The Future of Java SE [video] | Hacker News Reader

The Future of Java SE [video]

youtube.com

76 pointsjavinpaul73 comments

The end of this year is going to be such a dumpster fire for us. I think I've got about a hundred developers with a mix of Linux, Mac, and Windows. The corporate overlords are mostly sticking to JDK 1.8, which is fine -- RHEL has a LTS for that on Linux. The developer desktops are all over the place right now, however. The jump from 8 to 9+, for those of saddled with legacy crap code, has been somewhat painful. Got to wonder how many years I'll be seeing a 8u191 (or whatever the last release is) on the windows/osx boxes.

Curious, Is the adoptjdk or openjdk not an option for some reason?

I'm looking forward to seeing what they do but right now my impression is that they're still getting on their feet.

If you go to the page an select builds from the drop down half of them weren't available last I checked. They're new so I'd give them some space for maturation.

OpenJDK is on the Linux servers and workstations. The options for OSX, for example, are less clear. Folks are trying to see what non-Oracle option might work on desktops/laptops.

I wish they would add tail calls to the VM. It would really help functional programming. It's doable but apparently not considered a high priority.

We plan to they are on the list of things that will be looked at as part of project loom. It was originallly thought they would be closely related to continuations as they are both forms of non local return, but this hasn’t been the case so far.

protip: jump to minute 31 for the interesting stuff.

The first 30 minutes is about the new release cycle, which is... hm.. quite dry, boring and covered extensively elsewhere.

Yea, its death by gantt chart

Interesting road map - some features seem like would be moving java closer to go (walhalla, loom, Panama)

Yes indeed. Hopefully with the jdk modularization (Project Jigsaw) done and out and GA, efforts will finally pick up steam on those modernization projects.

Add GraalVM native image capability to that, and Java will be back to leading the pack again.

Imo, The most important part is earning developer trust more than the feature set.

The roughly decade long copyright battle around Java API use has killed any interest I had in Java

Why? (Or why should we care about your personal interest in java without any substantial information?)

I don’t believe in software patents, and Oracle’s litigious history indicates they are a strong proponent for software patents.

I loved working with Java/Spring at one time, but there are too many good alternatives and I choose to vote with my feet.

FYI, the company that extracted the most from Android through patent-related legal threats was Microsoft, that collected billions in patent royalties from Android vendors[1][2]. While more extensive, it generated less noise partly because the companies MS came after didn't have the same PR machine as the one Oracle came after... (a cynic would say that the result of the case was what would be expected from a battle between a company with a great legal department and a company with a great PR department).

Large software companies (except MS) are far more often the target of patent infringement lawsuits/threats than the perpetrator.

[1]: https://www.forbes.com/sites/ewanspence/2015/11/01/microsoft...

[2]: https://www.howtogeek.com/183766/why-microsoft-makes-5-to-15...

Microsoft has recently pledge their entire patent portfolio to the Linux system. That's not just the Linux kernel, but other packages built on it. I see this as a big change for them, since they need Linux on Azure.

It's not about the money. it's about the principle. For example, where does the copyright lawsuit leave Apache Harmony legally, given that it's an "unauthorized" Java API implementation? Is the codebase's Apache license still valid? If not, what does that say about the GPL license of GNU/Linux, which is similarly an "unauthorized" reimplementation of proprietary Unix?

There was nothing in the lawsuit about being authorised or not. The court ruled that Google's particular use of the Java APIs was not fair use, as it wasn't intended for interop[1]. Apache Harmony can only license what they own, and they don't own the Java API. I guess that if you use Harmony in a way that makes fair use of the APIs you should be fine, but that's not what Google did (at least according to the court's ruling).

[1]: https://en.wikipedia.org/wiki/Oracle_America,_Inc._v._Google....

For me, it is because if I ever wanted to embrace and extend Java, even for a personal project, I would feel that I was putting myself into a state of legal jeopardy.

You can argue all you want how I would be or I wouldn’t be, or that I would be fine as long as I didn’t call it Java, but as long as oracle remains super-litigatious...

I would rather put my efforts elsewhere.

(Would love to do or learn more about Graal, too, but... yuck. What am I going to get sued for if I re-implement any of it? No thanks.)

Even at the height of the court trial, even Google itself continued to use and fork OpenJDK (they rely on it heavily) and talk about it at Oracle[1]. OpenJDK is governed by its open-source license, which is GPLv2 (with the classpath extensions) -- just like Linux.[2]

[1]: https://video.oracle.com/detail/videos/featured-videos/video...

