot-a-system-for-m...

I'd vaguely rank the frameworks and libraries as follows:

  - Java/Jakarta EE: pretty dated at this point, you'll mostly see it in legacy projects, generally not a "modern" feeling experience, if you need to configure a separate web server instead of an executable .jar file, you're going to have a bad time
  - Spring: if you see plain old spring, you'll have a bad time, it's the definition of "enterprise" code and will waste lots of your time
  - Spring Boot: can be good if you start a new project with a recent version, still somewhat enterprise but lots of integrations are just plug & play
  - Dropwizard: takes a bunch of idiomatic Java packages and glues them together, comparatively lightweight, you can even get by without dependency injection, most of the cruft that you need to deal with ends up being because of the language itself instead of a framework with bunches of weird dynamic class loading and reflection
  - Quarkus/Vert.X/...: new frameworks and libraries that try to build something more modern and performant, generally you can look in the direction of these for many of the modern projects that you want to do, though their lack of maturity might show sometimes (e.g. if you try doing something unsupported, or just run into their learning curve, e.g. how the Vert.X event loop works)

I guess I can sum up my thoughts with:

  Is Java as a language inherently a bit more annoying to work with than some of the other web dev languages? Definitely.
  Have its frameworks and the language itself gotten better in the past years, especially after newer versions than Java 8 came out? Yes.
  Are many out there going to pick it for its decent static type system, good runtime performance (JVM is great) or bunches of stable packages? Sure.
  Are some going to be pigeonholed into working with horrible legacy projects that are absolutely demotivating and show off the worst of the language? Sadly, also yes.
  Are others going to look in the direction of something like Kotlin, which attempts to solve some of its nuisances, or maybe other languages altogether? Also yes.

But at the end of the day, a lot of it comes down to familiarity: if you feel like you'll be more productive in .NET, Node, Ruby, Python or another language, and there aren't external external factors that force your hand in a particular direction, pick whatever fits your needs the best.

[thesz]

Okay, I'll bite.

Java could be a much nicer language with much less allowance for architectural space programs.

Type classes were introduced into Haskell as early as 1988-1989. At 1991 you could have programming in language with parametric polymorphism, type inference and principled (albeit called ad-hoc) overloading, there was at least one.

The ST monad, one of the first effect-containing abstractions, was introduced as early as 1992 - I think it was close to introduction of IO monad.

So, at the time Java was introduced in 1995, it could have all these nice properties which greatly simplify programming and make architectures easier to understand and more future-proof (Haskell's solution to an Expression Problem [1] is one of the simplest, it uses what was available in 1995, more or less).

[1] https://en.wikipedia.org/wiki/Expression_problem

Thus, we could have a programming language that makes architecture rocket launch much harder. But we did not.

The slogan was "write once, run anywhere" without any regards on how much is needed to be written or how to safely compose what is written.

[pjmlp]

I guess you never seen Haskell code in production with endless GHC feature flags...

[thesz]

I saw, of course. I use Haskell professionally for 15 years now.

These feature flags are not "architecture space odyssey," though. Usually they reduce boilerplate, often significantly.

They are also artifact of Haskell being "research language," as mentioned in video above.

[OkayPhysicist]

Programming languages exist primarily to provide pretty paths for certain kinds of development, which is to say by the structure of the language they make certain kinds of solutions more natural to the developer, while making others more painful. Java's pretty path is large, object oriented towers of classes, with state scattered rough-shod throughout the application.

To be fair, there has been some effort to ease the insanity. Record classes, lambdas, and the like lessen the pain of trying to write Java against its original orthodoxy. But before that, a simple "sort list of lists by first value" method involved making a class available in the current scope that implemented Comparator<List<X>> which contains a method "compare".

As opposed to "sortby (comparing fst) list_o_lists"

[pjmlp]

Languages are orthogonal to the desires of technical architects, solution architects and business analysts, that control the helm of enterprise shops.

There was no Java when GoF book came out, CORBA and DCOM were used all over the place, SOAP was initially driven by Microsoft and its COM/.NET group (see Don Box), when enterprise architects get FP into the enterprise, scalaz happens.