ICNRG

Summary

The ICNRG (Information-Centric Networking Research Group) met at IETF 126. The session was co-chaired by Dave Oran and Ryo Yanagida. The meeting featured updates on active research group drafts, followed by six technical research presentations covering ICN integration with SCION, distributed AI computation, routing scalability, NDN access control, and cross-domain compute discovery.

Key Discussion Points

1. RG Status and Administration

CCNX Chunking Draft (draft-ietf-icnrg-ccnxchunking): Has completed last call, underwent IRSG review, and the authors have submitted a revised version addressing the feedback. It is currently waiting in the administrative queue to confirm that the IRSG reviewer is satisfied.
Versioning Draft (draft-ietf-icnrg-versioning): Remains an active research group draft. The chairs requested more feedback before progressing it to a last call.
FLIC Manifest Draft: An updated version is expected to be published for review prior to the IETF 127 meeting in November.

2. ICN+SCION: A Global Name-Based Network Service

Presenter: Jeremiah Davis
Slides: ICN+SCION: A global name-based network service
Discussion:
- The project connects NDN/CCNX autonomous systems (ASes) via the SCION secure inter-domain architecture. It incorporates the RHINE name service (a secure DNS/DNSSEC replacement) and Reflexive Forwarding to manage inter-domain routing states.
- Tillmann Wegmann asked about routing directly to clients versus using a centralized gateway. Jeremiah Davis clarified that the gateway architecture accommodates applications unaware of SCION and enables broader caching of name prefixes.
- Dirk Kutscher inquired about performance measurements; Jeremiah Davis noted that performance evaluations are scheduled as upcoming summer work.
- In the chat, David Oran questioned if consumers need to register with the mapping system, which would impact anonymity. Junxiao Shi and Tianyuan Yu clarified that only producers need to register. Kenneth Calvert and Tianyuan Yu discussed the need for on-demand mechanisms for routers to dynamically write popular prefixes into Prefix Egress Tables (PET).

3. ICN-Enabled Adaptive In-Network Computation for Resilient Distributed AI

Presenter: Tatet
Slides: ICN-Enabled Adaptive In-Network Computation for Resilient Distributed AI
Discussion:
- This work explores stateful, in-network computation for distributed AI (e.g., vehicular object tracking) utilizing a Request-to-Compute (RDC) model and contextual state management. It leverages versioning to track intermediate results and support seamless failover.
- Dave Oran asked where computation result caching is most beneficial. Tatet responded that in scenarios like autonomous driving, multiple vehicles approaching the same intersection can reuse cached object detection results (e.g., pedestrians, traffic lights) from nearby roadside units.
- Shelley/Tianxiang Li asked about node selection under dynamic resource availability. Tatet explained that compute node eligibility is dynamically evaluated based on resource availability and capacity scores.

4. From Map-and-Encap to BIER: Observations on Network Routing Scalability

Presenter: Lixia Zhang
Slides: From Map-and-Encap to BIER: Observations on Network Routing Scalability
Discussion:
- Lixia Zhang provided a historical review of routing scalability. The core issue is the conflict between provider-allocated topological addresses and stable end-system identifiers. Map-and-encap solutions (e.g., LISP, MPLS, BIER) resolve this conflict by separating the logical and topological namespaces.
- Ken Calvert raised SCION as an implementation of map-and-encap that routes to an AS based on AS numbers.
- Roland Bless highlighted KIRA in the chat as a scalable routing scheme that operates without requiring a centralized ID-to-locator mapping system. Dave Oran noted that hyperbolic routing is another scalable approach that trades off path stretch for reduced state.
- Lixia Zhang noted that the deployment of mapping systems remains a critical barrier for post-facto IP integration, whereas NDN integrates mapping directly into its routing architecture.

5. Architecting A Scalable Routing System for Named Data Networking

Presenter: Tianxiang Li
Slides: Architecting A Scalable Routing System for Named Data Networking
Discussion:
- This presentation detailed a map-and-encap routing architecture designed for NDN. Edge routers maintain a Prefix Egress Table (PET) synchronized via a Prefix State Database (PSD) application, while core routers forward interests based purely on topological router IDs. The architecture also supports stateless multicast via BIER bit strings.
- Dirk Kutscher raised questions regarding the scalability and performance of edge router synchronization using protocols like SVS. Tianxiang Li and Lixia Zhang responded that separating topological updates from application prefix updates significantly reduces BGP-style routing churn and packet overhead.
- Ken Calvert clarified that sync interests are only propagated for mapping table updates, not for general application sync.

6. Open-Source Decentralized Access Control Library for NDN

Presenter: Farhat
Slides: Open-source decentralized access control library for NDN
Discussion:
- Farhat presented a decentralized, open-source security library that sits between the application and network layers. It utilizes multi-authority Attribute-Based Encryption (ABE) and features an optimized immediate revocation mechanism.
- Tatet asked about the overhead on resource-constrained devices; Farhat confirmed that the library has been benchmarked on several hardware platforms, including Raspberry Pi.
- Dave Oran asked how the library handles large objects, noting that signing or encrypting every individual packet is costly. Farhat agreed that utilizing manifests to manage key-to-content mappings is the appropriate approach.
- Tianxiang Li and Lixia Zhang shared that the latest release of "Ownly" (an NDN collaboration application) has adopted the IETF Message Layer Security (MLS) standard (RFC 9420) to handle dynamic group key rotations efficiently.
- In the chat, Junxiao Shi warned about the lack of well-maintained ABE/ABAC libraries, citing incompatibilities between NAC-ABE (OpenABE) and OpenSSL 3.

7. Cross-Domain Compute Discovery

Presenter: Shelley (on behalf of Liu)
Slides: ietf126-icnrg-draft-liu-icnrg-cross-domain-compute-discovery-00-slides
Draft: draft-liu-icnrg-cross-domain-compute-discovery-00
Discussion:
- The presentation outlined the requirements for discovering compute services across heterogeneous space, terrestrial, and maritime domains. It emphasized separating stable service names from temporary execution instances, ensuring result validity, and verifying cross-domain trust.
- Due to time constraints, Dave Oran directed participants to continue the technical discussion on the mailing list.

Decisions and Action Items

Flick Draft: Authors committed to issuing an updated draft before the IETF 127 meeting in November.
Interim Meeting Planning: The chairs will poll the mailing list regarding holding an online interim meeting in mid-to-late September to address postponed topics and advance active drafts (including FLIC and the versioning draft).
NDN Updates: Lixia Zhang committed to posting NDN project updates and coordination announcements more regularly to the ICNRG and ndn-interest mailing lists to improve community synchronization.

Next Steps

CCNX Chunking: Move to final publication once the outstanding IRSG administrative review comments are resolved.
Versioning Draft: Solicit more community reviews on the mailing list to prepare draft-ietf-icnrg-versioning for a future last call.
Compute Discovery: Encourage discussion on the mailing list regarding the requirements outlined in draft-liu-icnrg-cross-domain-compute-discovery-00.

Automatic IETF Minutes