Six Months With Tauri: What You Gain and What It Costs

An engineer building a desktop database client (Tabularis) shares six months of production experience choosing Tauri over Electron and native development. The deciding factor wasn't the small installer size but the ability to write the backend as an ordinary Rust program instead of JavaScript: SQLx, Tokio, rustls and keyring compose with normal Rust types, while privileged operations cross an explicit IPC boundary through Tauri commands rather than being implicitly available.

That architecture has a real cost. Tauri doesn't bundle a browser engine; it uses whatever webview the OS provides — WKWebView on macOS, WebKitGTK on Linux, WebView2 on Windows. That means supporting three rendering engines instead of one, with real divergence in CSS handling, input behavior and GPU compositing. The sharpest pain point is Linux, where distribution-specific WebKitGTK builds and Wayland quirks have caused blank windows and crashes, and where the portable AppImage format — needed because you can't assume dependencies are present — balloons to roughly 94 MB versus 15-18 MB for other package formats.

The practical takeaway for engineers: Electron charges you for the runtime, native charges you for the number of platforms, and Tauri charges you for the seams between webview implementations. The small-binary pitch is real, but it's only half the story — the actual engineering cost shows up in cross-webview testing and in the messier corners of the Linux desktop ecosystem.