Backend API feature flags require stable targeting for incident reconstruction

·4 views

A DEV Community article outlines a method for implementing backend feature flags to reliably reconstruct checkout failures. It recommends using a deterministic, hash-based bucketing system to assign users to specific feature cohorts. The approach involves attaching a compact decision snapshot to a request trace, containing key details like flag variant and revision for debugging. This design prioritizes incident investigation over minimizing the flag function's size. The article also advises keeping user identifiers out of trace logs to protect privacy and system performance.

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)

OpenFEMA data reveals upcoming infrastructure projects funded for disaster recovery.

FEMA provides public data on obligated funds for post-disaster public repairs and pre-bid hazard mitigation work. This information, accessible via the OpenFEMA API, details project amounts, locations, and statuses for contractors and engineering firms. The data requires processing to combine different program datasets and resolve agency names. A tool has been created to streamline this data into a filterable list for tracking potential future solicitations.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Census building permit data now available in consolidated, user-friendly table

The US Census Bureau's monthly Building Permits Survey provides an early demand signal for the construction industry, tracking permits by structure type across states, counties, and cities. However, the data is published in dozens of separate, fixed-width text files with inconsistent valuation formats, making analysis cumbersome. A new tool automates the process of downloading, cleaning, and standardizing this data from the Census website. It consolidates the information into a single table with consistent dollar values and named columns, allowing for filtering and export. Analysts should note permits represent authorizations, not actual construction starts, and different structure types should be examined separately for accurate insights.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Analysis shows repacked Gemma 4 E2B model outperforms Google's 4-bit export

A technical analysis compares two 4-bit versions of Google's Gemma 4 E2B language model. The test was conducted on a Colab TPU v5e runtime using a pure-JAX engine. The repackaged model, which retains the original trained quantization grid, demonstrated significantly higher accuracy in next-token predictions compared to Google's official export. Google's export applies a second rounding step to the weights, diverging from the model's training. The repackaged version also matched original model outputs more closely while requiring 0.8 GB less storage.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Node.js feature flag implementation requires careful safety design for edtech rollouts

A DEV Community article outlines safe practices for implementing feature flags in Node.js applications, particularly for edtech rollouts. The core recommendation is to avoid network calls in the critical delivery path by using cached local values and background polling. Default flag values should be set to the safest state, typically false, to enable automatic rollback if the flag service fails. For a system requiring 60-second rollback capability, the article suggests a 15-second polling interval and a 45-second maximum cache age. The design prioritizes rollback safety over feature richness when choosing between different flag management solutions.

0 comments Read more at DEV Community