From 47dc99757093b43a5e36c91f4208db9876935dff Mon Sep 17 00:00:00 2001
From: Anssi Kostiainen <anssi.kostiainen@intel.com>
Date: Thu, 30 Jun 2022 15:11:12 +0000
Subject: [PATCH] Merge pull request #271 from
 webmachinelearning/privacy-considerations-refresh

SHA: 35c7a47c862d777b1cbc8bb476d50e07b422a1a8
Reason: push, by @anssiko

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
---
 index.html | 84 ++++++++++++++++++++++++++++++------------------------
 1 file changed, 46 insertions(+), 38 deletions(-)

diff --git a/index.html b/index.html
index 5c1dda96..56dc1fab 100644
--- a/index.html
+++ b/index.html
@@ -8,7 +8,7 @@
   <link href="https://www.w3.org/2008/site/images/favicon.ico" rel="icon">
   <meta content="Bikeshed version 37b7e7e68, updated Fri May 27 15:08:11 2022 -0700" name="generator">
   <link href="https://www.w3.org/TR/webnn/" rel="canonical">
-  <meta content="fe728b04f752714f2db13a9cd7a02d428a54d67c" name="document-revision">
+  <meta content="35c7a47c862d777b1cbc8bb476d50e07b422a1a8" name="document-revision">
 <style>
 /* Make <dl> blocks more distinct from their surroundings. */
 main dl:not(.switch) {
@@ -732,7 +732,7 @@
   <div class="head">
    <p data-fill-with="logo"><a class="logo" href="https://www.w3.org/"> <img alt="W3C" height="48" src="https://www.w3.org/StyleSheets/TR/2021/logos/W3C" width="72"> </a> </p>
    <h1 class="p-name no-ref" id="title">Web Neural Network API</h1>
-   <p id="w3c-state"><a href="https://www.w3.org/standards/types#ED">Editor’s Draft</a>, <time class="dt-updated" datetime="2022-06-28">28 June 2022</time></p>
+   <p id="w3c-state"><a href="https://www.w3.org/standards/types#ED">Editor’s Draft</a>, <time class="dt-updated" datetime="2022-06-30">30 June 2022</time></p>
    <details open>
     <summary>More details about this document</summary>
     <div data-fill-with="spec-metadata">
@@ -1100,6 +1100,9 @@ <h2 class="heading settled" data-level="4" id="privacy"><span class="secno">4. <
    <p>No information from the underlying platform is exposed directly. An execution time analysis may reveal indirectly the performance of the underlying platform’s neural network hardware acceleration capabilities relative to another underlying platform.</p>
    <p class="note" role="note"><span>Note:</span> The group is <a href="https://github.com/webmachinelearning/webnn/issues/85">soliciting further input</a> on the proposed execution time analysis fingerprinting vector and will augment this section with more information and mitigations to inform the implementers of this API.</p>
    <p>Unlike WebGPU, this API does not intrinsically support custom shader authoring; and as a result is not prone to timing attacks that rely on shader caches, or other persistent data. The API builds upon pre-existing shaders and lower level primitives of the browser or the underlying OS. Web developers who interface with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device①">GPUDevice</a></code> are expected to be aware of <a href="https://gpuweb.github.io/gpuweb/#privacy-user-agent-state">WebGPU compilation cache considerations</a>.</p>
+   <p>The WebGPU API identifies <a href="https://gpuweb.github.io/gpuweb/#privacy-machine-artifacts">machine-specific artifacts</a> as a privacy consideration. Given the WebNN API defines means to record an ML worload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers">GPUCommandBuffer</a></code>, compute unit scheduling may under certain circumstances introduce a fingerprint. However, similarly to WebGPU, such fingerprints are identical across most or all of the devices of each vendor, mitigating the concern. Furthermore, software implementations can be used to further eliminate such artifacts.</p>
+   <p>The WebNN API defines two developer-settable preferences to help inform <a href="#programming-model-device-selection">§ 6.2 Device Selection</a> and allow the implementation better select the most appropriate underlying execution device for the workload. <a data-link-type="dfn" href="#device-type" id="ref-for-device-type">Device type</a> normatively indicates the kind of device and is either "cpu" or "gpu". If this type cannot be satisfied, an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror">OperationError</a></code> exception is thrown, thus this type can in some cases add two bits of entropy to the fingerprint. <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference">Power preference</a> indicates preference as related to the power consumption and is considered a hint only and as such does not increase entropy of the fingerprint.</p>
+   <p>If a future version of this specification introduces support for new a <a data-link-type="dfn" href="#device-type" id="ref-for-device-type①">device type</a> that can only support a subset of <code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype">MLOperandType</a></code>s, that may introduce a new fingerprint.</p>
    <p>In general, implementers of this API are expected to be familiar with the <a href="https://gpuweb.github.io/gpuweb/#privacy-considerations">WebGPU Privacy Considerations</a>.</p>
    <h2 class="heading settled" data-level="5" id="ethics"><span class="secno">5. </span><span class="content">Ethical Considerations</span><a class="self-link" href="#ethics"></a></h2>
    <p>The Working Group has started documenting ethical issues associated with using Machine Learning on the Web, to help identify what mitigations its normative specifications should take into account. The Working Group publishes and maintains an Ethical Principles for Web Machine Learning document <a data-link-type="biblio" href="#biblio-webmachinelearning-ethics">[webmachinelearning-ethics]</a> open to contributions from the wider community via a dedicated <a href="https://github.com/webmachinelearning/webmachinelearning-ethics">GitHub repository</a>.</p>
