Commit Graph
601 Commits
Author SHA1 Message Date
Claude 4cded65064 fix(satellites): stop requesting the CelesTrak sar group, which does not exist
The core catalog asked CelesTrak for GROUP=sar on every update. CelesTrak
has no such group (sarsat is search and rescue), so that read failed each
time and the layer row read DEGRADED with "1 CelesTrak group unavailable"
even when every real group loaded. Each failed read also put the TLE feed
into a 60 s cooldown.

Drop sar from the catalog groups, the client allowlist, the proxy
allowlist and the group-to-class table.

Co-Authored-By: Claude Fable 5.2 <[email protected]>
Claude-Session: https://claude.ai/code/session_01DLPqdcmfWbcfzrXmBKZc5z
2026-11-20 17:54:30 +00:00
German Khokhlov 23efd3b343 Merge main into codex/fix-cctv-session-admission
Main's Street View budget tests in the media feed-type file now mount the
provider through the shared CCTV test harness, which also resets the Google
rate-limit setting.
2026-11-20 17:39:46 +04:00
Denis Groshev 050e0a3a7a Merge main into denis/iss-live-telemetry 2026-11-20 13:12:36 +01:00
German Khokhlov a5d6963d41 Merge branch 'main' into codex/cctv-shared-batch-20261019 2026-11-20 14:20:48 +04:00
German Khokhlov 4e03e82987 Merge main 2026-11-20 13:51:17 +04:00
German Khokhlov b2f746cdb2 Merge remote-tracking branch 'origin/main' into fix/places-no-store-on-errors 2026-11-20 13:48:55 +04:00
German Khokhlov ed38a1ab3f fix(cctv): keep an HLS camera's health through the refused viewer's frame
A viewer refused for capacity falls back to /frame after hls.js gives up.
For a pullable HLS camera without a snapshot, /frame only has a
placeholder or a Street View frame, which says nothing about the stream,
yet it wrote degraded over the health that /media owns. /frame now
leaves that health as it is; a dead upstream is still reported through
/media.

Tests: the middleware suite covers both fallbacks, a real upstream
failure that still degrades, and cross-site segment and release refusals
on a live HLS lease; capacity checks use HLS_LIMITS and HlsCapacityError;
the three CCTV suites share one provider harness in src/testSupport.
CURRENT-STATE and the changelog fragment describe the refusal.
2026-11-20 11:48:22 +04:00
German Khokhlov 01d4a3f454 CCTV Street View budget: key the client before the camera wait, share the default
- The /frame handler reads the client's network key (IPv6 /64) before it
  awaits the camera, so a client that drops the connection is still
  charged to its own window instead of the shared fallback key.
- The budget is charged only after the Google key and finite catalog
  coordinates are confirmed, and the key is read once.
- Places and the CCTV window share one default (GOOGLE_DEFAULT_PER_MIN).
- Tests: the catalog-only test checks heading/fov/pitch from the query at
  the catalog location; the budget test covers unknown ids, forwarded
  requests and /64 grouping; a real-socket test covers a dropped
  connection. The duplicate pose test is removed.
- .env.example, SECURITY.md, CURRENT-STATE and the changelog fragment
  describe the separate per-IP window, the catalog position and the
  placeholder for seed cameras.
2026-11-20 11:31:14 +04:00
German Khokhlov 247452c202 Parse Lithuania catalogs inside the shared reader so bad JSON cancels the sibling 2026-11-20 03:39:07 +04:00
German Khokhlov 67bead9eb9 test(places): pin no-store on every refusal path and reuse the catch
Tests now check Cache-Control on the proxy's own 400, 405, 429 and
transport 502, a geocode 503 with ZERO_RESULTS on both geocode routes,
and an upstream 429 with an HTML body on both Places routes; the geocode
status cases move to the geocode suite and cover reverse-geocode too.
An unreadable Places 200 now throws into the existing catch instead of
repeating its 502 reply. Comments and the CURRENT-STATE rule are shorter
and match the code.
2026-11-20 03:36:48 +04:00
Said Turksever 75f1a3c181 Frame the port-drawn gizmo in the CCTV QA drag fixture and share the catalog-pair reader
The drag-fixture framing looked the gizmo up in viewer.entities, which main no
longer populates, so it did nothing and five drag checks stayed inconclusive.
It now reads the arrow from the layer's drawn state and records the framed
shaft length as a check.

