I just hope with new lightweight threads I never have to write async reactive code again. It was a such an unproductive mistake, most applications do not need that level of complexity - now we can confidently say no applications need that level of added complexity.
Maybe add those links in the posts themselves. The first one mentions what the second one will be about, you could just turn that into a link. Or have a bullet list of all the links you add to each post in the series.
I really don't want to upvote this sentiment but I must anyway.
The fact is, despite Oracle being a menace to the tech industry, Java under their watch is thriving. Which is weird, because I don't know anyone who gives them money for Java. I'm genuinely curious who these companies are and what their incentives are!
I feel like that tide is changing in general. Not that two wrongs make a right, but don’t forget Microsoft once sold “Linux Licenses”, and then there’s the whole SCO vs IBM debacle. Lawyer driven revenue streams are falling out of style, unfortunately Oracle was a late bloomer. They’ve done an incredible amount of damage to their reputation.
Not everyone cares that much about their reputation outside FOSS circles.
Sun was aided by Oracle and IBM into the whole Java push during its early days.
Many apparently aren't aware of their roles into the whole Java history.
Oracle was the first RDMS with Java drivers, all GUI tooling rewriten into Java, JVM into the database, JSF famework, and acquisition of BEA and its JIT technology.
Also collaboration with Sun in the Network Computer thin client with its Java based OS.
Java gets more hate than it deserves. "There are languages everyone complains about and languages nobody uses."
Most of the hate comes from the overly complicated "enterprise design patterns" crap that took over the ecosystem in the late 90s into the 2000s, not the language itself. It's quite possible to write clean, clear, appropriately complex, well performing Java code.
On the plus side, of all the languages I've used Java is one of the absolute best when it comes to long term maintainability of code. This is why it's used so heavily in large enterprises with long-lived business critical code bases. Being the "COBOL of the 1990s/2000s" is not an insult, and as a language it is far superior to COBOL in every way. It's not a bad language to program in at all, while COBOL will make you hate your life.
It's also a safe language unless you break out of the JVM with JNI. It's the first safe language to get huge deployment if you don't count scripting languages. Safe doesn't mean you can't have security bugs of course, it just means you're not likely to have certain kinds of security bugs and stability problems like memory errors.
The JVM is really a fantastic piece of engineering and IMHO represents a whole direction in computing I feel sad that we didn't take. We opted to stay close to the metal with all the security, portability, code reuse, and other headaches that entails, instead of going into managed execution environments that make all kinds of compatibility and reuse and portability problems mostly go away.
The biggest current knock against Java I see is JNI, which unlike the core language is absolutely horrible. The second biggest knock is that the JVM is still kind of a memory pig. CPU performance is great, sometimes converging with C or Rust performance depending on work load, but it still hogs RAM.
In my experience, the problem of Java is the lack of standardized tooling.
To build an average Java software, you have to install a specific version of JDK, download a specific build system (Ant, Maven, Gradle, Bazel), hope everything works out on the first try - and if not, debug the most-likely-XML spec file searching for invalid dependency that's printed out on the 1000-line error output...
What Java is desperately missing is something like Python's `uv`.
---
Sibling comment mentioned that debugging Java itself is also a nightmare, which reminds me of the many Spring Boot projects I've had to debug. Stack traces containing 90% of boilerplate code layers, and annotations throwing me from one corner of the codebase to another like a ragdoll...
Admittedly, that's not inherently the problem of Java, but rather the problem of Spring. However, Spring is very likely to be found in enterprise projects.
I think this is really just anecdotal to your experience. Don’t you need to install a specific version of a compiler or interpreter for every language? Isn’t trying to build any complex project a pray on the first try? I’ve worked in Go codebases where it’s not simply “go build”. It’s: try go build, oops you need to use the make/justfile, oops don’t forget you need to install jq or some other random tool. Complex projects are complex.
I totally disagree, please don't add something like envs to Java, ever. Working with Python is a huge pain because of that.
Also, mostly just a JDK that is new enough, because of the strong backwards compatibility of the Java ecosystem.
> The biggest current knock against Java I see is JNI, which unlike the core language is absolutely horrible.
JNI was only ever designed to be good enough, and it is. The new FFM API aims to replace JNI in most cases, but it's designed to be "perfect". As a result, the new API took many years to develop, but JNI was quick to develop.