@@ -1150,13 +1153,13 @@ <h3 class="heading settled" data-level="6.1" id="programming-model-overview"><sp
    <p>The <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder">MLCommandEncoder</a></code> interface created by the <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑦">MLContext</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-createcommandencoder" id="ref-for-dom-mlcontext-createcommandencoder">createCommandEncoder()</a></code> method supports
 a graph execution method that provides the maximum flexibility to callers that also utilize WebGPU in their 
 application. It does this by placing the workload required to initialize and compute the results of the 
-operations in the graph onto a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers">GPUCommandBuffer</a></code>. The callers are responsible for the eventual submission 
+operations in the graph onto a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers①">GPUCommandBuffer</a></code>. The callers are responsible for the eventual submission 
 of this workload on the <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#queues" id="ref-for-queues">GPUQueue</a></code> through the WebGPU queue submission mechanism. Once the submitted workload 
 is completely executed, the result is avaialble in the bound output buffers.</p>
    <h3 class="heading settled" data-level="6.2" id="programming-model-device-selection"><span class="secno">6.2. </span><span class="content">Device Selection</span><a class="self-link" href="#programming-model-device-selection"></a></h3>
    <p>An <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑧">MLContext</a></code> interface represents a global state of neural network execution. One of the important context states is the underlying execution device that manages the resources and facilitates the compilation and the eventual execution of the neural network graph. In addition to the default method of creation with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions">MLContextOptions</a></code>, an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑨">MLContext</a></code> could also be created from a specific <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device②">GPUDevice</a></code> that is already in use by the application, in which case the corresponding <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface">GPUBuffer</a></code> resources used as graph constants, as well as the <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#texture-interface" id="ref-for-texture-interface">GPUTexture</a></code> as graph inputs must also be created from the same device. In a multi-adapter configuration, the device used for <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⓪">MLContext</a></code> must be created from the same adapter as the device used to allocate the resources referenced in the graph.</p>
    <p>In a situation when a GPU context executes a graph with a constant or an input in the system memory as an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView">ArrayBufferView</a></code>, the input content is automatically uploaded from the system memory to the GPU memory, and downloaded back to the system memory of an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView①">ArrayBufferView</a></code> output buffer at the end of the graph execution. This data upload and download cycles will only occur whenever the execution device requires the data to be copied out of and back into the system memory, such as in the case of the GPU. It doesn’t occur when the device is a CPU device. Additionally, the result of the graph execution is in a known layout format. While the execution may be optimized for a native memory access pattern in an intermediate result within the graph, the output of the last operation of the graph must convert the content back to a known layout format at the end of the graph in order to maintain the expected behavior from the caller’s perspective.</p>
-   <p>When an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①①">MLContext</a></code> is created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions①">MLContextOptions</a></code>, the user agent selects and creates the underlying execution device by taking into account the application’s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference">power preference</a> and <a data-link-type="dfn" href="#device-type" id="ref-for-device-type">device type</a> specified in the <code class="idl"><a data-link-type="idl" href="#enumdef-mlpowerpreference" id="ref-for-enumdef-mlpowerpreference">MLPowerPreference</a></code> and <code class="idl"><a data-link-type="idl" href="#enumdef-mldevicetype" id="ref-for-enumdef-mldevicetype">MLDeviceType</a></code> options.</p>
+   <p>When an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①①">MLContext</a></code> is created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions①">MLContextOptions</a></code>, the user agent selects and creates the underlying execution device by taking into account the application’s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference①">power preference</a> and <a data-link-type="dfn" href="#device-type" id="ref-for-device-type②">device type</a> specified in the <code class="idl"><a data-link-type="idl" href="#enumdef-mlpowerpreference" id="ref-for-enumdef-mlpowerpreference">MLPowerPreference</a></code> and <code class="idl"><a data-link-type="idl" href="#enumdef-mldevicetype" id="ref-for-enumdef-mldevicetype">MLDeviceType</a></code> options.</p>
    <p>The following table summarizes the types of resource supported by the context created through different method of creation:</p>
    <div class="note" role="note">
     <table>
@@ -1236,7 +1239,7 @@ <h4 class="heading settled" data-level="7.2.1" id="permissions-policy-integratio
 string "<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="webnn-feature" data-noexport id="webnn-feature">webnn</dfn></code>".
