Re: [w3ctag/design-reviews] WebGPU Compatibility Mode (Issue #1063)

SenorBlanco left a comment (w3ctag/design-reviews#1063)

Thank you for your comments! Your time and attention is appreciated.

1. The additional fingerprinting surfaces that we know of are:
  a. The 'core-features-and-limits' feature, which will be exposed on Core-supporting devices but absent on Compat-only devices.
  b. WebGPU limits in OpenGL ES drivers. These limits will be bucketed using the same approach is used in core WebGPU, and those buckets must be chosen carefullly in order to capture OpenGL ES capabilities at a sufficient granularity. For example, storage buffers in vertex shaders are unsupported on many popular Mali devices, so we've added a new limit ('maxStorageBuffersInVertexStage') which must be zero on those devices, and non-zero elsewhere.

Note that the presence or absence of Compatibility mode itself in Chrome will not be an additional fingerprinting surface, since Chrome plans to ship Compatibility mode JS API on all devices. (It could be used to distinguish Chrome from other browers, or older versions, but these are bits that are already available via other means.)

2. It's a good point, and one that we've discussed a fair bit in the group. However, it's not something we plan to spec in the short- or medium-term. There thinking is that there is not a great deal of incremental reach to be had going from Compat to "Compat-plus". However, if further research did reveal a large contingent of useful devices in between Compat and Core, we might consider it in the future.

I'll add some text to the explainer to summarize the above.

-- 
Reply to this email directly or view it on GitHub:
https://github.com/w3ctag/design-reviews/issues/1063#issuecomment-2884873715
You are receiving this because you are subscribed to this thread.

Message ID: <w3ctag/design-reviews/issues/1063/2884873715@github.com>