Fast JSON and Furious Runtime

An airhacks.fm conversation with David Kral about: Helidon JSON and Helidon JSON Binding, compile-time JSON binding with annotation processors generating readable Java source code, no bytecode manipulation and no reflection, debuggable generated converters, comparison with Micronaut compile-time generation, why Helidon avoids the JSON-B API and its reflection overhead, a proposed bridge to the JSON-B API for stable standards-based code, serializing Java Records with the @Json.Entity annotation, a package-info annotation idea for processing an entire package, Boundary Control Entity (BCE) structure for entity records, JSON-P style JsonParser and JsonGenerator entry points, HashMap-like dynamic JSON objects for versioned and evolving APIs, a single centralized entry point for discoverability, grounding LLMs with Jakarta EE and MicroProfile standards to avoid hallucinations, reducing dependencies by letting LLMs generate helper methods, Java 25 source-mode scripting and simplified main replacing python scripts, zero-dependency JSON parsing with a forked org.json, JMH benchmarks showing Helidon JSON around 400% faster than Jackson and faster than yasson, energy and cost savings from faster serialization, object pooling and virtual-thread-friendly design with Scoped Values David Kral on twitter: @VerdentDK

An airhacks.fm conversation with Simon Ritter about: Azul Zulu as a straight build of openJDK versus Azul Zing/Prime as the commercial high-performance JVM, improving Java performance without changing application code, C4 the Continuously Concurrent Compacting Collector, fully concurrent and compacting pauseless garbage collection, reaching a global safe point and millisecond-level pauses, a loaded-value read barrier intercepting every object read with a two-instruction test-and-jump, using bits in the object header limiting heaps to 20 terabytes, the one and a half day garbage collection pause anecdote on a terabyte-scale heap, latency-sensitive workloads like high-frequency trading and ad brokering, ZGC as a garbage collector heavily based on the published C4 work, Falcon a JIT compiler based on LLVM replacing HotSpot's C2, contributing the JIT wrapper and garbage collector integration back to LLVM, speculative optimizations and deoptimizations, throughput gains around 53 percent for Kafka and 30 percent for Cassandra, doing more with less to reduce cloud costs, ReadyNow capturing a profile of loaded classes, initialized classes, profiling data and compiled methods to reach about 98 percent of steady-state performance at startup, the Cloud Native Compiler offloading JIT compilation to a centralized cloud service, the Optimizer Hub storing profiles locally or centrally, soft real-time versus hard real-time JVMs, the Real-Time Specification for Java and immortal memory, garbage-collector-less programming, tiered compilation with C1 client and C2 server compilers, virtual threads and platform threads in relation to garbage collector threads, Azul Vega hardware history, Log4Shell as a potential security check in the compiler service Simon Ritter on twitter: @speakjava

An airhacks.fm conversation with Ian Rogers about: Java's early API design strengths versus verbose C++/STL manuals, the Java 7 String.substring behavior change from an O(1) view over the character array to an O(n) copy, substring performance regressions at Google, implementing a custom substring as a workaround, designing APIs without knowing future customers, the SPEC JVM98 jack parser generator throwing an exception per matched token, exceptions used for control flow violating the exceptions-are-exceptional principle, JVM benchmarks skewed toward fast exception throwing, ANTLR and JavaCC as later parser tools, class loaders pairing a type with a runtime notion, ClassNotFoundException thrown from nested jar files, the DaCapo benchmark as a test of exception-throwing speed, Effective Java advice to return interfaces such as Map instead of HashMap, why substring should have returned CharSequence, the argument that String should be an interface and CharSequence the concrete type, CharSequence length limited to int and capped at 2GB, Guava Rope as a collection of CharSequences unable to implement CharSequence because of the int length limit, signed versus unsigned sizing of arrays and strings, byte signed and char unsigned inconsistencies, StringBuilder versus StringBuffer, the race condition avoided by StringBuilder cloning its array in toString, StringBuffer passing array ownership under a lock, biased locking making uncontended locks cheap, the Attack of the Clones problem of defensive cloning, optimizing clones away in the JVM, transactional memory as an alternative to locking, copy-on-write and CopyOnWriteArrayList, Swift copy-on-write, Linux kernel read-copy-update and epochs, writing the Azul C4 garbage collector, the C4 read barrier now used in ZGC and Shenandoah, moving Azul from custom hardware read-barrier instructions to x86, the two-space invariant in concurrent copying collectors, IBM Metronome fixups versus two-space invariants, detecting same-page references by XOR of two pointers, trading computation for memory accesses in read barriers, the Transitive Corporation binary translator, Rosetta for Apple and a Sparc-to-Power translator behind IBM's planned Sun acquisition, writing Android Runtime ART, ahead-of-time compilation of Dex replacing Dalvik, disk-size constraints of AOT compilation, Hans Boehm and sticky mark bits for generational marking without moving objects, ART generational concurrent garbage collection in Android KitKat, HashMap interface dispatch replacing LinkedList iteration from JikesRVM, alphabetically sorted interfaces putting AbstractCollection first, Google Maps interface dispatch consuming half the frame time, frame-rate and pause-time gains from ART, WhatsApp broken by an unbalanced-lock bytecode obfuscator, balanced locks required for biased locking, thread safety annotations in Clang/LLVM guarding the Java heap, the mutator lock and stale pointer risks, a reader-writer lock model of the Java heap, managing risk when replacing an operating system runtime, CyanogenMod as a delivery path for ART, Project Valhalla value types, current work on the Linux perf tool and observability, OProfile origins, GPU versus CPU visibility in system tools Ian Rogers on linkedin: irogers

