FreeBSD 12.2 vs Linux on Ryzen 9 5900X: OpenZFS Benchmarks

Over the past nearly two months we have been running a lot of Linux benchmarks on the AMD Ryzen 5000 series, but what about the BSD operating systems with these Zen 3 desktop CPUs? Recently we got around to trying out several BSDs on a Ryzen 9 5900X desktop, as well as running FreeBSD 12.2 vs. Ubuntu Linux benchmarks, including with Linux on OpenZFS and Clang.

This benchmark suite is particularly valuable because it isolates multiple variables simultaneously: filesystem (ext4 vs OpenZFS), compiler toolchain (GCC vs Clang), and kernel (Linux vs FreeBSD) on identical Zen 3 hardware. The results reveal that OpenZFS on Linux narrows the historically wide FreeBSD-vs-Linux performance gap, suggesting that filesystem choice can matter more than kernel choice for storage-intensive workloads on modern AMD processors.

WSL 2 Gains NVIDIA CUDA GPU Passthrough for Linux

At Microsoft Build in the first half of the year, Microsoft demonstrated impressive new capabilities and improvements coming to Windows Subsystem for Linux 2, including the ability to share the host machine's GPU with WSL 2 processes. Then in June, Craig Loewen from Microsoft announced that developers on Windows Insider ring machines could now make use of GPU acceleration for Linux workloads. This NVIDIA CUDA support enabled developers and data scientists to use their local Windows machines for inner-loop development and experimentation.

GPU passthrough in WSL 2 addresses a genuine pain point for machine learning practitioners who were previously forced to dual-boot or use remote GPU instances. However, the implementation routes through a paravirtualized GPU driver rather than bare-metal access, which introduces latency overhead that makes it unsuitable for production training workloads. The real significance is for local prototyping and debugging -- it eliminates the context switch between writing code on Windows and testing CUDA kernels, which previously required a full OS reboot into a native Linux installation.

OpenJDK Project Lanai: Metal API Replaces OpenGL on macOS

Most graphical Java applications are written using the Swing UI toolkit, which renders via the Java 2D API. Internally, Java 2D can use software rendering plus a blit to the screen, or it can use a platform-specific API, such as X11/Xrender on Linux, Direct3D on Windows, or OpenGL on macOS. These platform-specific APIs typically offer much better performance than software rendering and generally offload the CPU. Metal is the new Apple platform API for such rendering, replacing the deprecated OpenGL API. (The name has nothing to do with the Swing "Metal" Look and Feel; that is just a coincidence.) We will update existing internal JDK code and create new code to utilize the Metal framework in a similar way as we already utilize the other platform-specific APIs. We will also investigate opportunities to leverage new capabilities of the Metal framework, such as more concurrent rendering.

This migration is not optional -- Apple deprecated OpenGL in macOS Mojave (2018) and has not updated it since. JEP 382 ensures Java desktop applications remain viable on macOS long-term. The early benchmarks comparing OpenGL vs Metal Java 2D rendering show Metal matching or exceeding OpenGL in most operations, with particularly strong gains in concurrent rendering paths. For enterprise Swing applications still deployed on macOS workstations, this is a necessary modernization that also unlocks future GPU compute possibilities through Metal's shader pipeline.

Wasmer Runtime Brings WebAssembly Beyond the Browser

Wasmer allows you to run WebAssembly modules either standalone or embedded within other languages such as C/C++, Rust, Python, Go, PHP, Ruby, and more. By design, the environment within which a WebAssembly module runs is completely isolated (or sandboxed) from the native functionality of the underlying host system. This means that by default, Wasm modules are designed to perform nothing more than pure computation. Consequently, access to OS-level resources such as file descriptors, network sockets, the system clock, and random numbers is not normally possible from WASM. However, there are many cases in which a Wasm module needs to do more than perform pure computation; it must interact with native OS functionality.

Wasmer represents a critical step in the WASI (WebAssembly System Interface) ecosystem that extends WASM's value proposition from browser sandboxing to server-side and edge computing. The key differentiator is the capability-based security model: unlike containers, which rely on kernel namespaces and cgroups, WASM modules get zero permissions by default and must be explicitly granted access to each resource. This makes WASM a natural fit for plugin systems, edge functions, and multi-tenant environments where the attack surface of container runtimes is a genuine liability.