It would be nice to have the FFM API much sooner, but alternatives like JNR and JNA have been around for a long time. There wasn't a rush to develop a JNI replacement.
I have mostly heard these complaints from people who haven't used a java ide and/or do not know that jvms allow waiting for a debugger before starting execution (helps with all sorts of spring or whatever errors and boilerplate).
That said, there is a shitload of "enterprise" fuckery in Java, but those Devs would have made a mess of any codebase anyway.
Surely this is written by an LLM.
Paying per core for “enterprise” just because you’re a business isn’t my idea of being a good steward. If anything we should be championing the OpenJDK folks. They are the real heroes.
The only thing that's paid is support, and the OpenJDK folks are Oracle employees (well, the ~90% of them who do ~95% of the work on OpenJDK). OpenJDK is an Oracle project in the same sense that Chromium is a Google project. In fact, OpenJDK (even more precisely - the OpenJDK JDK) is the name given to Oracle's implementation of the Java SE specification, but we do get contributions from other companies, such as this particular great enhancement to JFR (even external contributions also involve significant work by Oracle employees).
Anyway, if you don't want to buy a support service, either from Oracle or any of the other companies that sell it, the use of the JDK is free. There is no "enterprise" flavour of the JDK, paid features, or use restrictions as there used to be under Sun's management. Java is obviously freer now - as in beer or in speech - than it was 20 years ago.
GraalVM is a separate product developed by an unrelated team. Its enterprise flavour is not considered an enterprise flavour of the JDK. The closest to an enterprise JDK from Oracle I can think of is the "Enterprise Performance Pack" for the 12-year-old Java 8, but it has nothing that isn't in the free and open recent releases (which actually include many more performance enhancements).
The idea there is that it's cheaper for companies with legacy software that isn't actively maintained to pay for some portion of the performance improvements in modern JVM generations than to ramp up maintenance to upgrade to modern Java, and this can help fund the continued evolution of OpenJDK.
What rug pull do you picture could happen at this stage? OpenJDK is the reference spec. Fully open source. Stewarded by multiple companies. Even if Oracle somehow managed to force the whole thing closed source (not sure that's even possible?) you've got all the other contributors who'd "hell no", fork and away you go. Which version of Java do you think the community would go with? There's no way it'd work.
I never thought I would be excited for a new release of Java, but ever since Java 21, I have grown to actually enjoy writing the language. Whomever is running it has really done a good job making the language actually fun to write in the last few years.
Mark Reinhold runs it, having been there since 1.x days as the core tech lead. He also pushed the move to open source and fast release cycles. But mainly he seems to have recruited and developed good people. I can’t think of a more enduring lead facing more challenges, technical and organizational.
Modern tracing-moving GCs offer very efficient memory management algorithms that are often hard to beat on performance. Their price isn't so much performance anymore as memory footprint, while manual memory management (and refcounting GCs) optimise not for performance but for footprint (often at the cost of performance). Even this tradeoff is often misunderstood, as covered by this brilliant keynote from the most recent International Symposium on Memory Management: https://youtu.be/mLNFVNXbw7I
There isn't such thing, if you mean Rust, affine types systems require tree structures and have issues with multiple scopes, hence the memes with borrow checker.
Which I would refer as compiler assisted, although not really a proper term.
There is a reason outside Rust, everyone else is string to combine GC alongside affine/linear/effects/dependent types instead of one solution for everything.
The productivity of having a GC, with the type system capabilities to go lower level, when the profiler says so.
Although I have to conceded, Rust made a great job bringing ATS and Cyclone ideas into mainstream.
Additionally with AI powered languages, naturally most of this will be automated anyway.
As refcounting and tracing are the two primary starting points for garbage collection, they're its two pillars so to speak (e.g. https://web.eecs.umich.edu/~weimerw/2012-4610/reading/bacon-...), I don't know of any serious person in the memory management space who doesn't see refcounting as a GC, although some languages with a refcounting GC say "they don't have a GC" in their marketing material.
As a general rule, tracing seeks to optimise for throughput while refcounting seeks to optimise for footprint, although the two can become more elaborate and converge (as shown in the paper I linked).
Maybe not how you were brought up, and people argue... But Watson & Watson, back in '87, described ref counting as "An efficient garbage collection scheme for parallel computer architectures".
Even the dragon book refers to ref counting as a form of garbage collection.