[2]: http://openjdk.java.net/legal/

That's silly.

OpenJDK has been around for years.

There's also Kotlin.

I tried to compile the JVM 9 on a Debian this weekend, just to give it a try. The openjdk's mercurial repository returned 500 HTTP-errors 9 times out of 10. After relaunching the clone command 50 times, I managed to fetch the sources. Total size of the source code with sub-projects dependencies : 1.5 GB.

After that you have to further download a pre-compiled JVM to bootstrap the compilation, and install further debian packages such as X11 stuff (???). I looked at the Debian package's Makefile from the package maintainers, and then I gave up.

The complexity, and especially the dependencies, of the JVM are outstanding. I guess some work groups are paid to work on Java, but it seems some sub-projects are staling, while some modern expected features are still not in the JVM. For example, it's difficult to see where GraalVM is in the grand scheme. Is it replacing several previous project? Is it competing with ongoing sub-projects? Will it be compatible with most projects?

The problem I see with Java is : if people are not paid to work or develop it, I'm not sure people would continue to support the language and the stack, given the time and specific skills you have to invest in it.

At that point, Java looks more like a research project than a production-ready tool.

I think that hosting the jdk sources using a bunch of Python scripts on top of mercurial was a big mistake, or perhaps it was not a mistake when Java was opensourced more than 10 years ago, but it's a mistake now to not use gitlab/github. As you said the mercurial repository as stability issues, and when it works, it's super slow.

Java need a previous version of java to bootstrap because javac is written in Java. And you don't need x11 if you build the headless version.

For Graal, the goal is to replace c1 and c2 (the two JITs used by Hotspot and written in C) by Graal in order to have only one JIT written in Java. But Graal is competing with c2 which has tuned over two decades, so currently you can choose if you want to use c1/c2 or c1/Graal when running Java. We know that Twitter is running Graal in production, i've no more info on that.

I'm 27 years old, I do not develop software on a professional level, but my Master degree in C.S. and my knowledge are enough to be able to download and compile Rust, Go, Python or Julia on my machine in less than an hour for each (Julia taking a bit more time because it even compiles LLVM as a default, but it was as easy as git clone; get_dep.sh; configure; make, and some scientific libraries deps). Even the Linux kernel is rather well self-contained.

I would like to be able to hack on Java like I can on any of those above languages, but "JVM" is actually a too big of a project containing way more than the necessary, which is, I guess, a legacy from the business model and organization of Oracle. Today, there are ant, maven, Gradle, mercurial, JVM, 2 GCs, JRE, JDK, headless versions, tons of codenamed other projects, a giant README and a lot of documents to read.

If "Java" was just a single _lean_ compiler and standard libraries, easy clone; configure; make, like Go or Rust are, it would be more approachable for new people, it would give a better first impression.

That being said, I'm impressed by the academic work of people working on the GCs and the JIT technologies of Java and JVM.

If you've built Go from source, how were you so surprised about the bootstrapping requirement? Go is also bootstrapped and you need to install a binary Go compiler to build it.

In fact, you probably noticed I tested most trending compilers and they are all bootstrapped this way. But none of them is 1.5GB before the compiled compiler is downloaded. My reaction was more about the unexpected overall size and process to get a basic Java compiler built from scratch. In all the above, the binary compiler can be automatically downloaded as part of the dependencies (from their own servers, not packaged for any distribution in particular), which make things simpler.

Edit : Maybe other compilers were lighter because I cloned with --depth 1 on git. I don't know the equivalent on mercurial, so it downloaded more than necessary

> The openjdk's mercurial repository returned 500 HTTP-errors 9 times out of 10.

I don't know why that would happen - that's a shame. They're moving to GitHub soon though.

> Total size of the source code with sub-projects dependencies : 1.5 GB.

Is that a problem? How large are most other large, long-lived projects' repositories? Linux is also several GB, for comparison.

> After that you have to further download a pre-compiled JVM to bootstrap the compilation

Yes it's bootstrapped. Just like most other programming languages. Is this a problem?

> and install further debian packages such as X11 stuff (???).

Java includes windowing libraries. Is this a big problem?

> I looked at the Debian package's Makefile from the package maintainers

What were you looking for in there? You can just build Java without trying to build Debian packages.

> The complexity, and especially the dependencies, of the JVM are outstanding.

I really don't think this is the case compared to any other large systems project. What was the real problem you found? It needs a binary to bootstrap it? Yeah like almost all languages. It's a bit large? It's been around for decades. It provides windowing libraries? Wow what a crime.