WSL 2 Adds Native ext4 Filesystem Mounting via wsl --mount

Starting with Windows Insiders preview build 20211, WSL 2 will be offering a new feature: wsl --mount. This new parameter allows a physical disk to be attached and mounted inside WSL 2, which enables you to access filesystems that aren't natively supported by Windows (such as ext4). So, if you're dual booting with Windows and Linux using different disks, you can now access your Linux files from Windows.

The wsl --mount feature eliminates a long-standing pain point for dual-boot users who needed third-party tools like Ext2Fsd or DiskInternals to read Linux partitions from Windows. By routing disk access through WSL 2's real Linux kernel rather than a user-space driver, file permission semantics (ownership, ACLs, symlinks) are preserved correctly. This is also a strategic move by Microsoft: as desktop applications increasingly run in browsers and Electron wrappers, the underlying OS matters less -- making it essential for Windows to interoperate seamlessly with Linux rather than compete against it directly.

Nvidia Nears $40B Deal to Acquire ARM from SoftBank

According to those familiar with the matter, a cash-and-stock deal between the two companies could occur early next week and would be valued at over $40 billion, which could potentially be the largest deal ever in the semiconductor industry. SoftBank initially acquired ARM for $32 billion four years ago. The two companies have reportedly been in talks for weeks over a possible deal, and if completed, scrutiny may arise among antitrust regulators since Nvidia itself is currently a customer of ARM. Companies that utilize ARM technology would also not be in favor of a deal without explicit assurances that ARM's instruction set will continue to be available for equal licensing opportunities.

The antitrust concerns around this acquisition are well-founded. ARM's business model depends on being a neutral IP licensor -- Qualcomm, Apple, Samsung, and Nvidia itself all rely on ARM's architecture. An Nvidia-owned ARM creates an inherent conflict of interest that could fragment the ecosystem. Meanwhile, RISC-V is gaining traction as a royalty-free alternative (SiFive's HiFive Unmatched board shipped in this period), and x86 efficiency has improved dramatically in recent process nodes, narrowing the historical power-per-watt advantage that made ARM dominant in mobile. The deal's outcome will determine whether ARM remains an open platform or becomes a competitive weapon.

ARM, Nvidia, Intel

JDK 16 Vector API Brings SIMD to Java Applications

Vector computations consist of a sequence of operations on vectors. A vector comprises a (usually) fixed sequence of scalar values, where the scalar values correspond to the number of hardware-defined vector lanes. A binary operation applied to two vectors with the same number of lanes would, for each lane, apply the equivalent scalar operation on the corresponding two scalar values from each vector. This is commonly referred to as Single Instruction Multiple Data (SIMD). Vector operations express a degree of parallelism that enables more work to be performed in a single CPU cycle and thus can result in significant performance gains. For example, given two vectors each covering a sequence of eight integers (eight lanes), then the two vectors can be added together using a single hardware instruction. The vector addition hardware instruction operates on sixteen integers, performing eight integer additions, in the time it would ordinarily take to operate on two integers, performing one integer addition.

The Vector API (JEP 338) is the JVM's answer to both WASM SIMD and native vectorization in languages like Rust and C++. Before this API, Java developers had to rely on the HotSpot auto-vectorizer, which only optimized simple loop patterns and could not express explicit SIMD intent. The new API maps directly to hardware vector instructions (SSE, AVX-256, AVX-512 on x86; NEON on ARM), giving Java applications near-native throughput for numerical workloads like matrix operations, signal processing, and machine learning inference -- domains where the JVM was previously uncompetitive.

Java, JVM, JDK16

Intel Tiger Lake 11th Gen Benchmarks vs AMD Ryzen 4000

Intel Tiger Lake will soon begin appearing in laptops with an upgraded CPU architecture, the all-new Iris Xe (Gen12) graphics, new AI capabilities, Thunderbolt 4, PCI Express 4.0, WiFi 6, and other new functionality. The Gen12 graphics have me most excited but there should be healthy improvements as well on the CPU side, not to mention improved connectivity. Intel Tiger Lake is better positioned to compete with AMD's current generation Ryzen 4000 (Renoir) series mobile processors, but even still the top-end Tiger Lake (U) model is only four cores / eight threads. TDPs on the new parts range from 7 to 28 Watts.

