Agenda for CDNI, IETF 124, Montreal, Canada Chairs: Chris Lemmons, Kevin J. Ma, Sanjay Mishra IETF 124 Agenda: datatracker.ietf.org/meeting/124/agenda Charter: http://datatracker.ietf.org/wg/cdni/charter/ Mailing list: cdni@ietf.org Monday, 3rd November 2025, 17:00-19:00 Eastern (Montreal) Time , Room: Notre Dame Chair Slides: - The new charter was adopted. The new charter is simpler, accurately reflects the group's current work, removes references to completed work, and explicitly includes work on MEL and Processing Stages. - Milestones were reviewed and updates noted; to be entered by the chairs post-meeting. - Trigers version 18 submitted. Needs re-review by Kevin and Nir. - Kevin suggested aiming for December if only minor nits. - Alan suggestedFebruary might be more realistic if major changes are required. - Named Footprints needs to be pushed out to June 2027 - Alan: It is a massive undertaking and requires more thought on the approach - Chris: reminder that we might need to split the doc - Cache control depends on MEL so needs to be pushed out after MEL. Guillaume Bishot volunteered to shepherd. - MEL and Processing Stages targeted for June 2026. - Source Access Control is a "monster" draft. Pushed to August 2026, after MEL and Processing Stages. - Client Access Control also pushed to August 2026, to focus on MEL and Processing Stages. - Rajeev agreed that protected secrects and client access control could follow shortly after MEL and Processing Stages. - Delivery Metadata targeted for March 2026. - Gorry (as AD): Requests that drafts be spaced out and try to limit them to one per month for his review. - Chairs noted IANA early feedback and encouraged Triggers and Named Footprints authors to incorporate comments. Discuss Updates on In-Progress Work draft-ietf-cdni-delivery-metadata-00 (Guillaume Bichot) - Guillaume: In response to comments/feedback from the last IETF, simplified the draft, focusing on: traffic type, capacity, and transport (uni/multi cast) - Use case: If the uCDN knows a client doesnt support HTTP redirect, it can force the dCDN not to use HTTP redirect. - Chris [chat]: maybe think carefully about the name "CdnSelection". If you're really just selecting transports, the name might be causing confusing. - Alan: It's foundational, but needs better definition - Sounds like CDN virtualization, which has additional use cases: - geographical separation (e.g., scoping with footprints) - orchestration: staging vs prod, could be same footprint but different configurations - dCDN CDN selection sounds strange vs a dCDN advertising and a uCDN selecting - Alan: Traffic type needs a formalized definition of types, so that there is no ambiguity - The high level definitions are too "hand wavy". Doesn't work without out-of-band agreed upon labels. - Ben [chat]: Could you make it a registry and then other specs can add additional concrete definitions of types? - Chris [chat]: smells like a registry to me. Registries are inexpensive, ossification is very expensive. - Rajeev: CDN transport type feels too broad with just multicast and unitcast. Doesn't specify protocol. - Guillaume: We want the complexity of protocol, only the type - Rajeev: How does the uCDN choose between different protocols? - Guillaume: Could use the "hints" property - Chris: Please continue the disucssion on the list. - Glenn: The original intent for traffic types was high level concepts: live vs vod, with hints, that would seed defaults - with Level 3 found useful. - Rajeev [chat]: +1 to ben's suggestion of using a registry draft-ietf-cdni-ci-triggers-rfc8007bis-18 (Alan Arolovitch) - Addressed Kevin's comments: - Clarified RFC vs Spec required and mandatory vs optional properties - Refactored Trigger collections: new root level object with per state and per label collections - Removed redundant "Urgent" support - Added formal decision tree for trigger processing - Rajeev: Pulling trigger content vs just IDs could be very large - Alan: It's optional - Sanjay: Please review the draft and send comments to the list draft-ietf-cdni-named-footprints-02 (Alan Arolovitch) - Alan: RFCs 8006 and 8008 use footprints "inline", where RFC 9241 establishes canonical footprints - Want to establish canonical references that other CDNI interfaces can use, not using an external protocol like ALTO. - CDNI Objects should be able to use either inline or canonical footprints draft-ietf-cdni-logging-extensions-02 (Ben Rosenblum) - Ben: Updated draft has not been published yet but has significant changes for custom fields and sampling - Custom fields: dCDN and advertise additional fields for inclusion in any logging records, including per-transport. - Custom record type: dCDN can fully specify the record in FCI or MI. - Sampling: Percentage-based, MEL rule-based, Session-based, or Header/Query-param-based draft-ietf-cdni-protected-secrets-metadata-05 (Ben Rosenblum) - Ben: Draft not published yet, but removing HashiCorp reference and defining a registry for external store types. draft-ietf-cdni-metadata-expression-language-00 draft-ietf-cdni-processing-stages-metadata-00 - Glenn: Processing Stages: Four different points at which the request/response can be modified - Glenn: MEL: request/response processing/transformation draft-ietf-cdni-source-access-control-metadata-00 - Glenn: Adds Source Metadata Extended and Source Extended objects to support multiple origins for failover, load balancing, and detention. - Also added support for HTTP/2 and HTTP/3 CDNI Interim - Sanjay: With 10+ documents, it would be useful to have more regular meetings to make progress. - Rajeev: Great Idea. Hard to schedule in November/December - Chris: Target mid-January, post holidays. - Gorry (as AD): Advertise interim ASAP. It's ok to cancel them. Best to have them on the IETF calendar months in advance. Leave 2-3 week buffer around IETF meeting. - Chris: Doodle polls will be posted to the list.