The Java ecosystem has been slowly dying for years, it's already the new Cobol at this stage.
I have the impression that a lot of the mindshare has switched to the Node and Javascript ecosystem, especially the frontend devs, via the Typescript / Angular path, which is a natural path to Java developers due to the comfort given by familiar concepts like static types, classes, etc.
When I used to do Java, I always found that Spring was a much better framework than Java EE, and have used it in most projects I've worked.
Oracle did not take a note from the Microsoft book and is really behaving that an evil corporate overlord.
It's clear at this stage that they bought Sun in order to slowly turn Java proprietary, and charge billions to the companies that are now fully Java-based for the next few decades.
I'm glad I saw the writing on the wall and switched, but the problem is that Java is still being taught a lot at schools, and school programs especially universities are hard to adapt and only change every 4-5 years or so.
Working in the JVM world I'm always surprised to hear that its dying, the JVM and Java moves faster forward than ever, Graal is about to go 1.0, a bunch of very interesting (very) low latency GCs are in the pipe (for example ZGC and Shenandoah) and lots of other interesting projects like Valhalla promise to keep the JVM relevant for the next 20 years. Oracle have made some moves to monetize on parts of its investment, but the situation is largely misrepresented.
Too little too late, and I highly doubt that (5-year-old) Valhalla will deliver even one of the promised features (value types) in the next 3 years. Remember that the original plan was to have all Valhalla features in Java 10.
Java/JVM was in the ideal position to dominate the software world for the whole 21th century 15 years ago. Instead they did an IE6 and stagnated for so long that their competitors became plain better.
I'd say Java/JVM is now where IE was in 2008. Still massive majority market share, but in decline because people are switching to competitors. LLVM and webassembly are eating JVM's cake, and a multitude of languages like Kotlin, C#, Go, Rust and even Javascript and Python are chipping away at Java's market share.
At any given time most developers will neither have much real choice when it
comes to technology nor have a longer term perspective. Hence most developers will stick to what they know and what organizational inertia seems the path of least resistance.
The comparison to COBOL is good for several reasons.
> Working in the JVM world I'm always surprised to hear that its dying
So how is this "JVM world" escaping the grip of Oracle? E.g. graalvm is directly Oracle's, or am I missing something? Specifically, what is preventing any of the Java-related things people use now (including those that you mention) to end like the OP is describing?
Asking as somebody outside of that world, but interested "how much it's worth considering" under the given circumstances (Oracle being Oracle, demanding a lot of money for everything Oracle's that some company could use, and owning a lot of that "world"). Where are the boundaries? What is safe and outside of reach of Oracle? E.g. what is safe for a small company to use?
Java the language is slightly different than JVM but nevertheless the future is not very bright IMO. So far there are a few areas Java/JVM still holds a strong position but if you take a look at GitHub trending there is little or now new blood coming in.
- Android. No serious competitor to Java yet (JVM is irrelevant). But the Google lawsuit could have some complication.
- Spring stuff. I won't be surprised if they will be replaced by golang and nodejs (along with react/angular/vue etc). Same path for RoR.
- Data processing: Spark stuff, PrestoDB, Flink, Kafka, Hive, HBase, Lucene/Elastic etc. Java/JVM is still dominant but golang could be a future contender. A few new application databases/KV stores are implemented in golang.
The problem for Java is that the last category is mostly services (instead of libraries/frameworks) so you can potentially use any language to work with them, and the industry probably won't create many jobs for building generic services especially in the cloud era. So having the dominance doesn't provide a lot of protection.
> The Java ecosystem has been slowly dying for years, it's already the new Cobol at this stage.
This has been said for at least the last ten years. Not only is Java not dying, but companies like Apple and Facebook (not to mention Netflix, Amazon and Google, that are mostly or largely Java already) are increasing their investment in the platform. This was not the case when COBOL was 25.
The reason this is so is that Java is by far the most technologically advanced server-side platform in existence. It continuously presents technological breakthroughs in GC, compilation, and low-overhead profiling technology (just in the past few years it did so again in all three -- with ZGC, Graal and JFR). Node, Python, Go, Erlang -- none of them even come close. .NET comes closest, but even it is pretty far behind. This certainly wasn't the case when COBOL was 25.
It is true that for a short period of time, somewhere between 2002 and 2005, Java was more dominant than it is today, but that was a fluke. More "scripty" runtimes and languages were usually more successful in the "smallish software" domain. In 1995-2002 that was VB (and others); now it's Python. Also, there has been some decline in all languages due to the ubiquity of JavaScript on the browser, but Java did not suffer more than others. Maybe less so.
> Oracle did not take a note from the Microsoft book and is really behaving that an evil corporate overlord.
Oracle has just open sourced the entire JDK. Sun didn't do that and even .NET isn't entirely open source. I have no knowledge or involvement with EE, but it seems Oracle simply doesn't have faith in EE, and doesn't want the Java trademark associated with it.
(I work at Oracle on the JDK -- i.e. OpenJDK -- but speak only for myself)
I think you’ve drunk the kool-aid a little. Outside of complex projects headed by PhD compsci students (who used to learn java in uni but those courses are now taught in python) there isn’t a whole lot of mindshare going towards java these days. At least in regards to new FOSS projects.
I can say that my company’s use of java is diminishing rapidly over the last 5 years too.