> I guess some work groups are paid to work on Java

Hundreds. Hundreds of people are paid to work on Java full time, at many companies. I would guess probably more people work on Java implementation than any other programming language.

> if people are not paid to work or develop it...

Let me stop you there, because they are.

> At that point, Java looks more like a research project than a production-ready tool.

It demonstrably is production-ready. Massive companies use Java in production in enormous, complex, demanding deployments, and have been for decades.

Please see my other reply for some details of some of the points.

> It demonstrably is production-ready. Massive companies use Java in production in enormous, complex, demanding deployments, and have been for decades.

It's true, it powers a good part of the Android ecosystem, and probably a lot of stuff we use everyday, but Google has massive ressources!

Given the momentum of Java, and with the rise of GPGU, Tensor DSPs, accelerated hardwares, containers stuff, Kubernetes or maybe even new programming paradigm around AI, will Java be able to keep up and find its way? What about Scala, Clojure, etc

This is only my point of view, but C and C++ still look more future proof for speed, and everything else in Python or Go is good enough. I understand many investments were made by many companies on the JVM and at least one of its language tho

You were unlucky, the hg server went down during the weekend http://mail.openjdk.java.net/pipermail/jdk-dev/2018-October/...

This doesn't happen often.

It's definitely not a research project. There are many large companies who contribute to OpenJDK, for example Oracle, SAP, Red Hat, SAP, Google and Amazon

https://blogs.oracle.com/java-platform-group/building-jdk-11...

Project Loom looks pretty interesting!

I wish the future of java was less Oracle.

It is and it has been? They've open sourced a ton just this last year removing most differences between OracleJDK and OpenJDK. I think it's just a few small compatibility variences left.

Can you elaborate? I don't understand what you mean.

Oracle contributes patches to OpenJDK, for instance lambda support.

Do you want less of that? Are the commits made by Oracle employees especially bad?

It's not like you need to use the Oracle JDK. There are many different distros, just like Linux.

already porting everything away from Java

any GOOD reason why porting existing stuff that works ?

Doubt it honestly. Seems to me like the main group of people up in arms about it are those who never did serious Java development anyways.

Comments on HN vary very widely in quality, but articles about Java or the JDK currently seem to attract particularly passionate low-quality comments. Does it somehow tickle the same neurons that Microsoft did back in the days of Micro$haft Winblowz 95?

If you think Java has it bad, I invite you to read every thread about C++.

PHP projects also get a lot of hate from people that have no idea about PHP, they just learned that some noobs wrote bad code with it, it is not a pretty or cool language but only bad developers would port existing project from PHP just because the language is not cool.

Even if the hate is poor at least C++ gets quasi-technical hate. 90% of low effort complaints on HN are about Oracle v Google. I'd take morning and groaning about GC or boring verbosity language over that in a heart beat.

Because Java is an ecosystem that is dying, and the sooner you move away from it the less pain you will have in the long term.

So if you have years of developers hours n a Java project you throw it away and start over with cool new tech that may also be not cool when your port is done?

As long as the current code works, you can add new features and fixes to it, there is no reason to change. For new projects you should use teh best tool for the job.

No, not necessarily the latest “cool tech”, just something other than Java. Just because you’ve spent a lot of money and time on the Java platform doesn’t mean continuing to throw resources at it isn’t a example of the sunk cost fallacy at work.

Sunk Cost Fallacy? Sounds like the fallacy fallacy! [0][1]. Joking aside, that's incredibly misapplied.

More seriously, I'm curious where you work or what parts of the industry you're privy to. It's everywhere from where I'm standing. Tons of companies are built on millions of jvms. LinkedIn, Google, Twitter, Amazon, etc. I know personally but I'm sure every third other tech company in the valley at least too.

[0]: https://rationalwiki.org/wiki/Fallacy_fallacy [1]: https://en.m.wikipedia.org/wiki/Argument_from_fallacy

I'll take your fallacy fallacy and raise you a fallacy fallacy fallacy.

Just to remind you, there was an entire industry built around Flash and Silverlight too. The people responding to me seem to think I'm advocating a quick rewrite (whatever that means), but all I'm saying is that from where I stand Java will not last in the long term without serious changes within Oracle. It would be better to have a transition plan now than no transition plan on a dying platform.

I was sort of expecting to be flamed, but I can assure you I'm not interested in revealing to flamers where I work...

Curious, what change within Oracle do you want to see?

Java is still free, so it can't be that.