Tarktee and Lithuania read their two catalogs through one helper,
readCatalogsTogether, so both get the shared deadline, sibling abort and the
release of a response that arrives after the other feed failed.

An Ontario error response has its body released, and tests pin the refused
redirect and system-code log labels.
2026-11-19 18:06:58 +00:00
Said Turksever 4e1f004a83 fix(cctv): keep a camera's health when an HLS viewer is refused for capacity
A full lease or session budget fell into the general catch and marked the
camera degraded, although its upstream was fine and other viewers were
watching it. Capacity refusals now throw HlsCapacityError and still answer
503 with Retry-After, without touching health.

Tests: fill all 16 lease slots from one shared address and pin the refusal
of a ninth viewer and a third session, keep a second lease alive so the
released-lease 404 is the lease check, and share recordingResponse between
the two middleware test files.
2026-11-19 17:45:15 +00:00
Denis Groshev 0620d01140 Merge main into denis/iss-live-telemetry
Conflicts: the locale catalogs, package boundaries and format scope are
three-way merges (entries main removed, such as displayPanelScroll.js,
stay removed); applicationShell.js and ContextView.vue keep main's Space
Missions panel and the STATION and LIVE wiring, moved back into
_initGlobalContextPanel; installations take main's feedback module, and
their label box joins main's card-on-top click rule; ports.test.mjs lists
modelParts and modelNodes; CURRENT-STATE.md keeps both edits.

zh-Hant is regenerated from zh-CN, as main now requires. "Below" keeps
main's translations.
2026-11-19 17:07:48 +01:00
German Khokhlov 0699876e4e fix(cctv): keep main's HLS leases and test them through the middleware
Drop the per-peer lease budget and the lease-to-address binding. Main's
keyless same-site gate already refuses cross-site media requests, while
the budget capped a shared address (Docker gateway, reverse proxy, NAT)
at 8 leases across both sessions and still let one address hold both.
A lease now keeps working when its viewer reconnects from a new address.

Add a middleware test that drives the HLS playlist, segment and DELETE
for viewers on one shared address, on distinct addresses and after an
address change, and fold the media cases (segment read, DELETE, no
catalog request) into the cross-site table test. Remove the redundant
puppeteer flag, the CURRENT-STATE paragraph and the changelog fragment,
which no longer describe a change.
2026-11-19 19:59:01 +04:00
German Khokhlov 01368b95bf Merge remote-tracking branch 'origin/main' into fix/cctv-streetview-fallback-catalog
# Conflicts:
#	SECURITY.md
2026-11-19 19:12:56 +04:00
German Khokhlov 11e1c9d26d CCTV Street View fallback: the browser's pose at the catalog position
The fallback keeps the catalog position but takes heading, field of view
and pitch from the request again, as before, clamped as before. A catalog
camera with no pose (every world-cameras row on main) otherwise asked
Google for heading 0, pitch 0, FOV 80 while the browser draws it at the
id-hash heading, pitch -17 and FOV 74, and a calibrated camera lost its
calibration. Cost and the per-IP budget are unchanged.

- The Street View budget is built per proxy, so tests no longer depend on
  GEV_RATELIMIT_GOOGLE_PER_MIN in the shell; the fixture pins it.
- New tests: a pose-less catalog camera, and the per-IP budget (no call
  without a key, the cap, a separate budget per address).
- The client no longer sends lat/lon on /frame.
- CURRENT-STATE, the changelog fragment and SECURITY.md say the fallback
  has its own per-client limit, and that unlisted (seed) cameras get the
  placeholder.
2026-11-19 19:06:53 +04:00
German Khokhlov bdd92663d6 Cancel the Tarktee sibling on a body failure and settle review notes
Each Tarktee DATEX feed now reads its body as soon as its headers arrive, so a
failed or oversized body aborts the other request at once instead of waiting
for its headers up to the shared 30 s deadline. Status and redirect errors
name the feed again.

Ontario failures log the error class and system code, never the key-bearing
URL or upstream text. DATA_SOURCES.md and the camera licence field use the
Open Government Licence – Ontario statement the credit shows, and a test pins
that credit.