Tiger Lake's real competitive advantage lies in Iris Xe integrated graphics, which are genuinely competitive with entry-level discrete GPUs for the first time. On the CPU side, the 4-core ceiling at 28W TDP means Intel concedes the multi-threaded performance crown to AMD's 8-core Ryzen 4800U. However, both Intel and AMD are US-based companies subject to the same export restrictions. For regions evaluating sovereign chip strategies, neither vendor provides supply chain independence -- a factor that makes RISC-V and domestic ARM licensees increasingly relevant for geopolitical risk mitigation.

Intel, AMD, ARM

Lenovo Sells ThinkPad X1 Carbon with Fedora Preinstalled

As a follow-up from the news earlier this summer of Lenovo planning to certify their ThinkPad and ThinkStation lines for Linux from Ubuntu and Red Hat while also offering distribution choices like Fedora, that work is proceeding with Lenovo now offering their first system from the web store that comes preloaded with Fedora. The ThinkPad X1 Carbon Gen 8 is available with Fedora preloaded while still offering options from Core i5 through Core i7 10th Gen CPUs, 8GB / 16GB of RAM, a variety of display options (including 14-inch 4K), and more.

Lenovo offering Fedora on their flagship ThinkPad X1 Carbon marks a turning point for Linux desktop adoption. Previous preinstalled Linux options were limited to niche vendors like System76 and TUXEDO. Having a tier-1 OEM sell Linux through their mainstream web store eliminates the two biggest barriers for enterprise Linux desktop adoption: hardware certification risk and the perception that Linux laptops are not commercially supported. The choice of Fedora over Ubuntu also signals Red Hat's push to position Fedora as the enterprise developer workstation distribution.

Ubuntu, Schenker, Tuxedo

OxygenOS 11 Abandons Stock Android for Samsung One UI Design

The first OxygenOS 11 preview is now live, and we have rounded up all the latest features. The latest version of the interface comes with a host of new features, including an always-on display, updated Zen Mode, new system font, optimized system-wide dark mode, and more. But one change that OnePlus users may not welcome is the new design. OxygenOS always stood out for its clean user interface inspired by stock Android, with the skin likened to Google's Pixel Launcher on steroids. That is changing with OxygenOS 11, as OnePlus introduces a new design language that is more in line with Samsung's One UI than Pixel and Android One phones.

This design pivot is a strategic bet that reveals OnePlus's shifting market position. The stock Android aesthetic originally attracted power users and developer-oriented buyers, but that segment has limited scale. By adopting One UI-style one-handed navigation patterns (larger touch targets in the lower half of the screen, header content pushed upward), OnePlus is chasing Samsung's broader consumer market. The risk is alienating their existing community -- and the eventual merger with OPPO's ColorOS in 2022 proved this was less about UX philosophy and more about consolidating engineering resources across BBK Electronics subsidiaries.

Play Store, Huawei, App Gallery, Apple

Apple vs Epic Games: Fortnite App Store Battle Escalates

The dispute between Apple and the maker of the popular video game Fortnite over in-app purchasing rules continued Sunday with Fortnite hosting a free online competition billed as the last time gamers will be able to play together across all platforms. The #FreeFortnite Cup was scheduled two days after Apple asked a federal judge in California on Friday to deny Fortnite maker Epic Games' request to have the game returned to the App Store.

Epic's legal strategy is deliberately designed to set antitrust precedent, not just win a commercial dispute. By implementing a direct payment system that explicitly bypasses the 30% App Store commission, Epic forced Apple into a public confrontation that exposes the tension between platform control and competitive market access. The outcome of this litigation could reshape how all mobile software distribution works -- particularly whether iOS must allow alternative app stores or sideloading, which would fundamentally alter Apple's services revenue model that generates over $20 billion annually.

Play Store, Huawei, App Gallery, Apple

Apple Removes Fortnite from iOS App Store Over Payments

Apple has removed Epic Games' battle royale game Fortnite from the App Store after the developer on Thursday implemented its own in-app payment system that bypassed Apple's standard 30 percent fee. The decision marks a significant escalation in the feud between Epic and one of the world's most dominant mobile software marketplaces. It also comes at an especially fraught time for Apple as the iPhone maker navigates antitrust concerns over its operation of the App Store and the rules it imposes on certain developers.

