Platforms & Engineering

What counts as a Platforms piece

This desk is about the parts of a consumer internet service that a reader cannot see from the outside: how a request is routed, where the work is done, what is cached and for how long, and which of those decisions is expensive to reverse later. A vendor announcement is not a story here until somebody has said what it replaces and what it costs to run. An architecture diagram is not a story either, unless the piece can say which constraint the shape of the diagram is answering.

The recurring question on the desk is where the trade-off was made. Low latency is bought with duplication, strong consistency is bought with waiting, a smaller bundle is bought with somebody’s build time. A piece that describes one side of that exchange and not the other is incomplete, and the desk sends it back. Figures — percentile targets, timeouts, payload budgets — are quoted with their unit and their qualifier, and sourced to a specification or to the vendor’s own documentation with the date it was read.