 Its <a data-link-type="dfn" href="https://w3c.github.io/webappsec-permissions-policy/#default-allowlist" id="ref-for-default-allowlist">default allowlist</a> is <code>'self'</code>.</p>
    <h3 class="heading settled" data-level="7.3" id="api-mlcontext"><span class="secno">7.3. </span><span class="content">MLContext</span><a class="self-link" href="#api-mlcontext"></a></h3>
-    The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑤">MLContext</a></code> interface represents a global state of neural network compute workload and execution processes. Each <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑥">MLContext</a></code> object has associated <a data-link-type="dfn" href="#context-type" id="ref-for-context-type">context type</a>, <a data-link-type="dfn" href="#device-type" id="ref-for-device-type①">device type</a> and <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference①">power preference</a>. 
+    The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑤">MLContext</a></code> interface represents a global state of neural network compute workload and execution processes. Each <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑥">MLContext</a></code> object has associated <a data-link-type="dfn" href="#context-type" id="ref-for-context-type">context type</a>, <a data-link-type="dfn" href="#device-type" id="ref-for-device-type③">device type</a> and <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference②">power preference</a>. 
    <p>The <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="context-type">context type</dfn> is the type of the execution context that manages the resources and facilitates the compilation and execution of the neural network graph:</p>
    <dl>
     <dt>"<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="default-context" data-noexport id="default-context">default</dfn></code>"
@@ -1276,16 +1279,16 @@ <h3 class="heading settled" data-level="7.3" id="api-mlcontext"><span class="sec
     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-contexttype-slot"><code>[[contextType]]</code></dfn> of type <a data-link-type="dfn" href="#context-type" id="ref-for-context-type①">context type</a>
     <dd data-md>
      <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑧">MLContext</a></code>'s <a data-link-type="dfn" href="#context-type" id="ref-for-context-type②">context type</a>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-devicetype-slot"><code>[[deviceType]]</code></dfn> of type <a data-link-type="dfn" href="#device-type" id="ref-for-device-type②">device type</a>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-devicetype-slot"><code>[[deviceType]]</code></dfn> of type <a data-link-type="dfn" href="#device-type" id="ref-for-device-type④">device type</a>
     <dd data-md>
-     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑨">MLContext</a></code>'s <a data-link-type="dfn" href="#device-type" id="ref-for-device-type③">device type</a>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-powerpreference-slot"><code>[[powerPreference]]</code></dfn> of type <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference②">power preference</a>
+     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑨">MLContext</a></code>'s <a data-link-type="dfn" href="#device-type" id="ref-for-device-type⑤">device type</a>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-powerpreference-slot"><code>[[powerPreference]]</code></dfn> of type <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference③">power preference</a>
     <dd data-md>
-     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⓪">MLContext</a></code>'s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference③">power preference</a>.</p>
+     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⓪">MLContext</a></code>'s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference④">power preference</a>.</p>
    </dl>
    <div class="note" role="note"> When the <code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot②">[[contextType]]</a></code> is set to <a data-link-type="dfn" href="#default-context" id="ref-for-default-context①">default</a> with the <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑥">MLContextOptions</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontextoptions-devicetype" id="ref-for-dom-mlcontextoptions-devicetype①">deviceType</a></code> set to <a data-link-type="dfn" href="#device-type-gpu" id="ref-for-device-type-gpu①">gpu</a>, the user agent is responsible for creating an internal GPU device that operates within the context and is capable of ML workload submission on behalf of the calling application. In this setting however, only <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑤">ArrayBufferView</a></code> inputs and outputs are allowed in and out of the graph execution since the application has no way to know what type of internal GPU device is being created on their behalf. In this case, the user agent is responsible for automatic uploads and downloads of the inputs and outputs to and from the GPU memory using this said internal device. </div>
    <h4 class="heading settled" data-level="7.3.1" id="api-mlcontext-sync-execution"><span class="secno">7.3.1. </span><span class="content">Synchronous Execution</span><a class="self-link" href="#api-mlcontext-sync-execution"></a></h4>
-    Synchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑥">MLGraph</a></code> on the calling thread, which must be a worker thread, to produce results as defined by the operations in the graph. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②①">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑦">MLContextOptions</a></code>. Otherwise, it throws an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror">OperationError</a></code> exception. 