The App Store model's fundamental weakness is that it conflates distribution, payment processing, and quality control into a single gatekept channel. The web, by contrast, operates as an open bazaar: any developer can ship software to any user without intermediary approval, payment is handled by independent processors, and quality is enforced by users and search engines rather than a curator. As progressive web apps (PWAs) gain capabilities -- offline support, push notifications, hardware access via WebHID and WebUSB -- the technical justification for mandatory native app stores weakens, making this legal battle a question of when, not whether, walled gardens face structural competition.

Play Store, Huawei, App Gallery

Huawei Kunpeng 920: 64-Core ARM Server CPU at 7nm Process

Overview Huawei Kunpeng 920 is the industry's highest-performance ARM-based server CPU to date. Using the cutting-edge 7nm process, the CPU was independently designed by Huawei based on ARMv8 architecture license. It significantly improves processor performance by optimizing branch prediction algorithms, increasing the number of execution units, and improving the memory subsystem architecture and so on. At typical frequency, the Huawei Kunpeng 920 CPU scores more than an estimated 930 on SPECint_rate_base2006. At the same time, power efficiency is 30% better than that offered by industry counterparts. Huawei Kunpeng 920 provides much higher computing performance for data centers while slashing power consumption. TaiShan will enable computing platforms with high performance and low power consumption for enterprises. Main Specification Architecture: ARM v8.2 Core: up to 64 Typical Frequency: 2.6GHz Memory: 8 DDR4 Channels Coherent Interconnect: 2S&4S I/O: PCIe 4.0, CCIX, 100G, SAS/SATA 3.0 Max Power: 180W Process: 7nm Highlights Huawei Kunpeng 920 will enable computing platforms with high performance and low power consumption for scenarios including Big Data, Distributed Storage, Database, ARM-Native, Edge Computing and so on. It brings unique value in the coming diversified computing era. Huawei Kunpeng 920 SPECint_rate_base2006 estimate based on internal testing. The SPEC CPU 2006 benchmark has been retired and SPEC is no longer reviewing or publishing results with this benchmark. SPEC, SPEC CPU, and SPECint are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information.

The Kunpeng 920's 64-core ARMv8.2 design with 8 DDR4 channels and PCIe 4.0 positions it competitively against AWS Graviton 2 and Ampere Altra in the ARM server market. However, the US Entity List restrictions on Huawei's access to TSMC's advanced process nodes create existential uncertainty for Kunpeng's roadmap. Without access to 5nm or smaller nodes, Huawei cannot match the generational performance improvements that competitors will achieve. This situation has accelerated China's domestic semiconductor investment, but fabrication capabilities at leading-edge nodes remain years behind TSMC and Samsung.

ARM, Intel, Linux, Ryzen

Apple Silicon MacBook Targets $799 Entry Price Point

This MacBook could be the cheapest Apple laptop ever, according to a new report. According to newly leaked data, the new MacBook with Apple Silicon will be the most inexpensive Apple laptop to date. The MacBook will cost approximately $799 and the MacBook Pro 13-inch approximately $1,099. The leaked price points are $200 less than the 2020 versions of both the MacBook Air and the MacBook Pro 13-inch. Apple is expected to officially announce these new laptops as well as their official pricing in the fall. (Source: LaptopMag) Why this is important for your business: We're all trying to conserve cash in these very difficult times, so if you're thinking of replacing MacBooks in your office you may want to hold off for a few months.

The economics behind Apple Silicon pricing are straightforward: Apple's A-series chips cost an estimated $30-50 to manufacture on TSMC's 5nm process, compared to the $150-300 Intel charges for comparable mobile CPUs. This margin improvement lets Apple simultaneously cut retail prices and increase per-unit profit. Beyond pricing, vertical integration gives Apple control over the entire performance-per-watt curve -- something Dell, HP, and Lenovo cannot replicate because they depend on Intel and AMD for silicon. When total cost of ownership includes hardware reliability, OS update longevity (5+ years of macOS support), and resale value, the $799 entry point makes Apple Silicon MacBooks competitive even against mid-range Windows laptops.

