SShortSingh.
Back to feed

OCI Workload Identity Federation enables keyless auth for GitHub Actions and Kubernetes

0
·2 views

Oracle Cloud Infrastructure has introduced Workload Identity Federation (WIF), a mechanism that allows external workloads to authenticate with OCI using temporary credentials instead of permanent API keys or technical users. The feature is designed for CI/CD pipelines, Kubernetes environments, multicloud setups, and AI agents that require short-lived access to OCI resources. Rather than storing long-lived credentials in GitHub Secrets or Kubernetes Secrets, workloads present an identity token from their native platform, which OCI validates and exchanges for an ephemeral session token. The model eliminates the need to maintain standing OCI users per workload, reducing operational overhead around user provisioning, audits, and offboarding. OCI IAM can also leverage identity claims — such as repository, branch, or Kubernetes namespace — to make fine-grained authorization decisions.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

PostgreSQL vs MySQL: How Architecture and Workload Shape the Right Choice

Choosing between PostgreSQL and MySQL for high-traffic applications requires analyzing workload mechanics rather than relying on feature checklists or historical preference. Key factors include read-to-write ratios, transaction complexity, query patterns, data relationships, and concurrency and latency requirements. PostgreSQL uses a multi-process architecture with one server process per client connection, offering strong memory isolation through shared buffers, write-ahead logging, and heap-based MVCC. Its background processes — including autovacuum, checkpointer, and WAL writer — handle durability and storage maintenance under heavy production loads. Engineers are advised to decompose their specific workload profile before selecting a storage engine, as internal mechanics like write amplification and lock contention ultimately determine system stability.

0
ProgrammingDEV Community ·

Viral 'AI Fails' Often Reflect Poor Prompting, Not Flawed Technology

A developer on DEV Community ran an experiment to challenge the popular narrative that AI tools are unreliable and easily fooled. They deliberately instructed Claude, Anthropic's AI assistant, to produce an absurdly wrong answer about the number of R's in 'strawberry' — and the model complied with an intentionally ridiculous response. The author then used this self-generated example to illustrate how misleading it is to share such outputs as proof of AI incompetence. As of mid-2025, free-tier generative AI models have advanced significantly since ChatGPT first gained mainstream attention nearly four years ago. The writer argues that poor prompt quality, rather than inherent AI limitations, is increasingly the reason users receive unhelpful or inaccurate responses.

0
ProgrammingDEV Community ·

Developer builds 8-game browser arcade in plain TypeScript with no frameworks

Developer Minjia Hu has released a retro-styled browser arcade featuring eight classic games — Snake, Tetris, Breakout, Flappy Bird, 2048, Minesweeper, Sudoku, and Gomoku — playable instantly without installation or sign-up. The entire build weighs roughly 35 kB gzipped and uses only plain TypeScript and Canvas 2D with zero runtime dependencies. All eight games share a unified arcade-cabinet interface supporting both keyboard and touch input, while each game's logic is kept strictly separate from rendering code. The project includes 269 unit tests and 28 end-to-end Playwright tests running in continuous integration, and is released under the MIT license. The source code and live demo are publicly available on GitHub.

0
ProgrammingDEV Community ·

Developer Fixes WebRTC-Direct Race Condition and SCM Polling Bugs Across py-libp2p and Go Agent Orchestrator

Open-source contributor Yash spent the week working across py-libp2p and a Go-based agent orchestrator, shipping 8 commits and managing 11 pull requests. A key fix addressed a race condition in the WebRTC dialer where ICE credentials were not set correctly before first-contact replay, which had been blocking interoperability between Go and Python nodes. He also resolved socket leaks in the mplex and swarm components to prevent file descriptor exhaustion in long-running nodes. On the orchestrator side, a bug causing backslash-named workspace files to resolve incorrectly on Linux and macOS was patched, and a GitHub polling issue where open PRs were repeatedly re-listed due to a mishandled 'since' parameter was investigated. Additional work included documentation improvements for WebRTC-Direct dialing and an in-progress pull request introducing a more robust ICE-Lite listener.