No matter what you think of the technological advancement of the language and ecosystem- it matters not if people are not using it.
Oracle itself is a pariah, the vast majority of the tech industry avoid it.
Last time I had to download a JDK I got a warning about having to pay a license for commercial use. I think most enterprises care more about the licensing arrangement than access to the source.
JavaScript is practically stone age compared to Java. Debugging is a nightmare. Documentation is a nightmare. Scaling is a nightmare. NPM is a thousand monkeys at a thousand typewriters.
Java suffers from having some 20-year old codebases because it's been so popular for so long but I can open up a 2004 WebLogic project in Eclipse or InteeliJ and trace everything it's doing. And modern Java has added all the nice features we expect a modern language to have. Java's slow demise is due to terrible PR and Oracle's endless insistence on JEE when every sane developer is using Spring.
I am not very sensitive to all those issues, I find the JS/CSS/HTML web stack a bit antiquated (virtual DOM, LOL), but very usable.
But the build toolstack is a pure, unadulterated nightmare. I used to think Maven's XML orgies, archetypes and learning curve were bad, but diving into Webpack et al was a new dimension of horror. This time based on JSON, though.
Time for another sortie by the COBOL evangelism task force.
Java has been the new COBOL since its inception in the 1990s.
I programmed quite a bit in COBOL early in my career. It's an extremely efficient language for scanning records especially in an era when a large proportion of data was stored offline on mag tape. Most common business problems of the era could be expressed and solved effectively in COBOL, which is why it became so popular.
I programmed in both C++ and Smalltalk in the early 1990s, which were innovative but also difficult to use for large systems. Java hit a happy medium with its C++-like object syntax, built-in libraries, and virtual machine operation with garbage collection. Add in the network effects from wide adoption (leading to tools and libraries) and it became a huge hit.
Oracle seems to be going out of its way to submerge the language in confusion. But nobody should mistake this as a reflection on the inherent capabilities of Java or its fitness for use.
I think you're missing the point of the COBOL comparison. It is not meant to imply that Java is similar to COBOL as languages go, or that it is as bad as COBOL, or even that COBOL is bad.
In fact, you described exactly why it is the new COBOL: because it is also an efficient and popular language for common business problems of our era, that is hugely popular, but mostly in "unexciting" places like large internal enterprise apps, out of sight for many. And just like COBOL, it will keep ticking on like that, probably forever.
Spring is compliant with and reliant on many of the JavaEE specifications - JDBC, Servlet, JAX-RS, etc. It provides DI that is not JavaEEs CDI.
I'd consider Spring as more of a non-JavaEE application server. Pick either Spring or Wildfly/Glassfish/Websphere/etc and you're either using part of the JavaEE specification with the former or the full specification with the latter.
Avoiding JavaEE for micro-services requires moving to something like DropWizard. Instead of a 200MB application server you end up with a 20MB application. Unfortunately, running that application in a Docker container will spin up a 180MB JVM. So the main reason we're moving away from JavaEE is actually more related to our move to container orchestration and that was originally driven by our adoption of micro-services. If you write a micro-service with a single end-point why should it use 500MB of memory?
Java is much closer to C++ than COBOL. COBOL has been massively obsolete for decades. Java has a lot of historical cruft but the basic syntax is still somewhat modern, mostly based off of C++ and agruable better.
Like C++, it's modern features are being slowly added. There's a huge difference between Java 8 and 11, in performance and features. Innovation is hardly dead.
Java EE and Spring are both terrible. I recommend trying some of the newer frameworks out there to see how much nicer things have become. DropWizard is my favorite right now.
I'm no Oracle fan, but Java is more "open" in many ways than it has been since the beginning. It wasn't till a few years ago we had a feature-complete high performance OSS implementation as the standard.
Oracle will try to ruin it, but the GPL JDK and history of projects being forked away from them makes it fairly safe. It's not in the interest of any Java user to have more than one standard.
IMO Java isn't going anywhere. It's been top 3 most popular language for over a decade. JS is somewhat promising but will never match Java in performance and multi-threading support due to core design decisions. Go is too verbose to get industry buyin to replace Java in most use cases. WASM is promising in the far future but it will take years to have a full browser and server ABI. The only real contenter to replace Java is C# and so far it's not getting much traction.
If anything killed Java it would be C#. Value types, no type Erasure, unsafe code and some native vector support. But it hasn't and probably won't ever dethrone Java. Java has a combination of decent features and an insanely huge amount of library support that makes it the default language of many companies.
How can a languages and its ecosystem that is being used actively in many projects within ALL Fortune Global 500 company and most Fortune Global 2000 be dying?
When people say Ruby is dying, it would be more accurate to say they are not growing or they are shrinking. But given the size of Ruby marketshare of less than say ~2%. I would not really disagree with their notion.
But Java? I mean even from a Web Development prospective, given you mention Node, JS and Spring Framework, it is Far from dead.
And this is speaking as someone who likes Ruby and dislike Java.
Recently started working on Typescript project and it looks to me like java with really poor tooling, C++ like compilation times and huge mess in libraries (i.e. how to determine which version of @types package I need). I think that Typescript is really step back from java.
As a Delphi developer this is the first time I heard about Java is dying, and in a reply the other one said that some is saying so for over a decade :D
And I think Java is more alive then Delphi, at least no one is setting up a site like http://www.isdelphidead.com/ :D
On the contrary! They are likely very aware of Microsoft's strategy and probably regard it as being too soft. They intend to milk their technology to the last drop.