ARM, Intel, macOS, BigSur

Apple Silicon ARM Transition Forces Microsoft and Google Response

Not only is Apple disrupting the MacBook and Mac market with its move to ARM-based processors, it's also going to have an impact on the wider marketplace. Microsoft will be forced to accelerate its Windows 10 on ARM project to allow its partners to stay in touch with Apple. And Google will have to consider the danger to its Chromebook project. I've previously talked about the impact of Apple's move to ARM-based processors on the Windows 10 ecosystem. In summary it's going to force Microsoft's Windows 10 on ARM into the spotlight; rival laptop manufacturers will no doubt offer at least one ARM-powered line-up in the portfolio rather than let Apple take all the glory; and there will be an increased focus on fixing legacy app compatibility (something that Windows 10 and macOS will have to deal with quickly and smoothly).

Linux stands to gain disproportionately from the ARM laptop transition compared to Windows. The Linux kernel, GNU toolchain, and virtually all major open-source software already compile natively for ARM64 -- a multi-year head start built through Raspberry Pi, Android, and ARM server deployments. Windows, by contrast, still struggles with x86 emulation performance and has a massive legacy Win32 application catalog that cannot be easily recompiled. If ARM laptops become mainstream, Linux distributions could offer near-parity with macOS on ARM hardware while Windows fights compatibility battles, creating a window of opportunity for Linux desktop market share growth.

ARM, Intel, MIPS

Nvidia in Advanced Talks to Buy ARM from SoftBank

Nvidia Corp. is in advanced talks to acquire Arm Ltd., the chip designer that SoftBank Group Corp. bought for $32 billion four years ago, according to people familiar with the matter. The two parties aim to reach a deal in the next few weeks, the people said, asking not to be identified because the information is private. Nvidia is the only suitor in concrete discussions with SoftBank, according to the people. A deal for Arm could be the largest ever in the semiconductor industry, which has been consolidating in recent years as companies seek to diversify and add scale. But any deal with Nvidia, which is a customer of Arm, would likely trigger regulatory scrutiny as well as a wave of opposition from other users.

The potential $40B+ acquisition would give Nvidia control over the instruction set architecture used by virtually every smartphone, most embedded systems, and an increasing share of cloud server CPUs. The regulatory challenge is unprecedented: ARM's neutral licensing model underpins the entire mobile ecosystem, and major licensees including Qualcomm, Apple, Samsung, and Amazon (Graviton) would face a competitor controlling their fundamental CPU architecture. Multiple governments -- UK, EU, China, and US -- would need to approve the deal, making the regulatory path as complex as the technology implications.

ARM, Intel, MIPS

Google Cloud Launches Confidential VMs with In-Memory Encryption

Google Cloud encrypts data at rest and in transit, but customer data must be decrypted for processing. Confidential Computing is a breakthrough technology that encrypts data in use -- while it is being processed. Confidential Computing environments keep data encrypted in memory and elsewhere outside the central processing unit (CPU). Confidential VMs, now in beta, are the first product in Google Cloud's Confidential Computing portfolio. We already employ a variety of isolation and sandboxing techniques as part of our cloud infrastructure to help make our multi-tenant architecture secure. Confidential VMs take this to the next level by offering memory encryption so that you can further isolate your workloads in the cloud. Confidential VMs can help all our customers protect sensitive data, but we think it will be especially interesting to those in regulated industries.

Confidential Computing represents the most significant cloud security advancement since network encryption became standard. By leveraging AMD SEV (Secure Encrypted Virtualization) at the hardware level, Confidential VMs ensure that even the cloud provider's own hypervisor and host OS cannot read guest memory contents. This eliminates the "trust the provider" assumption that has been the primary legal and compliance barrier preventing financial services, healthcare, and government agencies from migrating sensitive workloads to public cloud. The technology enables entirely new use cases: multi-party computation where competitors can jointly analyze data without exposing it to each other, and sovereign cloud deployments where data residency requirements extend to runtime memory, not just storage.

VMware, AWS, Azure, Disruption

JDK 16 Upgrades HotSpot JVM to C++14 Language Features

Through JDK 15, the language features used by C++ code in the JDK have been limited to the C++98/03 language standards. With JDK 11, the code was updated to support building with newer versions of the C++ standard, although it does not yet use any new features. This includes being able to build with recent versions of various compilers that support C++11/14 language features. The purpose of this JEP is to formally allow C++ source code changes within the JDK to take advantage of C++14 language features, and to give specific guidance about which of those features may be used in HotSpot code.

