{ "agent": "incident.meteorology@web-summit.dinis", "state": "second session; re-read at 11:30Z: 0 changed state; next re-read about 14:35Z", "last_wake": "2026-11-11T11:36Z", "started_by": "incident.conductor@web-summit.dinis (session_018yvwtQ4XRqTiSE3vjVq9KD), at Dinis Cruz's word (BRIEF.md, ~02:20 UTC)", "done": [ "session key made (record sha256:c5d0a303df53cf64, encryption sha256:162a3344a5c28014), public half in data/keys.json", "meteorology/lineage.json (grammar: 6 layers, 10 verbs each with its inverse), nodes/*.json (103 nodes), edges/*.json (143 edges)", "tools/meteorology-read.mjs: fetches the page each edge and licence cites and stamps state, time, sha256, quote; 123 edges stated on the page cited, 14 sites 'not stated', the rest marked unconfirmed or not read", "meteorology/licences.json: 17 sources, four answers each (read, store, show, transform), every row checked against its page", "meteorology/sources.json and samples/: ten open sources probed at 02:29 UTC for Lisbon and Madeira (sizes, cadence, how to read), four small samples; no model field in git", "tools/meteorology-build.mjs (a build generator): the map as SVG from the data, a page and an entity-shaped JSON per node, licences, sources, Storm Charles through the chain (from the data room)", "02:37Z round 1 live as v0.1.293 (94fd6bc)", "round 2 (3a7f61a): 144 edges, 125 stated on the page cited (ECMWF's fact sheet for what IFS assimilates; EUMETNET for MeteoAlarm; Météo-France's data portal for ARPEGE and AROME; AEMET for the storm-naming group); Weather Underground says its forecast input includes ECMWF and GFS; GraphCast's weights are CC BY 4.0 (README); each node page opens with its own small graph; charles.html opens with one storm in six layers", "round 3: 110 nodes, 152 edges (133 stated on the page cited, 13 not stated, 6 not read); EUMETSAT's licensing guide read in full (Core: CC BY 4.0; Meteosat level 1 an hour old or more: any purpose with attribution, no redistribution of the original numbers; newer than an hour: no public showing); Aurora MIT, FourCastNet BSD-3 (code); IPMA's site terms; ECMWF ensemble, GEFS, the IBI wave model, Instituto Hidrográfico's buoys and tide predictions; index: 'which model is behind which site' and 'what the reading found'", "round 4: the six-model table on charles.html rebuilt to incident.conductor@'s conditions (IPMA's warnings for Lisbon and Madeira first, from warnings.json; six single models, each with its run time as Open-Meteo reported it, labelled 'a model's output, read at Open-Meteo, evidence'; no average, no pick; 'does not forecast'); IPMA's own statement that its automatic forecasts are statistical processing of ECMWF and AROME (api.ipma.pt); Open-Meteo's upstream services from its licence page (pointers from incident.weatherwatch@, each read by us)", "round 5: 177 nodes, 226 edges (150 stated on the page cited, 54 not stated, 22 not read); every one of the 106 sites has a node and a page (only what each says itself; press pages keep no excerpt); meteorology/build-on.json and build-on.html: ten proposals with size, attribution line and what each needs, led by EUMETSAT's hour-old imagery (an 800 × 640 picture of the box is about 52 KB, probed and not kept) and NOAA's public-domain GFS fields (1.7 KB per step for the box); the map draws only the sites that say something; tools/meteorology-read.mjs reports what changed state", "05:20Z HANDOVER.md written; the 05:59Z re-read wake of this session cancelled (the next session runs the re-read on its first wake)", "05:23Z second session: new session key (record sha256:05ede4a171ccc4be, encryption sha256:29969fb5974c7b2c), public half in data/keys.json as incident-meteorology-dinis-2; the first session's entry marked handed over", "05:23Z re-read of every edge and licence (--stale): 258 items stamped, 0 changed state, 126 pages changed bytes with the statement still standing, 22 not read (as before: refused, blocked, or not requested because of their terms)", "05:23Z joins stated on the nodes (board card a-graph-2): \"entity\" on 3 nodes (body-ipma, body-aemet, svc-meteoalarm), \"of\" on 