The focus-release rendering test now uses main's icon renderer and its
isVideoFeedCamera contract. The Tarktee cancellation tests disable the 30 s
deadline and time out after 2 s, so a regression fails fast instead of passing
late or hanging; a duplicated older test and an unused import are gone.

This PR's notes move into changelog.d/1266-cctv-shared-fix-batch.md: the 1133
fragment described main's tracking probe and terrain check, not this branch.
2026-11-19 18:59:04 +04:00
German Khokhlov 19542baf61 Merge remote-tracking branch 'origin/main' into fix/places-no-store-on-errors 2026-11-19 18:29:27 +04:00
German Khokhlov 31b6b25d8b fix(places): make no-store the default and refuse an unreadable Places 200
Every Google route now sets no-store right after the same-site gate and
switches to private, max-age=300 only on a confirmed answer, so the
proxy's own 400, 405, 429 and 502 replies are no longer left without a
header. Both Places routes answer an unreadable 200 body with 502 instead
of caching an empty result, as the geocode routes already did. The test
pins the allowlist with OVER_DAILY_LIMIT and UNKNOWN_ERROR and covers the
unreadable and empty Places bodies; the changelog fragment is shorter.
2026-11-19 18:29:24 +04:00
German Khokhlov 90e464eb5a Merge main into codex/fix-cctv-session-admission
Resolve the conflicts by keeping main's keyless CCTV gate, which already
refuses cross-site requests on every /api/cctv route and keeps the routes
reachable behind a reverse proxy, instead of the strict gate on /media/.
Keep the per-request client id the lease calls use, and both sides of the
CURRENT-STATE paragraphs and the media tests.
2026-11-19 17:58:14 +04:00
German Khokhlov d7db1090fd Merge main and resolve the CCTV batch conflicts
Main released #1228 and #1233 and deleted their changelog fragments, so they
stay deleted here. Main also rewrote the tracking arrival probe in
scripts/track-regression.mjs, which supersedes this branch's version; main's
file is kept unchanged.
2026-11-19 17:56:23 +04:00
Manjunath PaparajuandSameh Khamis d68adcea21 Remove the Cyber HUD layout and keep its sonar as a visual style
Display > HUD > Sonar now turns the sonar sweep on over Tactical or
Minimal. It is off by default, ON reveals its five sliders, and it keeps
its guards: an OFF control, the sweep only while the HUD is visible and
outside Cockpit, no animation under reduced motion, and repaint timers
re-armed only after a drawn frame that still shows contacts.

The Cyber layout, its shell skin and its Cockpit lanes are removed. Share
links, scenes, saved views and set_hud that name Cyber open Tactical with
Sonar on; links and scene shots record Tactical with Sonar under the same
code, and a restored state that names another layout turns Sonar off.
set_cyber_sonar keeps its name and works under any layout. A newer Sonar
toggle or voice setting wins over a share-link restore that lands later.

From 1200px wide the command dock and World Timeline sit 74px above the
bottom edge, as in the Omnibar layout, leaving the bottom row to the map
attribution (left) and the Power Up corner row (right). The attribution
moves to the bottom right when the page runs right to left, where the
corner row sits at the bottom left, and a long credit line wraps upward
before reaching the corner row. The Tactical bottom-right readout sits
8px above Power Up and the coordinates start level with it. The World
Timeline is narrowed to sit between the corners. Toasts keep clear of
the raised dock and the chat and voice panel height budgets subtract its
rise. Below 1200px the layout is unchanged. The left rail treats the
centered World Timeline and voice control as obstacles only where they
reach a laid-out panel.

On top of the Omnibar: sonar.css replaces cyber.css after the Omnibar
imports, the Omnibar's HUD guide step now says Layout picks Tactical or
Minimal, and the Omnibar's Tactical bottom-right readout keeps the same
8px above Power Up.