JEP 347 is an infrastructure modernization that directly impacts JVM performance. C++14 features like move semantics, constexpr functions, and lambda expressions allow HotSpot developers to write more efficient JIT compiler and garbage collector code. The practical benefit is that JVM internals -- particularly the C2 compiler's optimization passes and ZGC's concurrent collection paths -- can use modern C++ idioms that reduce memory allocations and enable better compiler optimizations in the JVM build itself. This compounds over time: every JDK release built with C++14 can be incrementally faster in areas that were previously constrained by C++98 limitations.

Java16, Kotlin, Rust

Corning Gorilla Glass Victus: 2m Drop and 2x Scratch Resistance

Gorilla Glass Victus is the latest in glass innovation from the people at Corning, who have been working on delivering better glass for smartphones since the original iPhone. While Gorilla Glass is up to number 6 in its line, Corning has announced new glass and new branding, calling it Victus. Sure, fine. But what is interesting here is that Corning says the new Gorilla Glass Victus now brings two-meter drop protection (over six feet), onto hard and rough surfaces, while doubling the scratch resistance over Gorilla Glass 6. That is quite something, because Corning has only ever announced better scratch resistance or drop protection -- the twin pillars of mobile glass -- but never both together. Glass strength is a battle. Just like diamonds are very hard but brittle, and sapphire glass as well, Corning tries to increase hardness against scratches, while making glass less brittle. So, this is an event. And the hints are the Galaxy Note 20 will be the first to have it equipped when Samsung announces that phone in just under two weeks. You can watch Corning test out the new glass here. There is some joy in this, although most people will be looking for a certain YouTuber's regular durability testing which usually tracks "scratches at level 6 with deeper grooves at level 7." Will that change with Victus?

Gorilla Glass Victus is the first generation to simultaneously improve both hardness (scratch resistance) and toughness (drop survival) -- properties that are fundamentally at odds in materials science because harder glass is typically more brittle. Corning achieved this through a deeper ion-exchange compression layer that extends further into the glass substrate. Apple's earlier exploration of sapphire crystal (Mohs hardness 9 vs Gorilla Glass's ~7) was abandoned because sapphire shatters catastrophically on impact despite being nearly unscratchable. Victus's engineering tradeoff -- slightly softer than sapphire but dramatically more impact-resistant -- is the pragmatic choice for devices that are dropped far more often than they are scratched by hard minerals.

iPhone, Android

Foxconn Begins iPhone 11 Assembly in India Near Chennai

Apple's contract manufacturing partner Foxconn has started to assemble the current generation of iPhone units -- the iPhone 11 lineup -- in its plant near Chennai, India, a source familiar with the matter told TechCrunch. A small batch of locally manufactured iPhone 11 units has already shipped to retail stores, but the production yield is currently limited, the person said, requesting anonymity as matters are private. Apple, in general, has ambitions to scale up its local production efforts in India, the person said. The local production of current iPhone 11 models illustrates Apple's further commitment to India, the world's second largest smartphone market, as it explores ways to cut its reliance on China, which produces the vast majority of iPhone models today.

Apple's India manufacturing expansion is driven by three converging pressures: US-China trade tensions that impose tariff risk on China-assembled imports, India's Production Linked Incentive (PLI) scheme offering 4-6% financial incentives for local manufacturing, and the need to reduce single-country supply chain concentration exposed by COVID-19 factory shutdowns. However, India's component ecosystem remains underdeveloped compared to China's Shenzhen corridor -- most sub-assemblies (displays, chips, batteries) are still imported and assembled locally. True supply chain diversification requires building the entire vendor ecosystem, not just final assembly, which is a multi-decade undertaking.

Politics, Apple, iPhone

Chrome 85 Ships Fetch Upload Streaming and WebHID API

Fetch Upload Streaming Fetch upload streaming lets web developers make a fetch with a ReadableStream body. Starting in Chrome 85, it is available for an origin trial. Previously, you could only start a request once you had the whole body ready to go. But now, you can start sending data while you are still generating the content, improving performance and memory usage. For example, an online form could initiate a fetch as soon as a user focuses a text input field. By the time the user clicks enter, fetch() headers would already have been sent. This feature also allows you to send content as it is generated on the client, such as audio and video. For more information, see Streaming requests with the fetch API. See the Origin Trials section for information on signing up and for a list of other origin trials starting in this release. This origin trial is expected to run through Chrome 87 in January 2021. WebHID API There is a long tail of human interface devices (HIDs) that are too new, too old, or too uncommon to be accessible by systems' device drivers. The WebHID API solves this by providing a way to implement device-specific logic in JavaScript. In Chrome 85, it is available for an origin trial.

