Selenium WebDriver Capability Interoperability Between ChromeDriver and GeckoDriver for Unrecognized Vendor Keys

Unrecognized Vendor Capabilities Handling Across Selenium Browser Drivers

Selenium WebDriver client bindings transmit capabilities as JSON objects during new-session creation. The W3C WebDriver specification defines standard capability names while browser-specific options use vendor prefixes such as goog:chromeOptions and moz:firefoxOptions. However, the specification does not mandate uniform behavior for capabilities a driver does not recognize; implementations can ignore, reject, or transform them depending on driver version and protocol dialect.

ChromeDriver and GeckoDriver have historically exhibited different behaviors when encountering unknown top-level capabilities. Some drivers return invalid argument or session not created errors, while others silently ignore unrecognized keys. This variance becomes more pronounced when switching between W3C protocol and legacy OSS dialect, and Selenium Grid introduces additional complexity by potentially filtering or rewriting capabilities at the hub-node boundary before session creation.

The behavior is inherently version-sensitive and binding-dependent. A capability that succeeds with one driver version may fail with another, even when using the same client binding. Grid proxy and node version skew can alter capability validation independently of client-driver consistency.

What is the current interoperability behavior when an unrecognized vendor-prefixed capability is passed during new-session creation with ChromeDriver and GeckoDriver using the W3C protocol? Does the driver reject the session with an error, silently ignore the capability, or exhibit other handling? How does this behavior differ between driver versions and when using Selenium Grid with mixed hub-node versions?