Java as a Runtime Platform? I don't think it's dying anytime soon either.
JAVA might be a little bit old (in terms of technology) but more modern languages has a lot to do to catch up in:
* Multiple Relational DB support (I am still need to see good support for Python or NodeJS for databases like SYBASE, INGRES, others)
* Library ecosystem (there is a library for almost anything in JAVA) * Installed based (there is A LOT of Java Software out there)
I can see modern languages for the JVM like KOTLIN or SCALA fulfilling the places where JAVA is cumbersome.
Having said all this, unlike what was real 10 years ago, I definitely see that right now there are more JOB offerings for engineers in other languages (JavaScript/Node, Typescript, Python, and maybe even GO).
[EDIT]: grammar.
Right now, the TIOBE Index[1] lists Java as the third most popular language. Its popularity might also be attributed to usecases like Android development.
Meanwhile, I personally observe that Rust is rising in popularity, given its growing community. However, there are not that many jobs out there using Rust, so its usage may still be limited. (Although I hope it expands!)
More than language it is the basics that counts
The thing is that the underlying infrastructure of web backends has fundamentally shifted towards horizontal scaling away from the days of big metal. As opposed to a small number of fixed servers that serve a predictable amount of requests, modern backends are autoscaling or serverless and handling massive amounts of requests at a moment’s notice. Concurrency doesn’t matter as much when compute resources scale to meet demand. What does matter is startup time, and that is where Java is weakest.
Go is orders of magnitude faster than Java for responding to the first request. It also comes along with strong concurrency primitives and solid web frameworks. So you can run it with very few instances by default and quickly scale to meet demand. It also requires less effort to manage memory and GC.
Java has a vast ecosystem which is both an asset and liability. You can be sure that whatever you’re attempting to do with Java has already been done at least 5 different ways. That can make it difficult to decide on what the best approach to use, which library to select, etc.
Like, Python is king for data science /ML, and js is king for front end.*
It translates into a community of java devs which is less hegemonic by its relative size than it used to be.
That said, it doesn't imply that Java is loosing ground in other capacities. It makes great and frequent progress in terms of the JVM, language features, inter-operability with other languages. That makes it perfectly adapted to cloud environments. Old and tired statements about start-up times and memory footprint ignore the developments of Java in the last 15 years. Now Java instances can start in milliseconds with total memory footprints of a few mb [0]
Comparisons between Java and Cobol, that are sometimes made (even on this thread), are not enlightning: as far as I know, Cobol didn't evolve very much in the past decades. Not the case for Java - it has a 6-month release cycle [1], and these releases bring ambitious and significant changes[2], like virtual threads for one.
I would advise to stay updated on Java development by following key accounts of advocates on Twitter (eg, @nipafx), subscribing to the Java Magazine[3] or watch the many videos being posted by advocate devs on Youtube[4]
* even this needs a caveat. Many market leading solutions in data science are actually built in Java / Scala, like the Data Science Studio by Dataiku or H2O.
0: https://www.javaadvent.com/2019/12/getting-started-with-graa...
1: https://endoflife.date/java#:~:text=Java%20as%20developed%20....
2: https://chriswhocodes.com/jepsearch.html
In my niche, a COTS application that was written in COBOL has been migrating to Java for the last few years (happens to be an Oracle product through acquisition, but that's probably not relevant). As of now, the product is wholly written in Java, though large parts were passed through a COBOL-to-Java converter.
Java isn't going away any time soon - there are massive amounts of enterprise software now written in the language that will need to be maintained for years to come - but it's probably not the language of choice for a startup writing something custom.
Keep in mind that the HN community may not be indicative of the broader industry.
Cutting edge systems are not targeting Java. A small example off the top of my head: EdgeDB has no vendor supported Java client. TypeScript, Python and Golang are supported, but no Java. Ten years ago Java would have been a tier one target for such work. There are many other examples of this.
Java has next to no play in ML.
In its traditional form of compiled bytecode+runtime Java is a bad choice in cloud environments. Excessive RAM usage and slow startup create limitations and increase costs. There are techniques to deal with this, but adopting these comes with their own compromises.
Java still doesn't have a lightweight concurrency story. Project Loom and its Fibers are still at "early access" stage and it will be many years before it is mainstream and well integrated into common dependencies.
Java is far behind some newer languages with regard to safety. Yes Java is essentially memory safe and this has provided great value, but NPEs and some other common faults are are serious problem for Java due to the design of the language. Better designed languages have superior solutions for handling failures and have left Java behind, and there is little evidence that this is going to change via language improvements.
I'm not big on Java, but that's really a runtime issue, not so much a language issue. If Oracle really wanted to optimize Java, I feel like they could do that.
Also, I'm skeptical on the "increase costs" claim. The reason Java "worked" at the turn of the century was because corporations would gladly reduce developer time for the tradeoff of spending more on hardware if they needed performance.
The most performant solution would be writing assembly. Everything above assembly is a trade-off, C/C++/Java/Go/Python, etc.
That's sounds great until you have to pay the cloud bills roll in. At the turn of the century options were limited. Now they're not, and that's the world Java has to compete in today. Java is slowing giving up ground as a result. An idiomatic Golang program, for instance, uses far less memory to perform the same tasks as the equivalent Java.