Permissions Policy Guide (Previously Feature Policy)

Permissions policy is the new name for feature policy with a new HTTP header which uses Structured Field syntax.

How to add a new feature to permissions policy

Permissions policy (see spec) is a mechanism that allows developers to selectively enable and disable various browser features an APIs (e.g, "fullscreen", "usb", "web-share", etc.). A permissions policy can be defined via a HTTP header and/or an iframe "allow" attribute.

Below is an example of a header policy (note that the header should be kept in one line, split into multiple for clarity reasons):

Permissions-Policy: web-share=(), geolocation=(self https://example.com), camera=*

• web-share is disabled for all browsing contexts;
• geolocation is disabled for all browsing contexts except for its own origin and those whose origin is "https://example.com";
• camera is enabled for all browsing contexts.

Below is an example of a container policy:

<iframe allowpaymentrequest allow='web-share; fullscreen'></iframe>

OR

<iframe allowpaymentrequest allow="web-share 'src'; fullscreen 'src'"></iframe>

• payment is enabled (via allowpaymentrequest) on all browsing contexts within the iframe;
• web-share and fullscreen are enabled on the origin of the URL of the iframe's src attribute.

Combined with a header policy and a container policy, inherited policy defines the availability of a feature. See more details for how to define an inherited policy for feature

Adding a new feature to permissions policy

A step-to-step guide with examples.

Shipping features behind a flag

If the additional feature is unshipped, or if the correct behaviour with feature policy is undetermined, consider shipping the feature behind a runtime-enabled feature.

Define new feature

1. Permissions policy features are defined in services/network/public/cpp/permissions_policy/permissions_policy_features.json5. Add the new feature, placing any runtime-enabled feature or origin trial dependencies in its "depends_on" field as described in the file's comments. This list is used to generate permissions_policy_helper.cc.

2. Append the new feature enum with a brief description as well in services/network/public/mojom/permissions_policy/permissions_policy_feature.mojom. The enum must have the same name as the name field in the json5 file from step 1. Run tools/metrics/histograms/update_permissions_policy_enum.py to update enums.xml from the mojo enum.

3. Append the new feature name to the PermissionsPolicyFeature enum in third_party/blink/public/devtools_protocol/browser_protocol.pdl.

4. Add the new feature name to third_party/blink/web_tests/webexposed/feature-policy-features-expected.txt and third_party/blink/web_tests/wpt_internal/isolated-permissions-policy/permissions_policy.https.html.

5. Send a Pull Request to the webappsec-permissions-policy github repo in order to propose the new permissions policy name. See: https://github.com/w3c/webappsec-permissions-policy/blob/main/features.md

Integrate the feature behaviour with permissions policy

1. The most common way to check if features are enabled is ExecutionContext::IsFeatureEnabled.

2. Examples:

• web-share: NavigatorShare::canShare()
• payment: AllowedToUsePaymentRequest()
• usb: USB::getDevices()

Write web-platform-tests

To test the new feature with permissions policy, refer to third_party/blink/web_tests/external/wpt/permissions-policy/README.md for instructions on how to use the permissions policy test framework.

Fenced frame considerations

Reach out to third_party/blink/renderer/core/html/fenced_frame/OWNERS when adding a new permissions-backed feature so they can audit it for use with the Fenced Frames API, or feel free to add to the audit at content/browser/fenced_frame/PERMISSIONS_POLICIES.md and send to third_party/blink/renderer/core/html/fenced_frame/OWNERS for review.