From 93467d73545ae454ae404b9bbfeb6b3549aae0b4 Mon Sep 17 00:00:00 2001
From: Markus Handell <handellm@google.com>
Date: Fri, 20 Sep 2024 17:19:43 +0200
Subject: [PATCH] add WebRTC-specific interactions with capture/receive/RTP
 timestamps

---
 index.html | 40 ++++++++++++++++++++++++++++++++++++++--
 1 file changed, 38 insertions(+), 2 deletions(-)

diff --git a/index.html b/index.html
index 7443dbe..4e33291 100644
--- a/index.html
+++ b/index.html
@@ -22,7 +22,7 @@
       // Add additional editors here.
       // https://github.com/w3c/respec/wiki/editors
                ],
-    "xref": ["webrtc", "hr-time", "mediacapture-streams", "webrtc-stats"],
+    "xref": ["webrtc", "hr-time", "mediacapture-streams", "webrtc-stats", "mediacapture-extensions"],
     "shortName": "webrtc-extensions",
     "specStatus": "ED",
     "subjectPrefix": "[webrtc-extensions]"
@@ -920,6 +920,42 @@ <h3>
                 </div>
               </dd>
             </dl>
-</section>
+  </section>
+  <section class="informative">
+    <h2>Timestamp behavior</h2>
+    <h2>RTCRtpSender timestamp effects on packet NTP and RTP timestamps</h2>
+    <p>
+      The user agent defines a <dfn class="export">frame timestamp</dfn> being the same as the [= capture timestamp =] of the frame being sent on the {{RTCRtpSender}} {{MediaStreamTrack}} if it is set.
+      If it is unset the user agent estimates a timestamp from the sent frame's [= presentation timestamp =] together with the
+      time it was received by the {{RTCRtpSender}}.
+    </p>
+    <p>
+      The NTP and RTP timestamps of the encoded frame corresponding to the frame being sent on the {{RTCRtpSender}} {{MediaStreamTrack}} are sourced from the [= frame timestamp =].
+    </p>
+
+    <h2>RTCRtpReceiver timestamps</h2>
+    <h3>Remote capture timestamp</h3>
+    <p>
+      <!-- TODO: How this is estimated is under specified and should be fixed, but the section holds the requestVideoFrameCallback text for now -->
+      For a frame produced in a {{RTCRtpReceiver}} track, the user agent computes a
+      <dfn class="export">remote capture timestamp</dfn>. It is a best-effort estimate of the local capture
+      time on the sender translated to the receiver clock, and can use methods like using RTCP SR
+      as specified in [[?RFC3550]] Section 6.4.1, or by other alternative means if use by RTCP SR
+      isn't feasible.
+    </p><p>
+      Each frame's [= capture timestamp =] is set to the [= remote capture timestamp =], if available.
+    </p>
+    <h3>Received RTP timestamp</h3>
+    <p>
+      For a frame produced in a {{RTCRtpReceiver}} track, the frame's [=RTP timestamp=] is set
+      from the RTP timestamp of constituent packets of the corresponding received encoded frame.
+    </p>
+    <h3>Receive timestamp</h3>
+    <p>
+      For a frame produced in a {{RTCRtpReceiver}} track, the [=receive timestamp=] is set
+      as the time the corresponding encoded frame was received by the platform, i.e. the time at which the
+      last packet belonging to this frame was received over the network.
+    </p>
+  </section>
 </body>
 </html>