Mar 11, 2026 | RUMCG Meeting

Agenda

• Interop 2026 updates (Nic)
• Unused Preload RUM (Yoav)
• Core Web Vitals (CWV) across the browsers (All)

Admin

• Next meeting: April 8, 2026

General Summary

• Interop 2026 Progress: Limited inclusion of key web performance features like Core Web Vitals impacts standardization efforts.
• Browser Vendor Influence: Feature exclusions relate to varying priorities among vendors, slowing web performance advancements via Interop.
• Unused Preloads Proposal: New RUM data proposal seeks to clarify unused preload impacts, aiding resource optimization for developers.
• Core Web Vitals Insights: Data shows discrepancies across browsers, especially in Safari, with ongoing bug fixes expected to improve reliability.
• Community Collaboration Plans: Future discussions will focus on data sharing, server timing, and bot detection to enhance web performance efforts.
• Next Meeting Topics: Anticipated subjects include bot detection, AI impacts on RUM data, and further analysis of Core Web Vitals.

Action items

Yoav Weiss

• Refresh and update Chrome patches implementing unused preload and speculation reporting proposal; open a feedback issue in the new Web Performance repo to collect community input (20:26)

Community members (including Karlijn Lowik, Barry Pollard)

• Provide developer feedback and support for the unused preload/speculation proposal via the planned issue and Slack channels (24:47)

Barry Pollard and Yoav Weiss

• Add frequently asked questions (FAQs) to the proposal documentation regarding early hints, DNS prefetch, and preconnect exclusions and potential future support (28:27)

Nic Jansma

• Share RUM Archive data and analysis publicly; blog about data insights and methods for community access (55:33)

Cliff Crocker

• Produce detailed blog posts on Core Web Vital data findings and server timing header adoption to stimulate community discussion (55:34)

Community members

• Prepare for next meeting’s discussion on bot detection techniques impacting Core Web Vitals data quality and bot exclusion strategies (54:46)

Notes

Interop 2026 Updates

• Interop 2026 is progressing but excludes some key web performance features important to this community.
• Core Vitals Progress Limited for 2026 (03:49)

• Interop focuses on aligning browser features across vendors through tests and implementations.
• Core Web Vitals inclusion last year drove significant progress, prompting today's discussion.
• Proposals to include CLS and scheduler APIs for 2026 were rejected, slowing new standardization on these fronts.
• Logan Frames also missed inclusion, indicating ongoing work is needed to push that forward.

• Browser Vendor Priorities Affect Feature Inclusion (06:13)

• Barry Pollard explained some features were excluded because not all browser vendors prioritize them equally.
• Despite exclusion in Interop, individual browsers like WebKit are working on related features like speculation rules.
• Vendors such as Mozilla have shown some interest, leaving room for these features to land later with or without Interop.
• This reality tempers expectations for big strides in web performance features via Interop in 2026.

• Community Monitoring of Browser Positions (07:32)

• Cliff Crocker stressed the importance of tracking browser vendor positions on web performance features regularly.
• Keeping updated records helps coordinate efforts and funding for missing features, especially those not in Interop.
• The group discussed the need to refresh and maintain position statements, such as WebKit’s stance on CLS.
• This ongoing monitoring supports strategic planning and advocacy for desired web standards.

• Safari's Hesitancy on CLS and Related Features (08:20)

• Barry noted Safari has been hesitant on CLS due to measurement approach differences and cost concerns.
• Some WebKit contributors remain open, but priority remains low and doubts about necessity persist.
• Community contributions and real-world use cases could influence Safari’s eventual adoption.
• A request for a formal WebKit standards position on CLS is pending, highlighting uncertainty.

Unused Preloads and Speculation Reporting Proposal

• A new proposal aims to provide accurate RUM data on unused preloads and navigation speculations to optimize resource usage.
• Problem and Proposal Overview (10:11)

• Yoav Weiss described the issue of many unused preloads triggering console warnings but lacking reliable RUM reporting.
• Current heuristics are unreliable and premature, causing confusion about resource usage and performance impact.
• The proposal intends to report used versus unused preloads and navigation speculations only when the page unloads.
• This approach helps developers measure the cost-benefit tradeoff of preloading and prefetching resources.