An airhacks.fm conversation with Steve Hannah about: meeting Steve Hannah at JavaOne years ago via Codename One, the Apple IIGS as a first computer, floppy disk games and Bob Whitehead's Hardball!, starting web development in 1997 with Perl and CGI, the O'Reilly Perl camel book, CGI security issues and the parallel to LLM tools, choosing Java Servlets over CGI for startup-time performance, moving to PHP for quick and dirty web development, PHP evolving toward enterprise patterns with Zend, Symfony and Laravel, PHP at Cloudbeds as a serious enterprise language, Facebook compiling PHP with HipHop, connection pooling and data sources impressing PHP developers, convergence of ideas across Java, JavaScript classes and React moving from object orientation to hooks, web components as JavaScript classes with properties, TypeScript versus JavaScript with JSDoc types in VS Code, LLMs reducing the importance of web frameworks that follow standards, porting React to Java on TeaVM using Claude, Flash and ActionScript for rich interfaces before HTML5, key frames hiding the full code base in Flash, VisualAge for Java showing only methods and printing source to understand it, the Xataface open-source PHP/MySQL CRUD framework started in 2005, the Dataface to Xataface rename, FileMaker as a low-code database tool and its licensing constraints, joining Codename One as the first hire in 2012, front-end Java versus back-end Java culture at JavaOne, the extreme GUI makeover Swing session, coding with LLMs in 2026 using Claude Code and Codex, building a personal Swing-based IDE for multi-repo multi-branch agentic workflows, work trees and terminal-organized agent sessions, a possible Swing and JavaFX revival because they ship with Java and LLMs know them, Java 25 source-mode shebang scripting replacing Python scripts, Java shipping more built-in functionality than Python for regular expressions and dependencies, Ruby versus Python popularity, LLMs understanding Java (~87%) better than JavaScript (~60-70%), type safety enabling LLM self-correction, grounding LLMs with MicroProfile and Jakarta EE normative specifications, the JSR normative language read by LLMs, business annotations and rich JavaDoc and package-info improving LLM output, reducing external dependencies so LLMs avoid decompiling JARs, token savings from standards-based Java, AWS Bedrock for stable and private Claude access with Claude Code, Quarkus on AWS Lambda with cold starts and Provisioned Concurrency, pinging a AWS Lambda every minute within the free tier to avoid cold starts, cost calculation for Quarkus on Lambda with 2 GB RAM, constructions.cloud Quarkus serverless CDK examples, Google Cloud Run versus AWS App Runner, AWS Lambda and Azure Functions, AWS CDK as infrastructure as code in Java, JDeploy for packaging and distributing desktop apps, the AIRails.dev skills site and the microprofile-server skill, zero-dependency Java scripts and agents, migrating a WordPress blog to a Jekyll static site with Claude Steve Hannah on twitter: @shannah78