+    Synchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑥">MLGraph</a></code> on the calling thread, which must be a worker thread, to produce results as defined by the operations in the graph. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②①">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑦">MLContextOptions</a></code>. Otherwise, it throws an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①">OperationError</a></code> exception. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext②②"><c- g>MLContext</c-></a> {
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed②"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
   <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined"><c- b>undefined</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export data-lt="compute(graph, inputs, outputs)" id="dom-mlcontext-compute"><code><c- g>compute</c-></code></dfn>(<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph⑦"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/compute(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-compute-graph-inputs-outputs-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcontext-compute-graph-inputs-outputs-graph"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarrayinputs" id="ref-for-typedefdef-mlnamedarrayinputs①"><c- n>MLNamedArrayInputs</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/compute(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-compute-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlcontext-compute-graph-inputs-outputs-inputs"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarrayoutputs" id="ref-for-typedefdef-mlnamedarrayoutputs①"><c- n>MLNamedArrayOutputs</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/compute(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-compute-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlcontext-compute-graph-inputs-outputs-outputs"></a></dfn>);
@@ -1418,7 +1421,7 @@ <h4 class="heading settled" data-level="7.3.1" id="api-mlcontext-sync-execution"
         <p>If there is an error returned by <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot③">[[implementation]]</a></code>, then:</p>
         <ol>
          <li data-md>
-          <p>Throw an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①">OperationError</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②">DOMException</a></code> and stop.</p>
+          <p>Throw an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②">OperationError</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②">DOMException</a></code> and stop.</p>
         </ol>
        <li data-md>
         <p>Else:</p>
@@ -1525,7 +1528,7 @@ <h5 class="heading settled" data-level="7.3.1.1" id="compilation-examples"><span
 </pre>
    </div>
    <h4 class="heading settled" data-level="7.3.2" id="api-mlcontext-async-execution"><span class="secno">7.3.2. </span><span class="content">Asynchronous Execution</span><a class="self-link" href="#api-mlcontext-async-execution"></a></h4>
-    Asynchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑨">MLGraph</a></code> on a separate timeline, either on a worker thread for the CPU execution, or on a GPU timeline for the submission of GPU workload on the command queue. The asynchronous nature of this call avoids blocking the calling thread while the computation for result is ongoing. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②③">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑧">MLContextOptions</a></code>. Otherwise, it throws an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②">OperationError</a></code> exception. 