Fetch Upload Streaming and WebHID are part of Chromium's ongoing campaign to close the capability gap between native applications and web applications. Streaming uploads enable real-time audio/video transmission without WebRTC's complexity, while WebHID gives web apps direct access to game controllers, medical devices, and industrial peripherals that lack standard HID class drivers. Each new API reduces the justification for platform-specific native apps -- and by extension, the gatekeeping power of proprietary app stores. The strategic implication is clear: as the browser becomes a universal application runtime, distribution control shifts from Apple and Google's stores to the open web.

Firefox, Chrome, Safari

TUXEDO Pulse 15: Ryzen 7 4800H Linux Ultrabook with 20h Battery

Augsburg (Germany), 20th of July 2020. Our brand new TUXEDO Pulse 15 redefines the class of super thin and lightweight laptops and raises the bar to a completely new performance level: Instead of using a common power-saving ULV (Ultra Low Voltage) processor, the 16.8 mm thin chassis made of magnesium alloy houses a high performance AMD Ryzen 7 4800H octa core CPU with a thermal design power of up to 54 W. In order to give users besides this great performance also excellent battery runtimes while unplugged from the power socket, a huge 91 Wh Lithium-Polymer battery provides the AMD-only ultrabook with energy for more than 20 hours.

The TUXEDO Pulse 15 solves a persistent problem in the Linux laptop market: getting high-performance hardware without paying for an unused Windows license (the so-called "Windows tax" typically adds $100-150 to OEM costs). By pairing the Ryzen 7 4800H (8 cores, 16 threads at 54W) with a 91 Wh battery in a 16.8mm chassis, TUXEDO delivers a machine that competes directly with Dell XPS 15 and ThinkPad T-series on specs while shipping with a fully tested Linux installation. The use of AMD's Renoir APU is strategic -- AMD's open-source Linux GPU driver (amdgpu) is upstream in the kernel, avoiding the proprietary driver headaches that still plague Nvidia-equipped Linux laptops.

Linux, macOS, Apple

Coinbase Skips Kubernetes for Custom Container Orchestration

Coinbase built and maintains their own platform that is working for them. Coinbase provided an analysis worth studying. The major takeaway for me: asking people to manage their own Kubernetes cluster is like asking people to manage their own hypervisors when they just want VMs.

Coinbase's decision to build a custom container orchestration platform instead of adopting Kubernetes highlights a recurring pattern in infrastructure engineering: the complexity cost of a general-purpose tool can exceed the complexity of the problem it solves. Kubernetes provides a superset of capabilities -- service mesh, custom resource definitions, operator patterns, multi-cluster federation -- that most organizations never use, yet still pay for in operational overhead, YAML configuration sprawl, and the steep learning curve for every engineer who touches the stack. For organizations with straightforward deployment patterns (stateless HTTP services, blue-green deploys), simpler alternatives like ECS, Nomad, or custom tooling often deliver higher reliability with lower operational burden.

K8s, Cloud, AWS

macOS Big Sur: Biggest Desktop Design Overhaul in a Decade

macOS Big Sur is a major update that advances the legendary combination of the power of UNIX with the ease of use of the Mac, and delivers our biggest update to design in more than a decade," said Craig Federighi, Apple's senior vice president of Software Engineering. "With its modern and clean look, huge improvements to key apps including Safari, Messages, and Maps, and new privacy features, we think everyone is going to love the breakthrough experience that macOS Big Sur offers.

macOS Big Sur (version 11) is architecturally significant beyond its visual redesign. It is the first macOS version to run natively on both Intel x86-64 and Apple Silicon ARM64, requiring Apple to ship Universal 2 binaries and a Rosetta 2 translation layer. The generous backward compatibility -- supporting Macs from 2013 and 2014 depending on model -- contrasts sharply with Microsoft's tight hardware requirements for Windows 11 a year later. Apple's long update support cycle (7+ years for some models) reduces e-waste and total cost of ownership, which is a competitive advantage that enterprise IT departments increasingly factor into procurement decisions.