You and I may regard it as evil, but it's a perfectly valid strategy. Perhaps it's more common outside IT? It's the brutally honest business strategy.
I can sort of appreciate that they don't play around. Perhaps not my first choice of technology, but you certainly know where you have them.
I've spend quite a bit of time doing javascript and typescript lately. Typescript definitely patches up a lot of the ugliness of javascript but you still deal with the same messy ecosystem of tools, libraries, and other cruft. It's been slowly getting better over the years but it's still not that great. I see typescript more as a gateway drug for people who seem to think that Javascript is all there is and then discover that actually having a type system unlocks all these amazing new tools that the rest of our industry has been enjoying for a few decades.
Guess what, if you like typescript, you'll love Kotlin. It's a much cleaner language, just as expressive, and it has a whole range of things that typescript simply can't do or that it simply does better/more elegantly because typescript is stuck preserving compatibility with javascript and all it's weirdness. MS has done a wonderful job of shoving a lot of that under the carpet with typescript but fundamentally types are optional because all javascript has to be valid typescript and therefore it has to also allow all the silly hacks, misguided features, and other cruft that javascript people have been coming up with for the past decades.
You might also like C#, Rust, Elixir, and several other languages that each offer their own tools, libraries, and features. I'm seeing a pattern with many frontend developers in the last few years that they at some point become full stack developers and shortly after start adding more languages to their repertoire. I know former node.js developers doing Rust, Go, Elixir, Kotlin, Rust, etc. now. Most of them are not very eager to move back to doing node.js projects.
Like Kotlin all these languages are all coming to the node.js/browser ecosystem through WASM. Full stack in 2020 is going to be a very different beast than just a few years ago.
Right now WASM already allows lots of languages to be used for just about anything that people currently do with javascript and typescript in browsers or in the node.js ecosystem. As this becomes more mature, people are inevitably going to change the ecosystem. This is already happening. E.g. Rust is being used for many low level npms for a few years now. However people have also been developing frameworks and libraries in Rust that work around/replace that ecosystem entirely. For a Rust person dealing with crates is preferable to dealing with npms, webpack and all the other stuff that is used in that ecosystem. So, they are fixing that. The goal there is to run pure rust in wasm in the browser or on v8 with 0 npms pulled in or used during the build process.
Likewise, there's an ecoysystem of multi platform kotlin libraries emerging and that allows you to build wasm or native applications and libraries without needing things like webpack, npm, etc (but using gradle instead). Same with C# and blazor. Each of those ecosystems uses browsers and node.js as runtimes but try to minimize/eliminate/replace most of the rest of that ecosystem. In the same way the Kotlin ecosystem is also starting to provide alternatives to stuff from the java EE and Spring ecosystem. E.g. Ktor is a multi platform web application framework that you can use instead of spring boot and run natively or on the jvm. It's also a drop in replacement for anything you might want to do with e.g. express.js. Doing express.js applications in Kotlin is of course also possible, but why would you when you can have ktor instead? In the same way building react applications in Rust is possible, but replacing it with a rust specific framework is much less awkward.
So, I agree things are changing but I'd say the node.js/browser landsace is changing as well and will not involve a whole lot of javascript/typescript for people coming from the JVM or other ecosystems.
You might make the point that there are all these millions of npms out there that can't be replaced easily. I'd counter that with the point that an awful lot of them seem to have very short shelf lives and are in fact replaced/discarded regularly.
I would not go as far to say it's dying but it's heavily slowing down for sure. First because of the big move out of the "entreprise" software and secondly because of Oracle legal team. It's definitely due to the licensing that Android is switching to Kotlin for example.
a) Java/JVM hasn't moved out of the enterprise space at all. Not sure where you got that idea.
b) Android isn't switching to Kotlin because of licensing since it is a JVM language after all. It's purely because it is a much better language than Java and has great tooling courtesy of Jetbrains.
Oracle makes Java a risk factor in all projects. I stopped approving new Java projects in my team 2 years ago after we switched to Go. This wasn’t a decision we took lightly and it took almost a decade for me to be comfortable with leaving Java behind.
But it is the right thing to do. Oracle is not trustworthy.
Strange attitude to take when you can just target OpenJDK and then it is completely irrelevant what Oracle does or does not do.
The fact is that every single one of the largest companies has some Java or Scala code within their core platform. And none of them have ever raised Oracle as a risk.
In this case we have Oracle using the JEE namespace discussion as leverage to try to force the Eclipse Foundation to distribute their VM with their products.
If Oracle had succeeded, would that mean that every developer that downloaded the Eclipse IDE would be subject to Oracle's licensing and a possible audit target? I believe that it would.
The most valuable corporation in the world as of this date uses zero Java for anything “core”.
Anecdotally, nearly every CIO/CTO I’ve encountered in the last few sees Oracle as a huge risk. Most particularly with regards to databases, which are even sticker than applications which have shorter life cycles.
What's the relationship between OpenJDK and Oracle, exactly?
As somebody who's not deep in the Java ecosystem, the precise relationship is not clear to me, but it does seem that Oracle is a major contributor to OpenJDK at least, most likely also involved with the project steering.