+    Asynchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑨">MLGraph</a></code> on a separate timeline, either on a worker thread for the CPU execution, or on a GPU timeline for the submission of GPU workload on the command queue. The asynchronous nature of this call avoids blocking the calling thread while the computation for result is ongoing. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②③">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑧">MLContextOptions</a></code>. Otherwise, it throws an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③">OperationError</a></code> exception. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext②④"><c- g>MLContext</c-></a> {
   <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise①"><c- b>Promise</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined③"><c- b>undefined</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export data-lt="computeAsync(graph, inputs, outputs)" id="dom-mlcontext-computeasync"><code><c- g>computeAsync</c-></code></dfn>(<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph①⓪"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/computeAsync(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-computeasync-graph-inputs-outputs-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcontext-computeasync-graph-inputs-outputs-graph"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarrayinputs" id="ref-for-typedefdef-mlnamedarrayinputs③"><c- n>MLNamedArrayInputs</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/computeAsync(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-computeasync-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlcontext-computeasync-graph-inputs-outputs-inputs"></a></dfn>, 
                     <a data-link-type="idl-name" href="#typedefdef-mlnamedarrayoutputs" id="ref-for-typedefdef-mlnamedarrayoutputs③"><c- n>MLNamedArrayOutputs</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/computeAsync(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-computeasync-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlcontext-computeasync-graph-inputs-outputs-outputs"></a></dfn>);
@@ -1650,7 +1653,7 @@ <h4 class="heading settled" data-level="7.3.2" id="api-mlcontext-async-execution
         <p>If there is an error returned by <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot⑧">[[implementation]]</a></code>, then:</p>
         <ol>
          <li data-md>
-          <p>Throw an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③">OperationError</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥">DOMException</a></code> and stop.</p>
+          <p>Throw an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④">OperationError</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥">DOMException</a></code> and stop.</p>
         </ol>
        <li data-md>
         <p>Else:</p>
@@ -1686,7 +1689,7 @@ <h4 class="heading settled" data-level="7.3.2" id="api-mlcontext-async-execution
     </ol>
    </div>
    <h4 class="heading settled" data-level="7.3.3" id="api-mlcontext-webgpu-interop"><span class="secno">7.3.3. </span><span class="content">WebGPU Interoperability</span><a class="self-link" href="#api-mlcontext-webgpu-interop"></a></h4>
-    Create <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder①">MLCommandEncoder</a></code> interface used to record the ML workload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers①">GPUCommandBuffer</a></code> to allow mixing of ML workload with other GPU workload in an application that leverages WebGPU. This method only succeeds on an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑤">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device⑤">GPUDevice</a></code>. Otherwise, it throws an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④">OperationError</a></code> exception. 
+    Create <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder①">MLCommandEncoder</a></code> interface used to record the ML workload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers②">GPUCommandBuffer</a></code> to allow mixing of ML workload with other GPU workload in an application that leverages WebGPU. This method only succeeds on an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑤">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device⑤">GPUDevice</a></code>. Otherwise, it throws an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑤">OperationError</a></code> exception. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext②⑥"><c- g>MLContext</c-></a> {
   <a data-link-type="idl-name" href="#mlcommandencoder" id="ref-for-mlcommandencoder②"><c- n>MLCommandEncoder</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export data-lt="createCommandEncoder()" id="dom-mlcontext-createcommandencoder"><code><c- g>createCommandEncoder</c-></code></dfn>();
 };
@@ -1709,7 +1712,7 @@ <h3 class="heading settled" data-level="7.4" id="api-mloperanddescriptor"><span
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mloperanddescriptor"><code><c- g>MLOperandDescriptor</c-></code></dfn> {
   // The operand type.
-  <c- b>required</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype"><c- n>MLOperandType</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLOperandDescriptor" data-dfn-type="dict-member" data-export data-type="MLOperandType " id="dom-mloperanddescriptor-type"><code><c- g>type</c-></code></dfn>;
+  <c- b>required</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype①"><c- n>MLOperandType</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLOperandDescriptor" data-dfn-type="dict-member" data-export data-type="MLOperandType " id="dom-mloperanddescriptor-type"><code><c- g>type</c-></code></dfn>;
 
   // The dimensions field is only required for tensor operands.
   // The negative value means an unknown dimension.
@@ -1754,7 +1757,7 @@ <h3 class="heading settled" data-level="7.7" id="api-mlgraphbuilder"><span class
   <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①"><c- n>MLOperand</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="constant(desc, bufferView)" id="dom-mlgraphbuilder-constant"><code><c- g>constant</c-></code></dfn>(<a data-link-type="idl-name" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①"><c- n>MLOperandDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(desc, bufferView)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-desc-bufferview-desc"><code><c- g>desc</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-desc-bufferview-desc"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview"><c- n>MLBufferView</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(desc, bufferView)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-desc-bufferview-bufferview"><code><c- g>bufferView</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-desc-bufferview-bufferview"></a></dfn>);
 
   // Create a single-value operand from the specified number of the specified type.