macOS, Apple, Windows, Linux, Fedora

ECMAScript 2020: Optional Chaining, BigInt, and Dynamic Imports

Specific features introduced by ECMAScript 2020 include: A "function-like" import() module loading syntax to asynchronously import modules with a dynamic specifier. The proposal adds an import(specifier) syntactic form, acting in many ways like a function. It returns a promise for the module namespace object of the requested module, created after fetching, instantiating, and evaluating a module's dependencies, along with the module itself. The specifier will be interpreted the same way as in an import declaration. While specifier is a string, it is not necessarily a string literal; thus, code like import(`./language-packs/${navigator.language}.js`) will work. This was not possible with the usual import declarations. With the plan, import() is proposed to work in both modules and scripts, giving script code an easy asynchronous entry point into the module world and allowing it to start running module code. BigInt, a new number primitive for working with arbitrary precision integers. BigInt can represent numbers larger than two to the 53rd power, the largest number JavaScript can represent reliably with the Number primitive. A BigInt is created by appending n to the end of the integer or by calling the constructor. The matchAll method for strings, to produce an iterator for all matched objects generated by a global regular expression. The rationale behind this proposal is that if a developer has a string and either a sticky or a global regular expression with multiple capturing groups, the developer might want to iterate through all the matches, for which there are currently several options but with limitations. String#matchAll addresses issues by providing access to all capturing groups and not visibly mutating the regular expression object in question. Promise.allSettled, a new Promise combinator that does not short-circuit. This returns a promise that is fulfilled with an array of promise state snapshots, but only after the original promises have settled, i.e. have been either fulfilled or rejected. globalThis, providing a universal way to access the global this value. A dedicated export * as ns from 'module' syntax to use within modules. Increased standardization of for-in enumeration order, partially specifying enumeration order in JavaScript. import.meta, a host-populated object in Modules that can contain contextual information. This serves as a JavaScript metaproperty, holding host-specific metadata about the current module. Nullish coalescing, a value selection operator for better handling of cases involving property accesses. It is a syntax feature to improve working with "nullish" values (null or undefined). Optional chaining, a property access and function invocation operator that will short-circuit if the value to access/invoke is nullish.

ECMAScript 2020's most impactful additions are optional chaining (?.) and nullish coalescing (??), which together eliminate the most common source of runtime TypeErrors in JavaScript applications. These are not just syntactic sugar -- they change how defensive coding patterns are expressed across the entire language ecosystem. Dynamic import() enables code splitting at the language level rather than relying on bundler-specific mechanisms like webpack's magic comments, which benefits languages that transpile to JavaScript (TypeScript, Kotlin/JS, Dart) because their compilers can now target a standard dynamic module loading primitive instead of emitting bundler-specific output.

Kotlin, VM, JavaScript, Chromium

WSL 2 Performance: 21% Faster Than WSL 1, Within 8% of Native

When taking the geometric mean of those 69 benchmarks, using WSL 2 on Windows 10 May 2020 Update meant around 21% better performance over the original WSL implementation. Ubuntu 20.04 running bare metal on the same system was faster by just 8%. With WSL 2 offering much faster I/O as the main bottleneck with the original WSL, the WSL 2 performance is now incredibly close to that of Ubuntu itself running on the system. All 69 benchmarks can be viewed via OpenBenchmarking.org.

The 21% WSL 2 improvement over WSL 1 comes almost entirely from replacing the syscall translation layer with a real Linux kernel running in a lightweight Hyper-V VM, which eliminates the per-syscall overhead that crippled I/O-heavy workloads. However, the benchmarks obscure a critical real-world limitation: cross-OS filesystem access remains painfully slow. Accessing /mnt/c (the Windows filesystem from WSL) or \\wsl$ (the Linux filesystem from Windows) traverses the 9P protocol, which adds substantial latency to operations like Gradle builds that perform thousands of small file reads and writes. For build systems that need to delete open files -- standard Unix behavior that NTFS does not support -- this mismatch produces deadlocks and spurious failures that synthetic benchmarks never capture.

Ubuntu, Debian, Fedora, WSL, RedHat