It's sane to be leery about entering into any kind of contract with a counterparty as grasping and merciless as Oracle has proven to be. If they find an exploit in the GPL, Java projects are in for a bad time.
Just curious, what makes you think Google trustworthy (in relation to Go)? It is not surprising that corporates control and drive the languages and ecosystems in their own favor. IMO, Rust and D are truly open source, even though D lacks the corporate sponsor like any other major language.
« Oracle already started a blocking politics at the Eclipse Foundation’s board, where Oracle has a seat, and where unanimous decisions are needed. Oracle now has the power, and apparently will use that power, to block the foundation’s future. It demonstrated that power already in a board meeting, where they had the sole vote against an otherwise unaimous move. »
I hope other organisations with a director from Oracle are taking note.
For me it's more a lesson about how unanimous votes, or unanimous - N can become problematic.
In the EU the same is happening. Poland and Hungary have found a loophole allowing them to ignore EU rules and sanctions while still profiting from subsidies and market access.
All sanctions against a member state require unanimous-1 (all states except the sanctioned member) votes and Poland is vetoing all sanctions against Hungary, and Hungary all sanctions against Poland.
The irony is that Poland was once the #1 or #2 power in Europe, but they were destroyed by unanimous voting. Want to raise taxes to build an army to repel foreign invasion? Everyone agrees, except that one person too far from Russia to care, and the taxes get vetoed, so Russia wins, thus Poland was destroyed. The Poles know how dangerous unanimity is.
> Poland and Hungary have found a loophole allowing them to ignore EU rules and sanctions while still profiting from subsidies and market access
AFAIK this was tit-for-tat: HU & PL closed/sold-off their industry in favor of WEU in exchange for subsidies. Removing these subsidies for any reason would violate the main economic agreement between them and EU. Mind you, there is no significant locally grown industry in those countries anymore (outside of a few state-owned industries), WEU profited from this enormously, especially Germany. As economic potential of this arrangement has been exhausted after 25 years already, incorporating UA is the next move to give EU (primarily DE/FR/NL) some nice economy boost for the next 10 years. EU is essentially playing Sim City.
You think e.g. Germany doesn't do the same with several satellite countries they play for with funds, trade deals, diplomatic pressure, and so on, to get a bunch of votes on EU matters based on their interests?
Hungary and Poland aren't the first to abuse the unanimous-1 veto.
The Netherlands always vetos Romania's application to enter the Schengen zone as Romania's harbors to the Black Sea will provide cheaper alternatives for import of goods to the Dutch ones.
UN security council is another classic example, where any permanent member can use the veto to block any resolution they want, regardless of how the other permanent or elected members vote.
Well if they believe similar things they vote accordingly. It's not a loophole, just two governments having similar views about those issues. It's for the better as well, if you allowed sanctions for refusing immigration policy you would end up with huge backlash and likely several countries exiting EU.
OK, but Poland and Hungary are far from the only nations ignoring EU rules (though they might be the only ones ignoring certain particular rules).
Take for example the Italian government's decision to run a deficit higher than what is allowed under EU rules (2% per year if I recall correctly). The EU tried to put a brave face on it: they held negotiations with the Italian government and announced a compromise, but it was pretty obvious to people who follow EU politics that the EU had no effective means to enforce the rule on Italy.
And this cap on deficit spending by governments of member nations is no minor rule. Many had stated over the years since the introduction of the Euro that the rule is necessary for the single currency to work.
I don’t understand how the org let that happen... Does someone know or have articles about it? I can believe that they left the unanimous vote and then had the good idea to invite Oracle at the table without some
Power play behind it.
I bet Oracle will have its MS pivot moment,
sooner that we think,
to face AWS, PostgreSQL, IBM/Red Hat, Pivotal Spring offers...
But it was a lot of time wasted:
Apache Harmony/Java
LibreOffice/OpenOffice
MariaDB/MySQL
Jenkins/Hudson
now Eclipse/Java
Right now, they are in sales hurry mode, calling at all occasions technical contacts each time you create a new account to download something on their web site...
> the worst case possible, as it not only voids the “WORA” (Write Once Run Anywhere) principle, it simply won’t happen in reality: After 18 months virtually no application vendor really wants to spend the time and money to update all customers with recompiled versions just for the sake of a renamed platform with a dubios future.
WORA was never perfect, I would be surprised if there was even a single JRE release that didn't note a breaking change in its release notes. So vendors should have a process to update their releases just for that already in place.
> The future is unclear because Oracle already started a blocking politics at the Eclipse Foundation’s board, where Oracle has a seat, and where unanimous decisions are needed.
They have no rules to out board members acting in bad faith?
Well, mmm, good riddance? I suppose this will cause a lot of hardships to a lot of people who need to maintain JavaEE, but the fact that Java EE is no more is nothing but good. It is a nightmare case of overengineering.
I never used a flow blown Java EE web application server, but the various APIs I used like JDBC, JAX-RS, and JPA were generally well thought out and pleasant enough to work with. It would be a shame if these interfaces/specifications became proprietary.
Without knowing it you have cherry picked possibly the only good parts of Java EE. The rest that constitute a 'full blown' Java EE stack are horrific disasters (JSF, EJB, etc).
I switched to C# late last year. Oracle is just too much baggage to have hanging around when thinking about your career / future for me. Really didn't like the direction they are taking things when they announced their new release train and licencing model. Compared to that Microsoft have done an about face in recent years with a commitment to open source and first-class support for Linux and revamping their core offering with .NET Core. It's been a mostly smooth transition.