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="constant(value, type)|constant(value)" id="dom-mlgraphbuilder-constant-value-type"><code><c- g>constant</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type"></a></dfn>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-double" id="ref-for-idl-double"><c- b>double</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-value"><code><c- g>value</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-value"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype①"><c- n>MLOperandType</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-type"><code><c- g>type</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-type"></a></dfn> = "float32");
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="constant(value, type)|constant(value)" id="dom-mlgraphbuilder-constant-value-type"><code><c- g>constant</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type"></a></dfn>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-double" id="ref-for-idl-double"><c- b>double</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-value"><code><c- g>value</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-value"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype②"><c- n>MLOperandType</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-type"><code><c- g>type</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-type"></a></dfn> = "float32");
 
   // Compile the graph up to the specified output operands synchronously.
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑥"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
@@ -3435,7 +3438,7 @@ <h3 class="heading settled" data-level="7.8" id="api-mlgraph"><span class="secno
      <p>The underlying implementation provided by the User Agent.</p>
    </dl>
    <h3 class="heading settled" data-level="7.9" id="api-mlcommandencoder"><span class="secno">7.9. </span><span class="content">MLCommandEncoder</span><a class="self-link" href="#api-mlcommandencoder"></a></h3>
-    The <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder④">MLCommandEncoder</a></code> interface represents a method of execution that synchronously records the computational workload of a compiled <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph②⓪">MLGraph</a></code> to a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers②">GPUCommandBuffer</a></code> on the calling thread. Since the workload is not immediately executed, just recorded, this method allows more flexibility for the caller to determine how and when the recorded commands will be submitted for execution on the GPU relative to other GPU workload on the same or different queue. 
+    The <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder④">MLCommandEncoder</a></code> interface represents a method of execution that synchronously records the computational workload of a compiled <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph②⓪">MLGraph</a></code> to a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers③">GPUCommandBuffer</a></code> on the calling thread. Since the workload is not immediately executed, just recorded, this method allows more flexibility for the caller to determine how and when the recorded commands will be submitted for execution on the GPU relative to other GPU workload on the same or different queue. 
 <pre class="idl highlight def"><c- b>typedef</c-> (<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface②"><c- n>GPUBuffer</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#texture-interface" id="ref-for-texture-interface①"><c- n>GPUTexture</c-></a>) <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlgpuresource"><code><c- g>MLGPUResource</c-></code></dfn>;
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlgpuinput"><code><c- g>MLGPUInput</c-></code></dfn> {
@@ -3598,7 +3601,7 @@ <h4 class="heading settled" data-level="7.9.2" id="api-mlcommandencoder-dispatch
         <p>If there is an error returned by <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot①③">[[implementation]]</a></code>, then:</p>
         <ol>
          <li data-md>
-          <p>Throw an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑤">OperationError</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪">DOMException</a></code> and stop.</p>
+          <p>Throw an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑥">OperationError</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪">DOMException</a></code> and stop.</p>
         </ol>
        <li data-md>
         <p>Else:</p>
@@ -3634,9 +3637,9 @@ <h4 class="heading settled" data-level="7.9.2" id="api-mlcommandencoder-dispatch
     </ol>
    </div>
    <h4 class="heading settled" data-level="7.9.3" id="api-mlcommandencoder-generate-gpu-command-buffer"><span class="secno">7.9.3. </span><span class="content">Generate GPU Command Buffer</span><a class="self-link" href="#api-mlcommandencoder-generate-gpu-command-buffer"></a></h4>
-    Complete the recording of ML workload and return a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers③">GPUCommandBuffer</a></code> containing the recorded workload. 
+    Complete the recording of ML workload and return a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers④">GPUCommandBuffer</a></code> containing the recorded workload. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑨"><c- g>MLCommandEncoder</c-></a> {
-  <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers④"><c- n>GPUCommandBuffer</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export data-lt="finish(descriptor)|finish()" id="dom-mlcommandencoder-finish"><code><c- g>finish</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor" id="ref-for-dictdef-gpucommandbufferdescriptor"><c- n>GPUCommandBufferDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/finish(descriptor), MLCommandEncoder/finish()" data-dfn-type="argument" data-export id="dom-mlcommandencoder-finish-descriptor-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish-descriptor-descriptor"></a></dfn> = {});
+  <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers⑤"><c- n>GPUCommandBuffer</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export data-lt="finish(descriptor)|finish()" id="dom-mlcommandencoder-finish"><code><c- g>finish</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor" id="ref-for-dictdef-gpucommandbufferdescriptor"><c- n>GPUCommandBufferDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/finish(descriptor), MLCommandEncoder/finish()" data-dfn-type="argument" data-export id="dom-mlcommandencoder-finish-descriptor-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish-descriptor-descriptor"></a></dfn> = {});
 };
 </pre>
    <div class="algorithm" data-algorithm="mlcommandencoder.finish">
@@ -3645,7 +3648,7 @@ <h4 class="heading settled" data-level="7.9.3" id="api-mlcommandencoder-generate
      <li data-md>
       <p><em>descriptor</em>: an optional <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor" id="ref-for-dictdef-gpucommandbufferdescriptor①">GPUCommandBufferDescriptor</a></code>. Descriptor of the command buffer.</p>
     </ul>
-    <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers⑤">GPUCommandBuffer</a></code>.</p>
+    <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers⑥">GPUCommandBuffer</a></code>.</p>
    </div>
    <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8. </span><span class="content">Examples</span><a class="self-link" href="#examples"></a></h2>
    <div class="example" id="example-8a846be1">
@@ -3722,11 +3725,11 @@ <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8.