So your boss comes to you and says that the users want some new feature X, you will say that it will take your 3 years since you need to rewrite everything in Rust or Go and I would say I can do it by using library or CLI app X and integrating in our project, it could take a few days.

What do you think will happen? Do you think that a sane boss will approve you wasting resources to port it to cool tech X instead of fixing the current needs. The only projects that have no choice and must do a porting are Flash based projects because soon will be probably impossible to get regular users to have Flash in a browser.

That's kind of my point really. When, as I believe is going to happen, Java inevitably becomes unusable, unmaintainable or costly to license (we are talking about Oracle here) then a new project to replace the existing system will inevitably be commissioned.

All I am saying is that good management is to look at the risks of sticking with Java and find a migration path earlier, because the longer you wait the more expensive it becomes to move away. And one day you will find you have no choice, and then it will be very expensive.

It is possible to maintain and even add essential features to a legacy system whilst migrating to a new system. Happens all the time.

The probloem is that you do not understand Java ecosystem and you are spreading FUD,

First of all you have OpenJDK that is part of Debian, everyone knows that Debian is very strict about licensing so if it is part of Debian there will be nothing that will make Java retroactively unusable without a paid license.

I never mentioned licensing.

Then what is your point, what is the reason to port? Everyone can keep running the existing Java projects with the current JVM and the current libraries for a long time.

I am not sure what your work experience is but I worked in small companies that use boring old tech, maybe it would be cool to use some cool language or cool framework but then you still get left behind when the language has a new incompatible version or the frameworks have a new version, the purpose is to keep the project working for the users, we not fix what is not broken.

To what if don’t mind asking?

I'm moving to Go, as much as I dislike the language. Java is bulky. I need to buy servers from Linode, but on a budget. Because Java 11 is not supported on the alpine docker, the apps take up 512 MB even with stack set to 128k and heap at 64m. I can do the same systems in Go for a fraction of the sitting still memory consumption. I also get better performance/MB of RAM.

Add to this that Spring is becoming a sprawling behemoth with massively outdated tutorials lurking in the top results of Google. At this point I wanted something that was easy, GC'ed, with good tooling. That's Go. DLang looked promising, but it's not easy. It also suffers from the out-of-date tutorials and books issue (my D book still says I have to pick between Pheobos and Tango collections). C++ seemed like overkill. Go provides a simple binary with a simple runtime with a simple RAM pattern.

I do like Java. I know it well. I'm literally typing this with a Java Ring on (https://www.ebay.com/itm/JAVA-RING-RARE-Sun-Microsystems-JAV...). Sadly, the JVM at this point doesn't fit my needs. Go is pretty terrible in some of aspects, errors and the way the community decides to carve files up, and the annoying way functions on structures float freely within a file, but I couldn't find another language that compiled to a binary, was simple, and had good library support for threading (so OCAMLs out)and RabbitMQ.

I suppose you ruled out Rust because of the lack of simplicity?

The choice for between them is like asking, "Do I start my project today or do I learn Rust and start it in 3-6 weeks?"

If you don't already know Rust I'd never recommend it for a project from a "your requirements meet its strengths" perspective. Python or Go you can pick up as you go pretty easily. Even Java can annoying if you're new just getting set up with Maven and Intellij.

Rust is on my "Things I'm Excited to Sit Down and Learn" list but it's not something you'd pick up in 2 days. Learning Rust is a project unto itself.

What about .NET Core?

Three things. Core, to my knowledge doesn't compile. It has an option to bundle the VM, but I've never played with it much. Two, I solely run Linux. My understanding is that Linux is still the redheaded step child of MS. Finally, I don't like MS. This is a bias that is waning with my use of VSCode. I don't want to spend the money on the licenses. Also, I got bit by a compiler error in VB.net where branches of IFs were ignored. You'd enter the first block (if true), then fall into the second (else). All that combined soured me on MS.

There's a lot of work being done on making Java more container friendly. E.g., Project Portola [1] (Java on Alpine Linux), jlink and modules, and Valhalla would likely reduce memory consumption. You can even compile some Java apps AOT these days with native-image. Plus, there are plenty of lightweight alternatives to Spring.

[1]: https://openjdk.java.net/projects/portola/

After almost 20 years, I've too moved away from Java for personal projects years ago. But I'm doing this just because Java has become boring, and I've always preferred natively compiled languages anyway. I can't see anything wrong with how Oracle has handled this. OpenJDK is the reference implementation, and is licensed under the GPL with classpath exception. What else is Oracle supposed to do?

Relevant - http://mcfunley.com/choose-boring-technology

Reply on news.ycombinator.com