JVM is still the best platform for any type of back-end application. It has the widest array of libraries, is proven to run on any type of hardware and comes standard on every OS and every cloud instance.
.Net is a great platform and all but it still hasn't done enough to really merit switching teams. Especially when you have Scala, Kotlin, Clojure etc to pick from.
hand up uh not as experienced as most people on HN, what's the hate for c#? I went from vb6 to VB.net then c# and while I fully admit I'm a casual coder, i haven't had an issue (with what I do, which is voice to string primarily) so far. Is there something better for this I perhaps should look into?
First C# needs to get into all the platforms where there is a Java compiler.
And a proper story for something similar to Swing, not even considering JavaFX here. There is Xamarin, but again, the platforms with Java compilers available.
And I am stating this as someone that works daily within .NET eco-system.
I think every Java developer should be familiar with C# and dotNet Core. There is a lot of similarities between the languages and it will be easy for any Java developer to pick up.
Any organization that bought into JEE is probably willing to pay Oracle's license fees. That being said, I've literally never seen an org that preferred JEE to Spring and that includes numerous massive enterprises.
>But once Eclipse products would be not vendor-neutral anymore, the EF’s tax exemption might become void, which would mean a financial fiasco, or possibly mean the end of the organization as a hole.
I think the problem is Oracle's a-hole, in this situation.
> And the idea that corporate developers or even frankly any developers are switching to LLVM and WebAssembly en masse is pretty ridiculous.
There is only 1 popular language released since 2009 that targets JVM: Kotlin. All other target JS, their own bytecode or LLVM: Go, Rust, Dart, Elixer, Julia, Typescript, Swift.
This is a significant difference from 2003-2009 languages which all targeted JVM unless they were Microsofts: Groovy, Scala and Clojure.
And who uses Dart? Elixir? Julia? Hardly anyone. Lots of attempts but also lots of failures, sometimes due to poor runtimes or libraries.
Swift is iOS/Apple specific, at least in usage. But mobile is dominated by Android, which is moving towards Kotlin. So Kotlin is already huge right there, much larger than Dart or Elixir and probably larger than Swift, soon if not already.
Meanwhile Go would probably benefit from targeting the JVM, or at least having the option. As is, it's stuck in its own isolated world with low quality tools, libraries and GCs.
At any rate number of languages is an odd metric. Kotlin by itself is a good enough language to probably kill off most attempts to compete with it over the next 5-10 years or so on the JVM, outside of research languages. Does the JVM really need dozens of languages?
A lot of new languages fail. It's still a good measure of where things will be 10-20 years later.
Look at new languages of 1990-2000: Haskell, Python, VB, Lua, R, Ruby, Java, Delphi, JS, PHP, Rebol. 4/11 are mostly irrelevant now. 4/11 are important niche languages. 3/11: Python, Java, and JS are dominant players.
1990-2000 might be special because of a massive shift toward IT systems in general, but it might not be and be an indication of languages lifecycle.
> stuck in its own isolated world with low quality tools, libraries and GCs
"Go needs a native library for ___" is used as an opportunity for safely tedious résumé-driven development. On the JVM you'd have to look for an unsolved problem which is probably harder.
Readit News
I have the impression that a lot of the mindshare has switched to the Node and Javascript ecosystem, especially the frontend devs, via the Typescript / Angular path, which is a natural path to Java developers due to the comfort given by familiar concepts like static types, classes, etc.
When I used to do Java, I always found that Spring was a much better framework than Java EE, and have used it in most projects I've worked.
Oracle did not take a note from the Microsoft book and is really behaving that an evil corporate overlord.
It's clear at this stage that they bought Sun in order to slowly turn Java proprietary, and charge billions to the companies that are now fully Java-based for the next few decades.
I'm glad I saw the writing on the wall and switched, but the problem is that Java is still being taught a lot at schools, and school programs especially universities are hard to adapt and only change every 4-5 years or so.
Ref:
- https://openjdk.java.net/projects/valhalla/
- https://www.graalvm.org
- https://wiki.openjdk.java.net/display/zgc/Main
- https://wiki.openjdk.java.net/display/shenandoah/Main
Java/JVM was in the ideal position to dominate the software world for the whole 21th century 15 years ago. Instead they did an IE6 and stagnated for so long that their competitors became plain better.
I'd say Java/JVM is now where IE was in 2008. Still massive majority market share, but in decline because people are switching to competitors. LLVM and webassembly are eating JVM's cake, and a multitude of languages like Kotlin, C#, Go, Rust and even Javascript and Python are chipping away at Java's market share.
At any given time most developers will neither have much real choice when it comes to technology nor have a longer term perspective. Hence most developers will stick to what they know and what organizational inertia seems the path of least resistance.
The comparison to COBOL is good for several reasons.
So how is this "JVM world" escaping the grip of Oracle? E.g. graalvm is directly Oracle's, or am I missing something? Specifically, what is preventing any of the Java-related things people use now (including those that you mention) to end like the OP is describing?
Asking as somebody outside of that world, but interested "how much it's worth considering" under the given circumstances (Oracle being Oracle, demanding a lot of money for everything Oracle's that some company could use, and owning a lot of that "world"). Where are the boundaries? What is safe and outside of reach of Oracle? E.g. what is safe for a small company to use?
