Session Date/Time: 11 Nov 2021 16:00

panrg

Summary

The panrg session covered administrative updates, including a brief IAB review of the research group's activities. Key discussions revolved around the status of the pan-rg-questions draft, the evolution of the vocabulary of path properties draft, and a presentation on describing satellite networks using panrg terminology. A significant portion of the session was dedicated to an analysis of multi-path Quick (MPQUIC) against the impediments to path-aware networking deployment (RFC 9049), which sparked a broader conversation about the future of multi-path in the IETF and the role of panrg.

Key Discussion Points

Administrative Updates
- The IETF/IRTF Notewell was presented, emphasizing IRTF follows IETF intellectual property rules.
- Meeting logistics (audio/video off unless speaking, headphones recommended, VPN issues, pre-loaded slides).
- Corey was thanked for serving as minutes taker.
- The agenda included presentations on vocabulary, satellite networks, multi-path selection, followed by 40 minutes of open mic time.
pan-rg-questions draft Status
- The pan-rg-questions draft is awaiting its document shepherd and final feedback from Gauri and GitHub.
- Brian Trammell committed to rolling a new draft version during the open mic session.
IAB Review of panrg
- The IAB conducted a review of panrg in September/October.
- Next Steps identified for panrg:
  - Continue answering the research questions in the pan-rg-questions draft. The path properties draft is a good answer to the first question.
  - Consider research applications of Path Aware Networking (PAN) properties to IETF protocols (e.g., Spencer's multi-path work).
  - Reached out to the ALTO Research Group to invite them to panrg for related work.
  - panrg should operate as a general venue for discussions at the intersection of routing and transport.
Vocabulary of Path Properties (Rees Kressin)
- Draft Updates:
  - Revised 'entity' definition to be more specific: entities play a role related to pathway networking for particular paths and flows (e.g., data plane for forwarding, control plane for influencing forwarding).
  - Formalized 'transparency' definition as a function f(flow F, meta-information M) = action A, where the node is transparent if A is constant regardless of M.
- Discussion on Defining Terms:
  - The draft aims to define key terms used in other panrg documents (e.g., "routing domain identifier" from the questions draft).
  - Spencer Dawkins requested adding a pointer to the GitHub repository in the draft.
  - Discussion on the scope of definitions: how deep to go? Brian Trammell suggested stopping when existing IETF definitions can be referenced, and for core terms, a "research group last call" approach.
  - Sabine D'Hondt suggested context-specific definitions (e.g., "endpoint" in routing vs. transport). Rees clarified the document aims for general definitions within the panrg scope.
  - Gauri emphasized defining "endpoint" and "end-to-end".
- Feedback Mechanism: GitHub issues are preferred for specific definition feedback; mailing list for general scope discussions.
Satellite Networks and Path Properties (Nikola):
- Background: Historical TCP tuning for GEO satellites, challenges with Quick over satellite, increasing GEO/LEO capacities, need for satellite-aware transport.
- Application of panrg Terminology: The authors attempted to describe generic satellite communication (satcom) systems using the vocabulary of path properties terms (host, node, path properties).
- Proposed Next Steps:
  - Rees Kressin suggested mapping the satcom problem space to the panrg research questions.
  - Brian Trammell expressed interest in a deeper exploration of satcom in panrg, potentially as an "architecture" document illustrating how the internet adapts to this technology.
  - Gauri offered to help split the existing TSVWG document (which was very transport-centric) into two: one defining satellite systems and their properties/challenges for panrg, and another for transport fixes for TSVWG. This was positively received.
Multipath Selection Strategies (Spencer Dawkins)
- Observation: Multi-path is becoming ubiquitous, common, deployable, and ambitious (SCTP, MPTCP, DCCP, and now MPQUIC).
- MPQUIC Analysis Against RFC 9049 (Impediments to PAN Deployment): Spencer assessed the proposed multi-path Quick extensions against the 13 lessons learned from RFC 9049.
  - Positive Outlook: MPQUIC performs well against most impediments, especially as extensions are endpoint-to-endpoint and encrypted, negating concerns about network operator involvement or trust of midpoints (which were major showstoppers for older PAN technologies).
  - Remaining Challenges/Considerations:
    - Justifying deployment depends on the path selection strategy (e.g., bandwidth aggregation).
    - "High quality" alternative paths are crucial, and "quality" is application-dependent (Gauri).
    - Support in endpoint stacks depends on advanced scheduling, which is currently out of scope for the initial MPQUIC draft.
    - Planning for failure requires knowing when to discard unhelpful paths.
- Discussion:
  - Med pointed out that benefits for early adopters have changed post-QUIC, with major players driving initial deployments.
  - Cyril highlighted that partial deployment benefits heavily rely on the existence of different high-quality alternative paths.
  - Maria agreed, pointing to the IAB workshop on measuring user experience report as relevant.
  - Tanji asked about the different strategies for multi-path (endpoint-based like MPQUIC vs. midpoint-proxy like ATSS/5G UPF). Spencer noted this is a key question for panrg.
  - Brian Trammell sees MPQUIC as an exciting "actuation side" testing ground for path awareness, even before full sensing mechanisms.
  - Chris pointed out the complexity of multi-path on phones (Wi-Fi + cellular), where different entities make steering decisions, potentially leading to competition.
Open Mic / General Discussion
- The conversation continued on multi-path, its connection to path awareness, and where various research aspects fit within the IETF (panrg, iccrg, or other groups).
- Zohar and Colin acknowledged the overlap across groups and generally favored broad groups over hyper-specialized ones.
- Sabine D'Hondt mentioned an ALTO WG draft under IESG review that conveys abstracted path information between endpoints, which could be relevant to panrg's discussions on path properties and application performance.

Decisions and Action Items

pan-rg-questions Draft: Brian Trammell rolled a new version (pan-rg-questions-11) during the meeting to incorporate feedback from Gauri and GitHub.
Vocabulary of Path Properties Draft:
- Rees Kressin will add a pointer to the GitHub repository in the draft.
- Community to provide specific definition feedback via GitHub issues; general scope discussions on the mailing list.
Satellite Networks Draft: Nikola and co-authors will revise the draft to focus on defining satellite systems and discussing panrg research questions, potentially splitting transport-specific content into a separate document or WG. Gauri offered direct assistance with this.
Multi-path Discussion Follow-up:
- Spencer Dawkins will send links to relevant ATSS/multipath-quick interim drafts to the panrg mailing list.
- Sabine D'Hondt will send a link to the ALTO working group draft on conveying abstracted path information to the panrg mailing list and consider presenting it at a future meeting.
- Chairs of panrg, iccrg, and IETF ADs will discuss appropriate venues for various multi-path research questions.

Next Steps

Community Engagement: Provide feedback on the vocabulary of path properties draft, particularly for core terms (entity, node, host, link, path element, path), via GitHub issues before the next meeting.
Satellite Networks Document: Nikola and co-authors to develop the panrg-focused version of the satellite networks document, leveraging panrg terminology and addressing research questions.
Multi-path Research: Continue to explore the implications of multi-path Quick and other multi-path technologies, particularly regarding advanced scheduling, path selection strategies, and their interaction with path awareness concepts.
Guiding Principles: Continue to follow the IAB review recommendations: answer panrg questions, consider research applications of PAN to IETF protocols, and operate as a venue for routing and transport intersections.