[zzylol]

Why

The first review of the three-pipeline demo asked for the trade-off, not only the sketch's cost. Computing in front of the database stores much less but drops the raw data; storing everything and querying later keeps it but costs storage, query CPU and freshness. RFC 0001 did not define freshness, did not count the log store's cost, and had no scaling or playground plan.

What

RFC 0001 gets a "Review feedback, 2026-10-19" section, marked as an update after approval, plus three inline update markers:

• The three approaches the blog compares (sketch, exact, store and query) and the axes it reports a winner on.
• Definitions: metric data points (not "signals"), freshness, store-and-query freshness (ingest plus query latency), end-to-end detection latency.
• Baseline and metric changes: a one-stage raw pipeline plus a periodic query job, whole-system CPU and memory including the log store, bytes per hop, a traffic ramp, and a visible exact-merge panel.
• Experiments 7 to 9: whole-system cost, freshness against store and query, and traffic scaling with sampling stated where used.
• A benchmark playground plan, answers to the review's questions (including how sketches are carried), and new open questions.

It also replaces a person's name in the throughput success criterion with "the OTAP benchmark framework".

How

Document only. Accepted RFCs are not rewritten (docs/rfcs/README.md), so the original text stays and the changes are additions marked as updates.

Before

• No freshness definition.
• The raw pipeline is treated as a forwarding chain, not a store-and-query baseline.
• Cost covers pipeline stages only.

After

• Freshness and detection latency are defined per approach.
• Store and query is a one-stage baseline whose answer comes from a measured query.
• Cost includes every container needed to get the answer.
• Scaling and playground work is planned.

Verification

• Rendered the RFC to PDF and checked the new section and images.
• Checked the transport answer against asap-precompute-rs/src/otap/mod.rs (Summary data points with sketch.envelope) and the deploy configs (exporter:otlp_http between stages).
• Grepped the diff for personal and company names.