36 site nodes (the entity that publishes the site), \"source\" on 9 (the id in sources/registry.json); in meteorology/nodes/.json, graph.json and nodes/.json", "05:23Z names for the linker: body-open-meteo is now \"The Open-Meteo project\" (the bare name \"Open-Meteo\" goes to data-open-meteo), body-microsoft is \"Microsoft Research\"; \"ICON\" and \"Windy\" kept (every use outside meteorology/ is in research/weather-sites/, in the right sense)", "08:24Z re-read (--stale): 258 stamped, 1 changed state: site-sc-negocios-live visualises site-windy went from stated to not found in this read (the live blog moved on; noted on the edge); 125 pages changed bytes with the statement still standing; 22 not read, as before", "11:36Z re-read (--stale): 258 stamped, 0 changed state in the end. EUMETSAT's licence page (three rows) first came back \"not read\": the page answered, but slowly, and the headless reader dropped its status; tools/meteorology-render.mjs now keeps the document's status when the browser gives up waiting, and the three rows read \"stated\" again. 22 not read, as before; the Negócios live blog edge still \"not found in this read\"" ], "next": [ "about 14:35Z: the three-hourly re-read (node tools/meteorology-read.mjs --stale), scheduled into this session; report only what changed", "licences still unclear: Copernicus EMS products and Marine (the licence documents), Météo-France per data set, the Met Office's DataHub, FourCastNet's weights", "22 addresses not read (refused, blocked, or not requested because of their terms): listed on the map page", "if Dinis says yes to a build-on item: a GRIB decoder outside the build (GFS box to small JSON); a person registers with EUMETSAT" ], "asks": [ "incident.entities@: node ids are stable (obs-, model-, data-, body-, svc-, site-); an entity-shaped file per node is served at meteorology/nodes/.json", "incident.archivist@: the addresses the edges cite are in meteorology/graph.json (source, read_utc, sha256): candidates for Wayback captures", "incident.weatherwatch@: a lineage page and JSON per site is at meteorology/nodes/site-.md / .json (39 sites so far); your site pages can link there", "Dinis: two decisions on meteorology/build-on.html: (1) hour-old Meteosat pictures need a person to register with EUMETSAT and accept its licence, and design rule 9 lifted for that source; (2) GFS fields redrawn by us (public domain) need a GRIB decoder run outside the build", "incident.graph@: answered (a-graph-2): nodes carry \"entity\" (the same thing as that entity), \"of\" (a site: the entity that publishes it) and \"source\" (the id in sources/registry.json), each stated by us, in meteorology/nodes/.json and in graph.json; the join by host can become a fact where one is present", "incident.entities@: two node names changed: body-open-meteo \"The Open-Meteo project\", body-microsoft \"Microsoft Research\" (ids unchanged)" ], "findings": [ "IPMA's public charts page offers one model in its selector: ECMWF; IPMA's own page says it runs AROME at 2.5 km and derives products from ECMWF's global model", "IPMA's open-data terms (api.ipma.pt): personal or public use, never for profit, the source always named; silent on storing and changing: 'unclear' for both", "the only global model we may transform with no condition is NOAA's GFS (public domain); a gust field for the Lisbon to Madeira box through NOMADS's filter is 1.7 KB", "no Copernicus EMS activation for Storm Charles at 02:29 UTC", "BBC Weather names its forecast partner (MeteoGroup) and no model; 14 sites read do not say where their picture comes from", "EUMETSAT: three licences by product and age; Meteosat level 1 newer than one hour may not be shown on a public site without a broadcaster's licence", "Weather Underground states ECMWF and GFS (and NAM) as forecast input; BBC Weather names MeteoGroup; IPMA's site: ECMWF", "ECMWF open data: also at 0.1° from 20 October 2026; the portal is limited to 500 to 2000 simultaneous connections" ], "by": "incident.meteorology@web-summit.dinis", "updated": "2026-11-11T11:36Z", "session": "session_018yvwtQ4WgJUHwF2G8Ezss3" }