Line 2 of Ref counting's Wikipedia page refers to garbage collection.
People have argued since the invention in 1960, but... Ref counting is one form of GC. [0] It might not be a tracing GC, but it is GC.
I just hope with new lightweight threads I never have to write async reactive code again. It was a such an unproductive mistake, most applications do not need that level of complexity - now we can confidently say no applications need that level of added complexity.
(author here) This is part of a series of blog post. The next blog posts are: https://mostlynerdless.de/blog/2025/07/30/java-25s-new-cpu-t... (on the implementation) https://mostlynerdless.de/blog/2025/08/25/java-25s-new-cpu-t... (on sizing the queue) and a fourth on improving the performance: https://mostlynerdless.de/blog/2025/09/01/java-25s-new-cpu-t...
Maybe add those links in the posts themselves. The first one mentions what the second one will be about, you could just turn that into a link. Or have a bullet list of all the links you add to each post in the series.
Thanks for the recommendation.
This is built on a Linux sampling API that offers more accuracy, but it is still sampling.
For CPU tracing, with no sampling errors, use Apple’s M4 with the latest Xcode’s Instruments.
(author here)
Xcode doesn't know about Java's internals, so it doesn't know about Java frames, although it can help with native traces.
The JVMin in the last 6-8 years has been a powerhouse of innovation and cool features. Incredibly impressive!
And a thank you to Oracle for being a good steward of the language.
I really don't want to upvote this sentiment but I must anyway.
The fact is, despite Oracle being a menace to the tech industry, Java under their watch is thriving. Which is weird, because I don't know anyone who gives them money for Java. I'm genuinely curious who these companies are and what their incentives are!
I feel like that tide is changing in general. Not that two wrongs make a right, but don’t forget Microsoft once sold “Linux Licenses”, and then there’s the whole SCO vs IBM debacle. Lawyer driven revenue streams are falling out of style, unfortunately Oracle was a late bloomer. They’ve done an incredible amount of damage to their reputation.
Not everyone cares that much about their reputation outside FOSS circles.
Sun was aided by Oracle and IBM into the whole Java push during its early days.
Many apparently aren't aware of their roles into the whole Java history.
Oracle was the first RDMS with Java drivers, all GUI tooling rewriten into Java, JVM into the database, JSF famework, and acquisition of BEA and its JIT technology.
Also collaboration with Sun in the Network Computer thin client with its Java based OS.
For real. When Sun got bought, I thought to myself: “At least they would also destroy Java…”.
But nooo, Java thrives and flourish under Oracle protection.
Java gets more hate than it deserves. "There are languages everyone complains about and languages nobody uses."
Most of the hate comes from the overly complicated "enterprise design patterns" crap that took over the ecosystem in the late 90s into the 2000s, not the language itself. It's quite possible to write clean, clear, appropriately complex, well performing Java code.
On the plus side, of all the languages I've used Java is one of the absolute best when it comes to long term maintainability of code. This is why it's used so heavily in large enterprises with long-lived business critical code bases. Being the "COBOL of the 1990s/2000s" is not an insult, and as a language it is far superior to COBOL in every way. It's not a bad language to program in at all, while COBOL will make you hate your life.
It's also a safe language unless you break out of the JVM with JNI. It's the first safe language to get huge deployment if you don't count scripting languages. Safe doesn't mean you can't have security bugs of course, it just means you're not likely to have certain kinds of security bugs and stability problems like memory errors.
The JVM is really a fantastic piece of engineering and IMHO represents a whole direction in computing I feel sad that we didn't take. We opted to stay close to the metal with all the security, portability, code reuse, and other headaches that entails, instead of going into managed execution environments that make all kinds of compatibility and reuse and portability problems mostly go away.
The biggest current knock against Java I see is JNI, which unlike the core language is absolutely horrible. The second biggest knock is that the JVM is still kind of a memory pig. CPU performance is great, sometimes converging with C or Rust performance depending on work load, but it still hogs RAM.
In my experience, the problem of Java is the lack of standardized tooling.
To build an average Java software, you have to install a specific version of JDK, download a specific build system (Ant, Maven, Gradle, Bazel), hope everything works out on the first try - and if not, debug the most-likely-XML spec file searching for invalid dependency that's printed out on the 1000-line error output...
What Java is desperately missing is something like Python's `uv`.
