The meeting focused on refining the QoE metrics document and its associated diagrams and metrics related to playback sessions and QoE. Tony and Maximiliano discussed issues with rendering Mermaid diagrams across different editors and platforms, and considered including the Mermaid code in the document for clarity and AI readability. The group reviewed and preferred one of Maximiliano’s diagram templates, and discussed whether to use singular or plural terms in the diagrams. They debated how to present complex equations, with Joan suggesting pseudocode instead of LaTeX images for better maintainability and accessibility. The team discussed definitions and measurement windows for startup metrics like start delay, start bitrate, start buffer length, and especially startup throughput, with Joan proposing to define it as the throughput between the start event and playback start or user quit. They also reviewed minor updates to an existing document, with Hali explaining the approach to versioning and revisions. The session ended with a brief update on the upcoming presentation for REN and logistical notes about remote access.
- Hali to send Tony the latest SVTA template for the QoE metrics document.
- Hali to double-check with Flor about the Error Codes document edits.
- Hali to share the Zoom links for the Rennes session (once platform is confirmed).
- Maximiliano help populate the template with the mermaid code and improve the diagram, especially the border issue.
- Tony to update the document to include the mermaid code in an annex for AI agents to read.
- Tony to redo the V1.1 document changes with track changes for clarity.
- Antitrust Statement and WG Discussion
Mermaid Diagram Tool Issues
- The team discussed issues with using Mermaid for creating diagrams, where Maximiliano noted that different editors (mermaid.live and mermaid.ai) produced different results even with identical code.
- Tony expressed confusion about the Mermaid interface and shared his experience trying to align elements, though he eventually lost the proper alignment.
- The discussion highlighted inconsistencies between different Mermaid platforms and editors, with both participants acknowledging they were not experts on the tool.
Document Formatting and Proposal
- Tony discussed challenges with formatting in a document, including issues with HTML markup not displaying properly.
- He mentioned Slavik’s request to transform their work into a proper proposal with a title page using the SVTA template.
- Tony also addressed a question about changing diagram terminology from "playback session" and "assets" to "playback sessions" and "assets" with an S, but he decided against this change, preferring to keep them as objects rather than instantiations.
Document Formatting and Metrics Discussion
- Maximiliano and Tony discussed document formatting, with Maximiliano suggesting to use singular versions of words and include Mermaid code in an annex rather than as images to help AI agents better understand the diagrams.
- Tony proposed changing metric: “Exits Before Playback Starts” from a count to a ratio to be more actionable.
- Joan asked if initial buffering is excluded from the buffering ratio and Tony confirmed it was. Buffering ratio only includes ‘preventable’ buffering, omitting user initiated buffering.
- Tony discussed creating documentation including a state diagram, usage guidelines, and equations, with plans to have Max help populate the template once completed.
- Tony shared his use of LaTeX to format equations but questioned whether this approach was too technical for the team.
- Joan suggested pseudocode as an alternative, noting concerns about the image format of LaTeX equations not being easily copyable or searchable.
Document Template Issues Discussion
- The team discussed issues with a document template where Tony encountered problems with green borders and copy-paste functionality that wouldn't work properly when shared.
- The group reviewed template options, with Rob indicating he preferred the version with helpful blocks after seeing Tony's version again.
Design and Measurement Timeframe Decisions
- The team discussed design decisions for visual elements, with Joan agreeing with Rob's suggestion about grouping events as a series.
- They then focused on defining measurement timeframes for bandwidth and throughput, with Tony introducing concepts like start bitrate, start buffer length, and start throughput.
- The group debated whether to use fixed time periods or variable segment lengths for measurements, with Maximiliano suggesting 30 seconds and Joan questioning if one minute was appropriate given video lengths.
- The discussion ended with the team still deciding between 10, 20, or 30 seconds for the measurement timeframe.
Startup Throughput and Video Performance Metrics Discussion
- Joan and Tony discussed defining startup throughput metrics to better understand why start delays occur.
- Joan proposed measuring throughput specifically between the start and playback start time rather than over a fixed period, as this would capture network conditions during the critical startup phase.
- Tony agreed this approach would be more effective than measuring over a set time window, as it would account for variable conditions like going through tunnels during startup.
- The team discussed metrics for measuring video playback performance, focusing on start throughput calculations and how to handle sessions that end before video starts.
- They agreed to exclude sessions that end via idle timeout or are abandoned from throughput measurements.
Error Codes Document Edits
- Tony presented updates and minor corrections to the Error Codes document.
- Hali advised that since they are minor edits, they do not need to publish a new revision, noting that she will confirm this with Flor.
- The team also confirmed plans for an upcoming presentation at Rennes next week, with Hali agreeing to share the remote access details once confirmed.
Next Meeting: Q4 2026 Member Meeting in Rennes, France – October 7-8, 2026 - See Online Agenda for WG Session Times