Co-authored-by: Sameh Khamis <[email protected]>
2026-11-18 15:47:27 -08:00
Sameh Khamis 9554d6ea71 Serve the Data Library from any server, and let a server shape it (#1807)
A server can answer the library with createLibraryHandler, a fetch handler, and shape what it offers: hide entries (listed: false) or route browser-direct ones through itself (fetch: proxied). View-bounded entries ask for a grid-snapped box, and GeoParquet is read on the server within it. Applications can pass their own tile and data fetchers (createApplicationTools({library: {fetchTile, fetchData}})), and with library.requireCatalog a catalog that does not answer offers nothing and is asked again later. Tile paths have one spelling: integer z/x/y without leading zeros, within the grid and the entry's zooms. A server's index and partition reads are bounded and counted against its subrequest limit.
2026-11-17 23:34:00 -08:00
Denis Groshev 2c2e9de14d ISS telemetry relay: cap frames on the socket, time out the handshake, slower ladder for flapping sessions
- The feed socket is opened with maxPayload at the 64 KB frame cap and a
  10 s handshake timeout: ws would otherwise buffer up to 100 MiB before the
  cap applied.
- Only a session that has run a minute brings the retry ladder back to its
  first step, so a feed that sends a snapshot and then END keeps backing off;
  the ladder's later steps are tested.
- The rate limit keys IPv6 clients by /64, and any path under the route is a
  JSON 404.
2026-11-17 16:28:57 +01:00
Denis Groshev 71d57481b7 Merge main (624c229b): STATION and LIVE sit in the Inspector tab after the entity card
Main split Context into Inspector and Properties tabs and renders the
Inspector only while Context is open; the two sections join it there, after
the entity card. Locales, same-as-english and package boundaries keep both
sides, deduplicated (the branch had added 54 duplicate module entries).
2026-11-17 16:18:26 +01:00
Denis Groshev 7700f25bfb Format the telemetry reconnect test 2026-11-16 21:54:19 +01:00
Kevin Rajan 79ab830824 fix(satellites): forget ISS telemetry from dropped sockets
A new ISSLive connection must rebuild its own joint snapshot and station-clock
liveness. Without clearing the prior session, a fast reconnect can mark old
joint positions live before fresh angles arrive.

Add a node:test reconnect regression and update the runtime contract.
Follow-up to PR #1767 by ZloyS; original ISS model credit: IMBrandao (#1690).

Tests: node --test server/providers/space/issTelemetry.test.mjs (UNRUN)
2026-11-16 21:54:05 +01:00
Denis Groshev aeaa192b5f Review fixes: station panel and model housekeeping, label-box click, telemetry null stamp
- Installations: the label-box click fallback reaches the controls' methods
  (it read an undefined member and never ran).
- Station model: the docked-companion scan runs once a second again; the
  hand-over pass runs once, when the model starts drawing.
- Station panel: an unchanged view is not republished every frame; one
  render pump across slider drags, stopped on destroy; destroying the panel
  hands the joints back to Automatic, and the joint drivers' reset forgets
  manual choices and the telemetry lock.
- Telemetry relay: a null TimeStamp is no time, not the first second of the year.
- Drop issCommandShouldShow and issCommandNoteText, which nothing called.
2026-11-16 19:31:32 +01:00
Denis Groshev 9cc077964f ISS live telemetry, STATION and LIVE sections, Tiangong model, drawn through the renderer's draw port
Builds on the ISS 3D model (PR #1690) and brings it onto main's renderer
seam: the station models are items of a `models` draw set, posed by
heading, pitch and roll from the station's flight axes
(`headingPitchRollOfAxes`), with the camera, canvas and clicks read from
the renderer's ports; no satellites file reaches Cesium. The joint drivers
write column-major matrices into node poses the model part keeps, for a
renderer that can pose nodes; until the draw contract has that member the
models draw rigid.

- ISS joints take their angles from NASA's ISSLive telemetry, relayed by
  the server behind ISS_TELEMETRY_ENABLED.
- STATION and LIVE sections of Context replace the floating command panel
  and player; both collapse like the Radio section.
- Cockpit rides the ISS from mounts on its model; the LIVE page in the
  Cockpit briefing.
- Tiangong as a second station model, with computed joints.
- Installations: a label-box click selects the site.
2026-11-16 18:39:39 +01:00
Manjunath 9e9ce1db22 Fix Stories navigation, sharing, and voice entry 2026-11-16 20:12:10 +05:30
Sameh Khamis 39c89d563b Draw transit vehicles as trolleybuses, cable cars, funiculars, gondolas and monorails
The transit layer knew five modes, so a trolleybus drew as a bus, a
funicular as a train and a coach as a city bus. The GTFS basic route types
are now modes of their own (trolleybus, cable_tram, monorail, funicular,
aerial_lift), each with an icon, colour, detection tier, card word, voice
name, speed ceiling and fleet model.