An airhacks.fm conversation with Donald Raab about: the Epson HX-20 as the world's first laptop and learning Basic, an acoustic coupler modem, the Atari 2600 and Pitfall, Montezuma's Revenge, Apple II clones and the Franklin Ace, running a bulletin board system, learning BASIC, Pascal, fortran, COBOL, Turbo prolog and Turbo Pascal, dBASE III Plus as language and database management system, DBF file format by Ashton-Tate, Clipper by Nantucket as a dBASE compiler with Blinker linker, T-Browse and cursor-based table access, FoxPro and the x-based language family, NDX/NTX/CDX index formats, SQLJ embedded SQL, Paradox and Delphi, learning Smalltalk at IBM Object Technology University, Alan Kay and pure object orientation, blocks as lambdas in Smalltalk and Clipper (via the Classy library), ENVY version control with method-level editions, VisualAge for Smalltalk and VisualAge for Java, method categories for organizing methods, migrating a Clipper application to Java using interfaces and static methods in a procedural style, transparent persistence with TopLink and EclipseLink, memory constraints in 32-bit Java on Solaris, building a custom caching framework, the origins of Eclipse Collections in 2004, waiting ten years for lambdas, the JSR 335 expert group with Brian Goetz, select/collect/inject versus filter/map/reduce, eager methods on collections versus lazy streams, code folding regions in IntelliJ to simulate method categories, the book "Eclipse Collections Categorically", JRuby and Asciidoctor, meeting Yukihiro Matsumoto at OOPSLA 2002 Donald Raab on twitter: @TheDonRaab

An airhacks.fm conversation with Olimpiu Pop about: discussion about the Java Advent calendar 2025, running AI inference in the enterprise with Java versus python, why inference needs a mature ecosystem, the confusion between machine learning, inference and LLMs, HTTP clients for remote models, TornadoVM for local model acceleration, avoiding premature optimization for LLM-generated code, grounding LLMs against normative Java and Jakarta EE specifications, the separation between API and SPI to eliminate hallucinations, Boundary Control Entity (BCE) structure with self-contained packages for scalable LLM code generation, maximal cohesion and minimal coupling, standards as an effective technique for LLM code generation, MDN web components and custom elements for dependency-free front-ends, reducing context and inference cost with short prompts, AIRails skills site, agents and skills across Copilot and other tools, Eclipse Collections, WebAssembly for Java with Chicory and TornadoVM, pattern matching, Java in industrial and IoT systems at ASML, Legionella water-temperature monitoring with wireless sensors, Apache IoTDB and Apache NiFi, migrating Java Advent to Roq static site generator Olimpiu Pop on twitter: @olimpiupop

An airhacks.fm conversation with Johan Haleby about: discussion about the Occurrent event sourcing library for Java, motivation behind building Occurrent as a set of building blocks for event sourced applications, using higher-order functions instead of distinct framework concepts, the library approach for gradually introducing event sourcing into regular projects, CloudEvents as the serialized event format and its evolution into CNCF, real use cases including querying the event store for consistent reads and per-user logs, using the CloudEvents subject attribute to index a user, enforcing unique constraints such as one email per person, dynamic consistency boundary, event sourcing fundamentals of storing a list of events instead of mutable state, basing decisions on prior events, the distinction between event sourcing, CQRS and CQS, materialized views and different read and write models, subscriptions to events after they are persisted, the difference between domain events and integration events for decoupling teams, auditing versus event sourcing and why intent matters, Hibernate Envers, Transaction Script pattern, folding events into state with Java Gatherers fold, modelling domain logic with sealed interfaces and records, static decision methods returning new events, the Redux analogy, snapshotting and closing the books for long streams, Kafka and Kafka Streams for event-driven architecture, cascading enriched events, ksqlDB materializing streams as tables, DynamoDB Streams and on-demand tables, storing state alongside events in one transaction, tractors and fitness trackers as event sources Johan Haleby on twitter: @johanhaleby

