IETF Last Call Review of draft-ietf-cats-usecases-requirements-10
review-ietf-cats-usecases-requirements-10-rtgdir-lc-dunbar-2025-12-13-00

Summary: This document provides a comprehensive and well-written collection of
scenarios that illustrate the motivation for Computing-Aware Traffic Steering
(CATS). The breadth of use cases (AR/VR, transportation, digital twin, AI
training/inference) clearly demonstrates the importance of jointly considering
network and compute resources when steering traffic, and the document does a
good job of motivating the general CATS problem space.

However, there are a few areas where the narrative alignment between examples
and CATS-specific mechanisms could be improved. Addressing the comments below
would, in my opinion, significantly strengthen the document and make the CATS
applicability clearer.

Major:
- The introduction opens with a discussion of CDN. It is not clear how CDN
directly relates to the CATS problem as developed in the main body of the
document. In particular, CDN is not explicitly used as a running example in the
CATS use cases, nor does it appear to drive any unique requirements beyond
those already captured by general service-instance selection.

Suggestion: the CDN discussion might fit more naturally in an appendix (as is
partially done today), allowing the introduction to stay tightly focused on
CATS itself.

- Section 4.4 introduces a computing-aware SD-WAN scenario. While the example
itself is clear, it is not obvious how this scenario derives CATS-specific
requirements. In particular, SD-WAN policy-based application steering or
centralized controller-based optimization does not necessarily require CATS
mechanisms.

Suggestion: it would be helpful to explicitly state:
- which aspects of this scenario cannot be addressed by existing SD-WAN
mechanisms alone, and - Whether the intent is to present SD-WAN as a deployment
environment that can benefit from CATS, or as a motivating example for
CATS-specific metric distribution and decision logic.

Warm Regards, Linda