 </pre>
    </div>
    <h2 class="heading settled" data-level="9" id="appendices"><span class="secno">9. </span><span class="content">Appendices</span><a class="self-link" href="#appendices"></a></h2>
-   <h3 class="heading settled" data-level="9.1" id="appendices-mloperandtype-arraybufferview-compatibility"><span class="secno">9.1. </span><span class="content"><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype②">MLOperandType</a></code> and <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView①①">ArrayBufferView</a></code> compatibility</span><a class="self-link" href="#appendices-mloperandtype-arraybufferview-compatibility"></a></h3>
+   <h3 class="heading settled" data-level="9.1" id="appendices-mloperandtype-arraybufferview-compatibility"><span class="secno">9.1. </span><span class="content"><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype③">MLOperandType</a></code> and <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView①①">ArrayBufferView</a></code> compatibility</span><a class="self-link" href="#appendices-mloperandtype-arraybufferview-compatibility"></a></h3>
    <table class="data">
     <thead class="stickyheader">
      <tr>
-      <th><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype③">MLOperandType</a></code> 
+      <th><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype④">MLOperandType</a></code> 
       <th><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView①②">ArrayBufferView</a></code> 
     <tbody>
      <tr>
@@ -4295,10 +4298,11 @@ <h3 class="no-num no-ref heading settled" id="index-defined-here"><span class="c
   <aside class="dfn-panel" data-for="term-for-command-buffers">
    <a href="https://gpuweb.github.io/gpuweb/#command-buffers">https://gpuweb.github.io/gpuweb/#command-buffers</a><b>Referenced in:</b>
    <ul>
-    <li><a href="#ref-for-command-buffers">6.1. Overview</a>
-    <li><a href="#ref-for-command-buffers①">7.3.3. WebGPU Interoperability</a>
-    <li><a href="#ref-for-command-buffers②">7.9. MLCommandEncoder</a>
-    <li><a href="#ref-for-command-buffers③">7.9.3. Generate GPU Command Buffer</a> <a href="#ref-for-command-buffers④">(2)</a> <a href="#ref-for-command-buffers⑤">(3)</a>
+    <li><a href="#ref-for-command-buffers">4. Privacy Considerations</a>
+    <li><a href="#ref-for-command-buffers①">6.1. Overview</a>
+    <li><a href="#ref-for-command-buffers②">7.3.3. WebGPU Interoperability</a>
+    <li><a href="#ref-for-command-buffers③">7.9. MLCommandEncoder</a>
+    <li><a href="#ref-for-command-buffers④">7.9.3. Generate GPU Command Buffer</a> <a href="#ref-for-command-buffers⑤">(2)</a> <a href="#ref-for-command-buffers⑥">(3)</a>
    </ul>
   </aside>
   <aside class="dfn-panel" data-for="term-for-dictdef-gpucommandbufferdescriptor">
@@ -4401,10 +4405,11 @@ <h3 class="no-num no-ref heading settled" id="index-defined-here"><span class="c
   <aside class="dfn-panel" data-for="term-for-operationerror">
    <a href="https://webidl.spec.whatwg.org/#operationerror">https://webidl.spec.whatwg.org/#operationerror</a><b>Referenced in:</b>
    <ul>
-    <li><a href="#ref-for-operationerror">7.3.1. Synchronous Execution</a> <a href="#ref-for-operationerror①">(2)</a>
-    <li><a href="#ref-for-operationerror②">7.3.2. Asynchronous Execution</a> <a href="#ref-for-operationerror③">(2)</a>
-    <li><a href="#ref-for-operationerror④">7.3.3. WebGPU Interoperability</a>