A vehicle record may carry its route's GTFS route_type as routeType. It
settles the mode after the feed's fleet hint and before its route hint, and
picks the finer model (a coach for 200, a high-speed set for 101) when the
type is a kind of the vehicle's mode. Retained history classifies epochs
through transitModeFor too, so trails join the live vehicle.

Route hints from each operator's routes.txt: Prague's trolleybus lines
(51-59, 232, 275) and Petrin funicular, and the trams, trolleybuses and
metro lines of Bucharest, Cluj-Napoca and Turin. A feed may also name
route corrections that outrank the route type: Le Havre's routes.txt types
its funicular (8-FUNI) as a bus, so it is corrected to a funicular.
2026-11-14 12:03:12 -08:00
Sameh Khamis 52507a0755 Draw no buoy for a virtual aid to navigation
A virtual aid (AIS message 21's VirtualAtoN flag) is a mark broadcast
where nothing physical stands, so drawing it as a buoy showed a mark
that is not there. The vessel source record now carries `virtual`
(false unless a source says true; any source may set it), AISStream
sets it from VirtualAtoN, and the vessel layer keeps such an aid's
disc: vesselModelSpec returns null for it and the layer gathers no
model item for it. Its card, selected card and readout add VIRTUAL
after its kind.
2026-11-14 11:07:13 -08:00
Sameh Khamis 6c7fe48f40 Show an aid to navigation's kind, not a speed or heading
A buoy's card, selected card and HUD readout showed "0.0KT · 135°" from
whatever its source reported. A fixed mark has neither, so they now show
its kind alone, whole ("CARDINAL MARK N", which the 14-character type cut
read as "CARDINAL MARK "), in the bone white of its disc.

isAidToNavigation now knows every kind an aid report names ("Light, with
sectors", "Fixed structure off shore" had no aid word), leaves out a
"Buoy Layer" as it does a tender, and matches each type text once: the
vessel layer asks per vessel on its rotation pass, which now skips the
aid discs it never turns.

Aid rows no longer cost ships their place: aisStreamRows lists every
vessel ahead of every aid, so the layer's 12,000-row request never drops
a ship for a buoy, and vessels_in_area (and the brief built on it)
counts ships only. An aid's name runs on into its NameExtension.
2026-11-14 10:02:59 -08:00
Sameh Khamis 279aedb2f4 Draw aids to navigation as buoys
A buoy, beacon, RACON, light or IALA mark in the vessel feed was drawn
as a ship. A vessel record whose type names one, or whose MMSI is an
aid-to-navigation station's (99MIDxxxx), is now isAidToNavigation and
vesselModelSpec draws it as the fleet set's buoy of its mark (cardinal
N/E/S/W, port or starboard hand, safe water, isolated danger, special
mark; a red can when the type names no mark), at its built size and
flagged fixed so its model does not turn to a heading. Its icon is a
disc instead of a chevron.