• API Design Using Page Hide Event (15:01)

• The proposal adds a speculations attribute to the page hide event to report preload usage and navigation speculation details.
• Reporting at page hide waits until the last moment, avoiding misleading early data that could cause wrong conclusions.
• It includes metadata such as resource attributes and speculation eagerness to guide optimization.
• Ongoing discussions focus on edge cases like cache expiration and cross-process reporting complexity.

• Implementation Status and Community Support (20:22)

• Chrome patches implementing this exist but are outdated and not yet landed, awaiting refreshed feedback.
• Yoav and Barry welcomed developer feedback to prioritize and push the feature forward.
• The group expressed strong interest and willingness to support the feature’s adoption.
• A new GitHub issue for feedback will be opened to centralize community input.

• Scope Clarifications and Future Extensions (28:01)

• The current focus excludes cheaper speculation types like preconnect and DNS prefetch but may add them later if successful.
• Early hints usage in link headers might be added to distinguish preload triggers in RUM data.
• Discussion emphasized balancing feature complexity with usefulness and avoiding unnecessary scope creep.
• The proposal avoids tying reporting to other APIs like Fetch Later to maintain flexibility.

Core Web Vitals Data Insights

• Multiple RUM data sets reveal browser discrepancies, bugs, and evolving support impacting Core Web Vitals interpretation.
• Safari Data Anomalies and Bug Reporting (33:38)

• Erwin Hofman highlighted inflated INP numbers and abnormally low processing times in Safari data, traced to bugs.
• These issues mainly affect back/forward cache navigations and presentation delay metrics.
• Bug reports and fixes are in progress, expected in coming Safari releases.
• The current Safari INP data is unreliable for meaningful comparison until fixed.

• Usefulness of INP Subparts for Diagnostics (38:15)

• Barry Pollard questioned how valuable INP subparts are on aggregate versus individual interactions.
• Erwin explained subparts help narrow down where to look in DevTools, such as identifying long-running scripts or event listeners.
• The group noted INP subparts are messier than LCP subparts due to interaction variability but still useful for troubleshooting.
• Further discussion on INP metric utility and inclusion in public data sets is ongoing.

• Cross-Browser LCP and INP Comparisons (41:38)

• Cliff Crocker shared data showing Mac OS LCP trends with Safari performing slightly better on mobile than Chrome.
• INP data showed large deltas between Safari and Chrome, attributed to the known Safari bugs.
• Nic Jansma presented Impulse data showing traffic distribution with 23% Chrome desktop, 13% Chrome mobile, 20% Safari mobile, and 2.5% desktop Safari (47:47).
• LCP support rollout dates correlate with improved data quality and narrowing performance variability, especially for Safari starting late 2023.

• Influence of Bot Traffic and Data Categorization (53:38)

• Nic raised concerns about bot traffic inflating variability in performance metrics and misclassified user agents.
• The group agreed to hold a future discussion on bot detection and exclusion strategies to improve data accuracy.
• Collaboration between companies with bot management products was encouraged to share best practices.
• This effort aims to refine RUM data quality across browsers and platforms.

Future Topics and Community Collaboration

• The group plans to continue exploring performance data, server timing, and bot detection with collaborative data sharing.
• Server Timing Investigation Delayed (55:32)

• Due to time constraints, server timing header adoption research will be deferred to the next meeting.
• Cliff Crocker plans to publish a blog post with findings to stimulate discussion.
• This topic ties into broader performance monitoring and optimization efforts.

• Data Sharing and Transparency Encouraged (55:55)

• Nic Jansma emphasized the value of sharing competing RUM data sets to provide a holistic industry perspective.
• The shared Collab data and upcoming blog posts aim to empower others to explore similar queries independently.
• This open approach fosters transparency and collective advancement across web performance communities.

• Community Engagement to Support Proposals (24:08)

• Karlijn Lowik encouraged the group to actively support the unused preload proposal through feedback and issue commenting.
• Barry Pollard and Yoav Weiss committed to creating centralized feedback channels for easier community involvement.
• Such engagement is critical to accelerate adoption and refine proposals based on real-world needs.

• Next Meeting Focus Areas (02:01)

• Potential future topics include bot and bad actor detection, AI browser impacts on RUM data, and further Core Web Vitals analysis.
• Volunteers to lead these discussions are welcomed to shape the agenda.

• This ensures continued relevance and responsiveness to community priorities.