An airhacks.fm conversation with Stanislav Bashkyrtsev about: discussion about testing terminology and the difference between unit tests, component tests, System Tests, and integration tests, defining component tests as in-process invocations without HTTP, using RestAssured with MockMvc-style direct endpoint calls, avoiding mocks in favor of real system tests, why code coverage is a misused management metric, the anti-pattern of using reflection to inflate coverage, distinguishing line and branch coverage from actual verification, using coverage from system tests to detect dead code for pruning, mutation testing with PIT to measure assertion quality, testing Quarkus applications, the default Guice and Guava dependencies in Quarkus RESTEasy, starting a new microservice with a separate system-test module, calling endpoints over HTTP with the MicroProfile REST Client or the Java HTTP client, deploying Quarkus on AWS Lambda as a production-like environment, backward compatibility testing with multiple production versions, turning system tests into stress and load tests, testing connection pools and metrics under load, introducing a test-only private API to verify state changes in serverless systems, contract-driven work in large consulting projects, generating JSON and JSONB directly in PostgreSQL and returning it over JDBC, mapping database rows to Java records instead of DTOs, running GraalVM inside the Oracle Database for stored procedures and table triggers, the pendulum between database-centric and application-centric logic, the convergence of SQL and NoSQL databases, CI/CD pipelines with Jenkins and manual production deployment steps, avoiding Jenkins access to production via CGI shell scripts behind nginx, AWS CodePipeline and CodeBuild with CDK-defined infrastructure, event-driven pipelines triggered by S3 put-object events, multi-account roles with short-lived STS credentials, the size of the AWS SDK and reducing it by excluding unused HTTP clients, health checks and Kubernetes liveness and readiness probes, why health checks make little sense for short-lived Lambdas, a version endpoint for deployment smoke tests Stanislav Bashkyrtsev on twitter: @sbashkirtsev

An airhacks.fm conversation with Tomasz Ptak about: discussion about the guest's path from an Atari and a 486 to professional Java development, loading games from cassette tapes, building a clock with the Logo programming language, making websites with PHP for a community, studying data management and computer science, learning Perl, Bash, Pascal, Python, C, C++, Ruby and Java, Java 1.4 and Java 5 with generics and annotations, an island optimization algorithm switching from Python to Java for memory control, preference for strictly typed languages, first job at motorola Solutions building a server-side Java configuration system with SNMP and SNMP4J, moving from Tomcat to Netty, using Ant and Maven, managing a Jenkins server, rebuilding a buggy no-code Spring CRUD generator, rewriting an application with Apache Wicket for stateful web development, comparing Wicket structure coupling with Jakarta Faces, event sourcing with the Axon Framework and domain objects, bitemporal awareness and Hibernate Envers versioning, the Naked Objects pattern and object-oriented UI generation, third job at Open Market stabilizing a legacy Java SMS gateway, weekly outages and same-day retrospectives, containerizing bare-metal systems with Testcontainers and docker Compose, near zero-downtime deployment with Ansible, migrating from Maven to Gradle and removing the Buck build tool, upgrading legacy systems from Java 1.4 to Java 8, minimalistic Maven usage, a zero-dependency Java builder zb and zero-dependency unit runner zunit using only built-in compiler and jar tools, Java 25 as an automation tool replacing Python scripts, executable JARs without external dependencies, shebang instance-method scripting, reactive or infinite streams and stream gatherers, Git-tag-based versioning for monorepos, the AWS DeepRacer and AWS AI community, the mediocris blog Tomasz Ptak on linkedin: https://www.linkedin.com/in/tomasz-ptak

An airhacks.fm conversation with Thorsten Hoeger (@hoegertn) about: discussion about CloudFormation as underlying infrastructure as code using JSON/YAML to define desired state, CDK introduction as tool to define infrastructure using programming languages like Java that synthesizes to CloudFormation, explanation of CDK construct levels L1 (direct CloudFormation mapping), L2 (type-safe with opinionated defaults), L3 (patterns and abstractions), Custom resources for missing CloudFormation implementations or actions like cleaning S3 buckets, CDK resource provider framework simplifying custom resource creation by handling AWS Lambda lifecycle, Self-mutating pipelines concept using CDK to update own pipeline, CDK for Terraform (CDKtf) enabling Terraform synthesis using CDK constructs, jsii facilitating multi-language support for CDK constructs, Mixins feature allowing immediate modification of constructs without full L2/L3 abstraction, Refactoring constructs to move code between construct tree levels, CloudFormation Naming conventions using PascalCase for constructs and kebab-case for CloudFormation resource names Thorsten Hoeger on twitter: @hoegertn