Browser Features Check - Compatibility and APIs

doc

Review browser feature support to understand compatibility, privacy exposure, and API availability across different browser profiles and configurations.

Feature detection asks what the browser can do, not what it says it is. The answer is a set — WebRTC yes or no, WebGL yes or no, WebAssembly yes or no, permissions available or not — and that set is more informative than any header, because it reflects the actual build and the configuration it is running under.

The features worth checking

FeatureWhat its presence impliesPermission needed
WebRTC data channelsPeer-to-peer capability; historically leaks local addressesNo, for data channels
WebGL and WebGPUGPU acceleration path presentNo
WebAssemblyModern build; nearly universal nowNo
Service workersBackground and cache capabilityNo, but requires secure context
WebAuthnHardware-backed authentication availableYes, at use time
Media device enumerationCameras and microphones presentYes, before labels appear
Notifications, geolocationPrompt-gated capabilitiesYes, each time

The top rows are the discriminative ones, because they vary by hardware and by policy. The bottom rows are gated behind prompts, so a page that has not asked learns only that the API exists — which every modern browser reports.

Why the set matters more than the answer

Any single feature is close to universal. WebAssembly is available almost everywhere; so is WebGL, in some form. The useful signal is the combination, and the combination is shaped by three independent forces:

  • Build. Which browser and which version. Older builds lack newer APIs; that is expected and not suspicious.
  • Policy. Enterprise and privacy configurations disable specific APIs. A browser with WebRTC disabled is configured, not broken.
  • Hardware. A machine without a discrete GPU may lack WebGL2 or WebGPU. The absence reflects the device.

A combination that cannot be explained by any of these three is worth a second look. A combination that can be explained by one of them is ordinary.

Feature detection is the correct engineering practice

This is the part often lost in privacy discussions. Detecting features is how a site decides which code path to run, and it is more reliable than parsing a user agent. The same call that tells a site "this browser supports WebGPU" also tells a fingerprinting script "this machine has a discrete GPU". Both are reading the same surface.

Practical consequences:

  1. Breaking feature detection to reduce fingerprinting breaks functionality. A site that cannot detect WebGPU falls back to a slower path or fails.
  2. Reporting an absent feature is itself a signal. Disabling WebRTC removes one property and adds a distinctive one.
  3. Uniformity is the only defence. A browser that reports every feature as present is as distinctive as one that reports none.

Reading a feature report

  • Check the secure-context requirement. Some features are only available over HTTPS. Their absence on an HTTP page is expected, not a defect.
  • Separate presence from permission. A page can see that the camera API exists without seeing any camera. Labels appear only after consent.
  • Expect version-driven gaps. Missing newer APIs usually means an older build. Missing older APIs usually means a hardened configuration.
  • Do not read absence as intent. Disabled WebRTC is a setting. It is not evidence of anything about the user.

What cannot be changed

You can enable or disable specific APIs through browser settings and enterprise policy. You cannot add hardware you do not have, and you cannot make an older build support APIs it never shipped. The realistic goal is to know what your configuration reports and why, not to make it look like someone else's.

Common misreads

  • "WebRTC is available, so my local IP is exposed." Availability is a precondition. Whether addresses leak depends on the browser's mDNS policy and the specific API used.
  • "My feature set is unusual, so I am being tracked." Unusual means configured. Tracking would require the site to correlate, which is a different question.
  • "A missing API means the browser is broken." It usually means a policy or an older build, both legitimate.
  • "Feature detection is fingerprinting." It is a shared surface. The difference is what is done with the answer, not the call itself.

FAQ

Is feature detection the same as fingerprinting?

Mechanically, yes — the same existence checks serve both. The distinction is purpose and disclosure, not the API call.

Why does a feature work on one site and not another?

Secure context, permissions policy, and per-origin consent. The browser can disable a feature for one origin and allow it for another.

Can I hide which features I support?

Not meaningfully. Removing a feature is visible; the absence becomes part of the profile.

Should I enable everything for uniformity?

Only if nothing needs to be off. A configuration chosen for uniformity is still a configuration, and it has its own signature.