Observations from the field — on software, technology, people, and the everyday challenges of making IT projects work.
The performance characteristics of HotSpot (Temurin) and OpenJ9 (Semeru) are evaluated. A primary focus is placed on the startup superiority of HotSpot when utilizing Leyden AOT compared to the memory efficiency and Shared Class Cache (SCC) tuning of the OpenJ9 runtime.
Read more →A data-driven comparison between Quarkus (Java) and .NET (C#), moving beyond industry "feelings" to analyze empirical performance. Key findings reveal that while Quarkus remains the leader in raw throughput, .NET excels in memory efficiency for standard dynamic runtimes. The analysis also explores the impact of Project Leyden on Java startup times and identifies Quarkus Native as the top all-around performer when dynamic features are not required.
Read more →Roq's markdown plugin already emits the right markup for Prism. With the Quarkus Web Bundler, wiring up self-hosted, cache-busted syntax highlighting takes one Maven dependency and one JavaScript entry point.
Read more →Roq does not ship i18n primitives, but its conventions are flexible enough that adding a second language is mostly content layout and a single layout-side check. Here is how we did it.
Read more →Roq has no built-in image optimization, and a related Web Bundler issue is still open. Here is how we wired up automated WebP conversion using the tools Roq already exposes — a CDI producer, a Vert.x route, and a Qute user tag.
Read more →A behind-the-scenes look at how the ITBH website and blog came to life through a conversation with an AI coding assistant — and what that experience revealed about the future of software craftsmanship.
Read more →AI is a perfect instruction-follower. If your input is vague, contradictory or non-committal, the output will be exactly that — garbage in, garbage out, dressed up in fancy code.
Read more →Java is not the slow, outdated enterprise relic some still imagine. Here is why we chose Java and Quarkus over the Microsoft stack — and why that decision has aged well.
Read more →We run Nextcloud with a self-hosted LLM via LocalAI and vLLM. Response times were unpredictable — here is what we found and how we fixed it.
Read more →