---
Sibling comment mentioned that debugging Java itself is also a nightmare, which reminds me of the many Spring Boot projects I've had to debug. Stack traces containing 90% of boilerplate code layers, and annotations throwing me from one corner of the codebase to another like a ragdoll...
Admittedly, that's not inherently the problem of Java, but rather the problem of Spring. However, Spring is very likely to be found in enterprise projects.
I think this is really just anecdotal to your experience. Don’t you need to install a specific version of a compiler or interpreter for every language? Isn’t trying to build any complex project a pray on the first try? I’ve worked in Go codebases where it’s not simply “go build”. It’s: try go build, oops you need to use the make/justfile, oops don’t forget you need to install jq or some other random tool. Complex projects are complex.
> What Java is desperately missing is something like Python's `uv`.
JBang exist and (if I'm not mistaken) predates uv. See jbang.dev
We don't need Rust written tools in Java.
I totally disagree, please don't add something like envs to Java, ever. Working with Python is a huge pain because of that. Also, mostly just a JDK that is new enough, because of the strong backwards compatibility of the Java ecosystem.
> The biggest current knock against Java I see is JNI, which unlike the core language is absolutely horrible.
JNI was only ever designed to be good enough, and it is. The new FFM API aims to replace JNI in most cases, but it's designed to be "perfect". As a result, the new API took many years to develop, but JNI was quick to develop.
It would be nice to have the FFM API much sooner, but alternatives like JNR and JNA have been around for a long time. There wasn't a rush to develop a JNI replacement.
I hate Java because debugging java code is worse than debugging assembly
As a Intellij-user I am more than a little confused. What are features missing from Java's debugging story?
Yeah this comment is wild to me. Java’s debugging is so good I can debug a remote server from my local machine.
I have mostly heard these complaints from people who haven't used a java ide and/or do not know that jvms allow waiting for a debugger before starting execution (helps with all sorts of spring or whatever errors and boilerplate).
That said, there is a shitload of "enterprise" fuckery in Java, but those Devs would have made a mess of any codebase anyway.
I hate to say it, but this is because you are not a Java developer. It’s very easy to debug.
Found out another printf debugger.
And a thank you to all the other amazing companies. The work on the JEP is chiefly sponsored by SAP (with help from Datadog and Amazon).
Surely this is written by an LLM. Paying per core for “enterprise” just because you’re a business isn’t my idea of being a good steward. If anything we should be championing the OpenJDK folks. They are the real heroes.
The only thing that's paid is support, and the OpenJDK folks are Oracle employees (well, the ~90% of them who do ~95% of the work on OpenJDK). OpenJDK is an Oracle project in the same sense that Chromium is a Google project. In fact, OpenJDK (even more precisely - the OpenJDK JDK) is the name given to Oracle's implementation of the Java SE specification, but we do get contributions from other companies, such as this particular great enhancement to JFR (even external contributions also involve significant work by Oracle employees).
Anyway, if you don't want to buy a support service, either from Oracle or any of the other companies that sell it, the use of the JDK is free. There is no "enterprise" flavour of the JDK, paid features, or use restrictions as there used to be under Sun's management. Java is obviously freer now - as in beer or in speech - than it was 20 years ago.
There is an enterprise flavor of the JDK. It's called GraalVM enterprise edition.
That has nothing to do with OpenJDK, GraalVM is its own thing.
You’re the only one who put “Open” in there. Both your parent and grandparent said JDK.
GraalVM is a separate product developed by an unrelated team. Its enterprise flavour is not considered an enterprise flavour of the JDK. The closest to an enterprise JDK from Oracle I can think of is the "Enterprise Performance Pack" for the 12-year-old Java 8, but it has nothing that isn't in the free and open recent releases (which actually include many more performance enhancements).
The idea there is that it's cheaper for companies with legacy software that isn't actively maintained to pay for some portion of the performance improvements in modern JVM generations than to ramp up maintenance to upgrade to modern Java, and this can help fund the continued evolution of OpenJDK.
Because people keep forgetting Java is like C and C++, there are plenty of JDKs to chose from, and not all of them are related to the same codebase.
Zulu from Azul might be one? I think it comes with a difference JIT and GC in the enterprise version.
OpenJDK is the specification implementation. A huge amount of the OpenJDK development is paid for by Oracle (And others).
Because they have a financial interest in rug pulls.