- Android. No serious competitor to Java yet (JVM is irrelevant). But the Google lawsuit could have some complication.
- Spring stuff. I won't be surprised if they will be replaced by golang and nodejs (along with react/angular/vue etc). Same path for RoR.
- Data processing: Spark stuff, PrestoDB, Flink, Kafka, Hive, HBase, Lucene/Elastic etc. Java/JVM is still dominant but golang could be a future contender. A few new application databases/KV stores are implemented in golang.
The problem for Java is that the last category is mostly services (instead of libraries/frameworks) so you can potentially use any language to work with them, and the industry probably won't create many jobs for building generic services especially in the cloud era. So having the dominance doesn't provide a lot of protection.
This has been said for at least the last ten years. Not only is Java not dying, but companies like Apple and Facebook (not to mention Netflix, Amazon and Google, that are mostly or largely Java already) are increasing their investment in the platform. This was not the case when COBOL was 25.
The reason this is so is that Java is by far the most technologically advanced server-side platform in existence. It continuously presents technological breakthroughs in GC, compilation, and low-overhead profiling technology (just in the past few years it did so again in all three -- with ZGC, Graal and JFR). Node, Python, Go, Erlang -- none of them even come close. .NET comes closest, but even it is pretty far behind. This certainly wasn't the case when COBOL was 25.
It is true that for a short period of time, somewhere between 2002 and 2005, Java was more dominant than it is today, but that was a fluke. More "scripty" runtimes and languages were usually more successful in the "smallish software" domain. In 1995-2002 that was VB (and others); now it's Python. Also, there has been some decline in all languages due to the ubiquity of JavaScript on the browser, but Java did not suffer more than others. Maybe less so.
> Oracle did not take a note from the Microsoft book and is really behaving that an evil corporate overlord.
Oracle has just open sourced the entire JDK. Sun didn't do that and even .NET isn't entirely open source. I have no knowledge or involvement with EE, but it seems Oracle simply doesn't have faith in EE, and doesn't want the Java trademark associated with it.
(I work at Oracle on the JDK -- i.e. OpenJDK -- but speak only for myself)
I can say that my company’s use of java is diminishing rapidly over the last 5 years too.
No matter what you think of the technological advancement of the language and ecosystem- it matters not if people are not using it.
Oracle itself is a pariah, the vast majority of the tech industry avoid it.
Last time I had to download a JDK I got a warning about having to pay a license for commercial use. I think most enterprises care more about the licensing arrangement than access to the source.
Java suffers from having some 20-year old codebases because it's been so popular for so long but I can open up a 2004 WebLogic project in Eclipse or InteeliJ and trace everything it's doing. And modern Java has added all the nice features we expect a modern language to have. Java's slow demise is due to terrible PR and Oracle's endless insistence on JEE when every sane developer is using Spring.
But the build toolstack is a pure, unadulterated nightmare. I used to think Maven's XML orgies, archetypes and learning curve were bad, but diving into Webpack et al was a new dimension of horror. This time based on JSON, though.
Java has been the new COBOL since its inception in the 1990s.
I programmed quite a bit in COBOL early in my career. It's an extremely efficient language for scanning records especially in an era when a large proportion of data was stored offline on mag tape. Most common business problems of the era could be expressed and solved effectively in COBOL, which is why it became so popular.
I programmed in both C++ and Smalltalk in the early 1990s, which were innovative but also difficult to use for large systems. Java hit a happy medium with its C++-like object syntax, built-in libraries, and virtual machine operation with garbage collection. Add in the network effects from wide adoption (leading to tools and libraries) and it became a huge hit.
Oracle seems to be going out of its way to submerge the language in confusion. But nobody should mistake this as a reflection on the inherent capabilities of Java or its fitness for use.
Edit: correct typo
In fact, you described exactly why it is the new COBOL: because it is also an efficient and popular language for common business problems of our era, that is hugely popular, but mostly in "unexciting" places like large internal enterprise apps, out of sight for many. And just like COBOL, it will keep ticking on like that, probably forever.
Where I am (geographically, company wise and ecosystem) Java and the Java ecosystem are very much thriving.
I'd consider Spring as more of a non-JavaEE application server. Pick either Spring or Wildfly/Glassfish/Websphere/etc and you're either using part of the JavaEE specification with the former or the full specification with the latter.
Avoiding JavaEE for micro-services requires moving to something like DropWizard. Instead of a 200MB application server you end up with a 20MB application. Unfortunately, running that application in a Docker container will spin up a 180MB JVM. So the main reason we're moving away from JavaEE is actually more related to our move to container orchestration and that was originally driven by our adoption of micro-services. If you write a micro-service with a single end-point why should it use 500MB of memory?
Not in any industry I know.
Like C++, it's modern features are being slowly added. There's a huge difference between Java 8 and 11, in performance and features. Innovation is hardly dead.
Java EE and Spring are both terrible. I recommend trying some of the newer frameworks out there to see how much nicer things have become. DropWizard is my favorite right now.
I'm no Oracle fan, but Java is more "open" in many ways than it has been since the beginning. It wasn't till a few years ago we had a feature-complete high performance OSS implementation as the standard.
Oracle will try to ruin it, but the GPL JDK and history of projects being forked away from them makes it fairly safe. It's not in the interest of any Java user to have more than one standard.