-    <li><a href="#ref-for-operationerror⑤">7.9.2. Dispatch Execution Commands</a>
+    <li><a href="#ref-for-operationerror">4. Privacy Considerations</a>
+    <li><a href="#ref-for-operationerror①">7.3.1. Synchronous Execution</a> <a href="#ref-for-operationerror②">(2)</a>
+    <li><a href="#ref-for-operationerror③">7.3.2. Asynchronous Execution</a> <a href="#ref-for-operationerror④">(2)</a>
+    <li><a href="#ref-for-operationerror⑤">7.3.3. WebGPU Interoperability</a>
+    <li><a href="#ref-for-operationerror⑥">7.9.2. Dispatch Execution Commands</a>
    </ul>
   </aside>
   <aside class="dfn-panel" data-for="term-for-idl-promise">
@@ -5288,8 +5293,9 @@ <h2 class="no-num no-ref heading settled" id="issues-index"><span class="content
   <aside class="dfn-panel" data-for="device-type">
    <b><a href="#device-type">#device-type</a></b><b>Referenced in:</b>
    <ul>
-    <li><a href="#ref-for-device-type">6.2. Device Selection</a>
-    <li><a href="#ref-for-device-type①">7.3. MLContext</a> <a href="#ref-for-device-type②">(2)</a> <a href="#ref-for-device-type③">(3)</a>
+    <li><a href="#ref-for-device-type">4. Privacy Considerations</a> <a href="#ref-for-device-type①">(2)</a>
+    <li><a href="#ref-for-device-type②">6.2. Device Selection</a>
+    <li><a href="#ref-for-device-type③">7.3. MLContext</a> <a href="#ref-for-device-type④">(2)</a> <a href="#ref-for-device-type⑤">(3)</a>
    </ul>
   </aside>
   <aside class="dfn-panel" data-for="device-type-gpu">
@@ -5302,8 +5308,9 @@ <h2 class="no-num no-ref heading settled" id="issues-index"><span class="content
   <aside class="dfn-panel" data-for="power-preference">
    <b><a href="#power-preference">#power-preference</a></b><b>Referenced in:</b>
    <ul>
-    <li><a href="#ref-for-power-preference">6.2. Device Selection</a>
-    <li><a href="#ref-for-power-preference①">7.3. MLContext</a> <a href="#ref-for-power-preference②">(2)</a> <a href="#ref-for-power-preference③">(3)</a>
+    <li><a href="#ref-for-power-preference">4. Privacy Considerations</a>
+    <li><a href="#ref-for-power-preference①">6.2. Device Selection</a>
+    <li><a href="#ref-for-power-preference②">7.3. MLContext</a> <a href="#ref-for-power-preference③">(2)</a> <a href="#ref-for-power-preference④">(3)</a>
    </ul>
   </aside>
   <aside class="dfn-panel" data-for="power-preference-default">
@@ -5417,9 +5424,10 @@ <h2 class="no-num no-ref heading settled" id="issues-index"><span class="content
   <aside class="dfn-panel" data-for="enumdef-mloperandtype">
    <b><a href="#enumdef-mloperandtype">#enumdef-mloperandtype</a></b><b>Referenced in:</b>
    <ul>
-    <li><a href="#ref-for-enumdef-mloperandtype">7.4. MLOperandDescriptor</a>
-    <li><a href="#ref-for-enumdef-mloperandtype①">7.7. MLGraphBuilder</a>
-    <li><a href="#ref-for-enumdef-mloperandtype②">9.1. MLOperandType and ArrayBufferView compatibility</a> <a href="#ref-for-enumdef-mloperandtype③">(2)</a>
+    <li><a href="#ref-for-enumdef-mloperandtype">4. Privacy Considerations</a>
+    <li><a href="#ref-for-enumdef-mloperandtype①">7.4. MLOperandDescriptor</a>
+    <li><a href="#ref-for-enumdef-mloperandtype②">7.7. MLGraphBuilder</a>
+    <li><a href="#ref-for-enumdef-mloperandtype③">9.1. MLOperandType and ArrayBufferView compatibility</a> <a href="#ref-for-enumdef-mloperandtype④">(2)</a>
    </ul>
   </aside>
   <aside class="dfn-panel" data-for="dom-mloperandtype-float32">