Data Sync

Source

RFC data is mirrored from rfc-editor.org via rsync:

rsync -avz --delete \
  rsync.rfc-editor.org::rfc-index/rfc-json.tar.gz \
  rsync.rfc-editor.org::rfc-index/rfc-text.tar.gz \
  rsync.rfc-editor.org::rfc-index/rfc-html.tar.gz \
  data/

Incremental reindex

scripts/ingest.sh performs an incremental reindex — only changed RFCs are reparsed. This is fast (~seconds) on subsequent runs after the initial full build.

Schedule

Inside the container, scripts/entrypoint.sh runs a background loop:

Sync data from rfc-editor (immediately on first boot)
Run python -m rfc_mcp.ingest (incremental)
Sleep RFC_SYNC_INTERVAL seconds (default 24h)
Repeat

The server starts immediately; the health endpoint reports "starting" until the index is ready.

Manual sync outside Docker

scripts/sync_data.sh
scripts/ingest.sh           # incremental
scripts/ingest.sh --full    # force full rebuild

Section parsing tiers

Each RFC's sections come from one of two sources:

Tier	Source	Reliability
`structured`	xml2rfc HTML with `<section id>` markup	High (last ~10 years)
`heuristic`	Pattern-matched from plain text	Best-effort (older RFCs)
`null`	No sections could be determined	—