IMO Java isn't going anywhere. It's been top 3 most popular language for over a decade. JS is somewhat promising but will never match Java in performance and multi-threading support due to core design decisions. Go is too verbose to get industry buyin to replace Java in most use cases. WASM is promising in the far future but it will take years to have a full browser and server ABI. The only real contenter to replace Java is C# and so far it's not getting much traction.
If anything killed Java it would be C#. Value types, no type Erasure, unsafe code and some native vector support. But it hasn't and probably won't ever dethrone Java. Java has a combination of decent features and an insanely huge amount of library support that makes it the default language of many companies.
When people say Ruby is dying, it would be more accurate to say they are not growing or they are shrinking. But given the size of Ruby marketshare of less than say ~2%. I would not really disagree with their notion.
But Java? I mean even from a Web Development prospective, given you mention Node, JS and Spring Framework, it is Far from dead.
And this is speaking as someone who likes Ruby and dislike Java.
At least it would be if Oracle wasn't trying to squeeze it dry.
And I think Java is more alive then Delphi, at least no one is setting up a site like http://www.isdelphidead.com/ :D
You and I may regard it as evil, but it's a perfectly valid strategy. Perhaps it's more common outside IT? It's the brutally honest business strategy.
I can sort of appreciate that they don't play around. Perhaps not my first choice of technology, but you certainly know where you have them.
Guess what, if you like typescript, you'll love Kotlin. It's a much cleaner language, just as expressive, and it has a whole range of things that typescript simply can't do or that it simply does better/more elegantly because typescript is stuck preserving compatibility with javascript and all it's weirdness. MS has done a wonderful job of shoving a lot of that under the carpet with typescript but fundamentally types are optional because all javascript has to be valid typescript and therefore it has to also allow all the silly hacks, misguided features, and other cruft that javascript people have been coming up with for the past decades.
You might also like C#, Rust, Elixir, and several other languages that each offer their own tools, libraries, and features. I'm seeing a pattern with many frontend developers in the last few years that they at some point become full stack developers and shortly after start adding more languages to their repertoire. I know former node.js developers doing Rust, Go, Elixir, Kotlin, Rust, etc. now. Most of them are not very eager to move back to doing node.js projects.
Like Kotlin all these languages are all coming to the node.js/browser ecosystem through WASM. Full stack in 2020 is going to be a very different beast than just a few years ago.
Right now WASM already allows lots of languages to be used for just about anything that people currently do with javascript and typescript in browsers or in the node.js ecosystem. As this becomes more mature, people are inevitably going to change the ecosystem. This is already happening. E.g. Rust is being used for many low level npms for a few years now. However people have also been developing frameworks and libraries in Rust that work around/replace that ecosystem entirely. For a Rust person dealing with crates is preferable to dealing with npms, webpack and all the other stuff that is used in that ecosystem. So, they are fixing that. The goal there is to run pure rust in wasm in the browser or on v8 with 0 npms pulled in or used during the build process.
Likewise, there's an ecoysystem of multi platform kotlin libraries emerging and that allows you to build wasm or native applications and libraries without needing things like webpack, npm, etc (but using gradle instead). Same with C# and blazor. Each of those ecosystems uses browsers and node.js as runtimes but try to minimize/eliminate/replace most of the rest of that ecosystem. In the same way the Kotlin ecosystem is also starting to provide alternatives to stuff from the java EE and Spring ecosystem. E.g. Ktor is a multi platform web application framework that you can use instead of spring boot and run natively or on the jvm. It's also a drop in replacement for anything you might want to do with e.g. express.js. Doing express.js applications in Kotlin is of course also possible, but why would you when you can have ktor instead? In the same way building react applications in Rust is possible, but replacing it with a rust specific framework is much less awkward.
So, I agree things are changing but I'd say the node.js/browser landsace is changing as well and will not involve a whole lot of javascript/typescript for people coming from the JVM or other ecosystems.
You might make the point that there are all these millions of npms out there that can't be replaced easily. I'd counter that with the point that an awful lot of them seem to have very short shelf lives and are in fact replaced/discarded regularly.
b) Android isn't switching to Kotlin because of licensing since it is a JVM language after all. It's purely because it is a much better language than Java and has great tooling courtesy of Jetbrains.
You can't compare these frameworks. They're in entirely different spaces, with different requirements and constraints.
But it is the right thing to do. Oracle is not trustworthy.
The fact is that every single one of the largest companies has some Java or Scala code within their core platform. And none of them have ever raised Oracle as a risk.
If Oracle had succeeded, would that mean that every developer that downloaded the Eclipse IDE would be subject to Oracle's licensing and a possible audit target? I believe that it would.
https://developer.ibm.com/javasdk/support/commercial-licensi...
Everyone should move off of Oracle's VM. I think Oracle is going to be more agressive in discuraging this or making it more difficult.
Anecdotally, nearly every CIO/CTO I’ve encountered in the last few sees Oracle as a huge risk. Most particularly with regards to databases, which are even sticker than applications which have shorter life cycles.
So Oracle's attitude to OpenJDK and the possibility that it cuts back on development is a likely factor for people considering the platform.
Yeah, Google thought the same
If Oracle's copyright argument holds up, then they can kill or cripple OpenJDK at will by denying it the ability to implement new Java APIs.