What rug pull do you picture could happen at this stage? OpenJDK is the reference spec. Fully open source. Stewarded by multiple companies. Even if Oracle somehow managed to force the whole thing closed source (not sure that's even possible?) you've got all the other contributors who'd "hell no", fork and away you go. Which version of Java do you think the community would go with? There's no way it'd work.
Why do you think a good steward shouldn't have a financial interest?
Nope, people keep forgetting no one wanted to buy Sun, not even Google after torpedoing it (which would save them from their J++ like lawsuit).
IBM kind of thought of it, but ended up withdrawing the offer.
So the anti-Oracle folks would have seen Java wither and die in version 6, and the MaximeVM technology would never had been released as GraalVM.
`sdk install java 21.0.8.fx-librca`
No pre-core fee needed.
I never thought I would be excited for a new release of Java, but ever since Java 21, I have grown to actually enjoy writing the language. Whomever is running it has really done a good job making the language actually fun to write in the last few years.
Mark Reinhold runs it, having been there since 1.x days as the core tech lead. He also pushed the move to open source and fast release cycles. But mainly he seems to have recruited and developed good people. I can’t think of a more enduring lead facing more challenges, technical and organizational.
The more I search for a new language to learn the more I want to go back to Java. I feel so nostalgic :)
I have repeatedly said this in multiple occasions: You start to really appreciate Java after you work with other languages.
If you care about performance at that level, you should not be using Java or any other language with a GC.
Modern tracing-moving GCs offer very efficient memory management algorithms that are often hard to beat on performance. Their price isn't so much performance anymore as memory footprint, while manual memory management (and refcounting GCs) optimise not for performance but for footprint (often at the cost of performance). Even this tradeoff is often misunderstood, as covered by this brilliant keynote from the most recent International Symposium on Memory Management: https://youtu.be/mLNFVNXbw7I
Thankfully not everyone agrees.
https://www.ptc.com/en/products/developer-tools/perc
We have already have had enough from anti-GC cargo cult from "manual memory management is great" folks.
Having a GC (which RC is also an algorithm subset), doesn't preclude having other features.
There is also non manual, non GCed manual management.
I agree that a GC can be a viable implementation of memory management though.
There isn't such thing, if you mean Rust, affine types systems require tree structures and have issues with multiple scopes, hence the memes with borrow checker.
Which I would refer as compiler assisted, although not really a proper term.
There is a reason outside Rust, everyone else is string to combine GC alongside affine/linear/effects/dependent types instead of one solution for everything.
The productivity of having a GC, with the type system capabilities to go lower level, when the profiler says so.
Although I have to conceded, Rust made a great job bringing ATS and Cyclone ideas into mainstream.
Additionally with AI powered languages, naturally most of this will be automated anyway.
Ref counting is also no GC by the definition of a GC I was brought up with.
But yes, I was also thinking about the borrow checker.
Not necessarily in Rust, as other languages start to adopt similar techniques.
As refcounting and tracing are the two primary starting points for garbage collection, they're its two pillars so to speak (e.g. https://web.eecs.umich.edu/~weimerw/2012-4610/reading/bacon-...), I don't know of any serious person in the memory management space who doesn't see refcounting as a GC, although some languages with a refcounting GC say "they don't have a GC" in their marketing material.
As a general rule, tracing seeks to optimise for throughput while refcounting seeks to optimise for footprint, although the two can become more elaborate and converge (as shown in the paper I linked).
Maybe not how you were brought up, and people argue... But Watson & Watson, back in '87, described ref counting as "An efficient garbage collection scheme for parallel computer architectures".
Even the dragon book refers to ref counting as a form of garbage collection.
Line 2 of Ref counting's Wikipedia page refers to garbage collection.
People have argued since the invention in 1960, but... Ref counting is one form of GC. [0] It might not be a tracing GC, but it is GC.
[0] https://dl.acm.org/doi/10.1145/367487.367501
Then somehow someone has taught you badly, any proper CS reference book in academia has it, like the famous GC Handbook,
https://gchandbook.org/
The other languages adopting such type systems, also keep automatic resource management around for productivity.
The ones that don't, are still in the looking for meaningful adoption stage.
Most academic definitions of garbage collection include reference counting. It’s just a form of it.
https://gchandbook.org/contents.html
There are ways to mitigate this, the fly weight pattern, etc.