As somebody who's not deep in the Java ecosystem, the precise relationship is not clear to me, but it does seem that Oracle is a major contributor to OpenJDK at least, most likely also involved with the project steering.
Far as I understand having any Oracle products is a liability. They can't 'audit' you if you don't use any of their products.
I hope other organisations with a director from Oracle are taking note.
https://www.linuxfoundation.org/about/board-members/
In the EU the same is happening. Poland and Hungary have found a loophole allowing them to ignore EU rules and sanctions while still profiting from subsidies and market access.
All sanctions against a member state require unanimous-1 (all states except the sanctioned member) votes and Poland is vetoing all sanctions against Hungary, and Hungary all sanctions against Poland.
https://en.wikipedia.org/wiki/Liberum_veto
AFAIK this was tit-for-tat: HU & PL closed/sold-off their industry in favor of WEU in exchange for subsidies. Removing these subsidies for any reason would violate the main economic agreement between them and EU. Mind you, there is no significant locally grown industry in those countries anymore (outside of a few state-owned industries), WEU profited from this enormously, especially Germany. As economic potential of this arrangement has been exhausted after 25 years already, incorporating UA is the next move to give EU (primarily DE/FR/NL) some nice economy boost for the next 10 years. EU is essentially playing Sim City.
The Netherlands always vetos Romania's application to enter the Schengen zone as Romania's harbors to the Black Sea will provide cheaper alternatives for import of goods to the Dutch ones.
If a EU state refuses to do a thing, that's basically it.
https://piie.com/blogs/realtime-economic-issues-watch/france...
Take for example the Italian government's decision to run a deficit higher than what is allowed under EU rules (2% per year if I recall correctly). The EU tried to put a brave face on it: they held negotiations with the Italian government and announced a compromise, but it was pretty obvious to people who follow EU politics that the EU had no effective means to enforce the rule on Italy.
And this cap on deficit spending by governments of member nations is no minor rule. Many had stated over the years since the introduction of the Euro that the rule is necessary for the single currency to work.
But it was a lot of time wasted:
Apache Harmony/Java
LibreOffice/OpenOffice
MariaDB/MySQL
Jenkins/Hudson
now Eclipse/Java
Right now, they are in sales hurry mode, calling at all occasions technical contacts each time you create a new account to download something on their web site...
...
https://netbeans.apache.org/
https://cloud.oracle.com/containers/kubernetes-engine
...
In O/WebLogic, javax.* imports are by default private.
Open9J vs Graal
https://github.com/eclipse/omr/issues/1118
https://blogs.eclipse.org/post/mike-milinkovich/update-jakar...
Note: Oracle ECM, WebCenter Content has no simple Virtual Document solution.
Deleted Comment
WORA was never perfect, I would be surprised if there was even a single JRE release that didn't note a breaking change in its release notes. So vendors should have a process to update their releases just for that already in place.
> The future is unclear because Oracle already started a blocking politics at the Eclipse Foundation’s board, where Oracle has a seat, and where unanimous decisions are needed.
They have no rules to out board members acting in bad faith?
JVM is still the best platform for any type of back-end application. It has the widest array of libraries, is proven to run on any type of hardware and comes standard on every OS and every cloud instance.
.Net is a great platform and all but it still hasn't done enough to really merit switching teams. Especially when you have Scala, Kotlin, Clojure etc to pick from.
And a proper story for something similar to Swing, not even considering JavaFX here. There is Xamarin, but again, the platforms with Java compilers available.
And I am stating this as someone that works daily within .NET eco-system.
I think the problem is Oracle's a-hole, in this situation.
Do not work for, do business with, depend on, or attempt to negotiate with the lawnmower, either!
There is only 1 popular language released since 2009 that targets JVM: Kotlin. All other target JS, their own bytecode or LLVM: Go, Rust, Dart, Elixer, Julia, Typescript, Swift.
This is a significant difference from 2003-2009 languages which all targeted JVM unless they were Microsofts: Groovy, Scala and Clojure.
How will these languages ever catch up with innovations like JavaBeans, Java Server Pages, and Java Server Faces?
Dead Comment
You are going to upset all the Clojure fans with that :-)
But what would be the point in more JVM languages? I feel like a big factor in that is just that every niche is well filled with mature languages now.
Swift is iOS/Apple specific, at least in usage. But mobile is dominated by Android, which is moving towards Kotlin. So Kotlin is already huge right there, much larger than Dart or Elixir and probably larger than Swift, soon if not already.
Meanwhile Go would probably benefit from targeting the JVM, or at least having the option. As is, it's stuck in its own isolated world with low quality tools, libraries and GCs.
At any rate number of languages is an odd metric. Kotlin by itself is a good enough language to probably kill off most attempts to compete with it over the next 5-10 years or so on the JVM, outside of research languages. Does the JVM really need dozens of languages?
Look at new languages of 1990-2000: Haskell, Python, VB, Lua, R, Ruby, Java, Delphi, JS, PHP, Rebol. 4/11 are mostly irrelevant now. 4/11 are important niche languages. 3/11: Python, Java, and JS are dominant players.
1990-2000 might be special because of a massive shift toward IT systems in general, but it might not be and be an indication of languages lifecycle.
"Go needs a native library for ___" is used as an opportunity for safely tedious résumé-driven development. On the JVM you'd have to look for an unsolved problem which is probably harder.