AISStream now subscribes to AidsToNavigationReport by default, and the
server names each report's aid type code (0 to 31) as its kind instead
of reading it as a ship type.
2026-11-14 09:54:01 -08:00
Sameh Khamis 7b93045895 feat(vessels, transit, rail): draw fleet models under the 3D toggle
Vessels, Transit and Rail turn nearby icons into fleet models at the
aircraft layers' distances. A vessel is drawn at its AIS length and beam
(static dimensions now reach vessel rows), standing at the sea surface; a
transit vehicle or train stands on its road or rail, raised by its type's
elevation, and is painted in its mode or category colour where the model
marks it.
2026-11-13 07:16:25 -08:00
Said Turksever 24ccdd4dc3 Fix Data Library share restore, dense tile layers and EPC partition reads
- Share links: the compact /share/<code> format carries the switched-on
  library entries (field 29), so Copy Link keeps them; a share opened in the
  same tab restores them; any share decides what is on, so one with every
  library entry off no longer switches on the recipient's stored entries.
- A tile entry too dense at its minimum zoom (Norway cadastre z14, depth z11)
  is hidden below that scale via minimumTerrainLevel instead of allocating
  millions of tiles when zoomed out.
- EPC partitions keep the full extent for exclusion and the 98 % extent for
  ranking only; the old trimmed-only extents are retired.
- Partition index footers that fail transiently are retried; an incomplete
  index is never cached as complete, a complete prior index is kept, and it
  is rebuilt after five minutes.
- Views that leave touching partitions unread are partial; listing and
  footer bodies are capped and ranges validated.
- The Browse dialog relabels after a language switch.
2026-11-12 22:37:34 +00:00
Said Turksever db80d82b8d Add the Data Library: live regional datasets, tiles and packs
DATA LAYERS keeps its curated core layers; regional and niche sources
live in the Data Library, opened from Browse data library at the foot
of the panel. Each entry is a JSON manifest in src/library/entries/
pointing at a live upstream (query API, provider-hosted file or tile
endpoint), so nothing is bundled and a self-hosted instance hosts no
data. Sources without CORS go through one allowlisted /api/library
route. Kinds: dataset (GeoJSON, CSV, JSON records, ArcGIS, partitioned
GeoParquet), tiles (XYZ, WMS/ArcGIS bbox export, terrarium relief) and
pack. Selections persist and travel in share links as lib=, separate
from the core layer tokens. Library rows get Zoom and Remove icon
chips that work while the layer is off, rendered through RailChips.
Strings are translated in every locale.
2026-11-12 12:26:45 +00:00
Denis Groshev 9b3ca481ce Merge upstream main into denis/rail-layer 2026-11-11 23:23:58 +01:00
German Khokhlov 94ff78be3b refactor(routes): inline the single-use 502 responder
The outage branch is the only caller, so it writes its 502 directly.
Same status, headers and body.
2026-11-11 21:23:35 +04:00
German Khokhlov 82e2b716bc docs(routes): tighten the outage comments and notes
Shorter comments, changelog and CURRENT-STATE text for the 502 outage
path; the CURRENT-STATE paragraph now also names the 429 exception and
which budget expiry answers busy and which answers 502. Text only.
2026-11-11 21:21:07 +04:00
German Khokhlov a29611ca17 Merge branch 'main' into upstream/route-outage-unavailable 2026-11-11 21:16:50 +04:00
Denis Groshev 13818d0b9f Merge upstream main into denis/rail-layer 2026-11-11 18:02:52 +01:00
Denis Groshev 830c080803 feat(rail): add a Rail layer with keyless live trains
Live trains from 30 keyless operator feeds: GPS positions from 29 and
Irish Rail's station reports, shown as the operators publish them. 28
work out of the box; Renfe long and medium distance and KTMB appear once
their pack is built, because their vehicles name no route.

Packs are opt-in (scripts/build-rail-paths.mjs): the operator's own
route lines, stop names, next stops and trip-update delays for live
trains. Nothing downloads when the app starts, no train is placed from a
timetable, and no request carries a deployer's key.
2026-11-11 16:06:22 +01:00
Said Turksever f878efe22d Merge commit '0ceee8df40b97c6c693e86a4443794cd4ed1ff35' into HEAD 2026-11-11 08:54:28 +00:00
Sameh Khamis bc786f8c55 fix(news): Conflicts reports of the same time come in a fixed order
GDELT exports are read in parallel, so two reports with the same time
arrived in whichever order their exports finished, and the feed test that
pages through them failed now and then. Same-time reports now order by
most mentions, then by id.
2026-11-10 22:40:15 -08:00
Sameh Khamis b89f81b729 fix(voice): keyless Gemini voice drops the microphone hint
The Gemini token route now answers a missing key with code API_KEY_MISSING,
the session carries it to the error state as missingKey, and the voice
control hides "Check microphone permission" for that error, as it already
does for OpenAI. The OpenAI controller sends the same flag, so the shared
control and the controller agree whichever updates last.
2026-11-10 17:06:50 -08:00
Said Turksever 09d02b6e7a Keyless voice: say where to add the OpenAI key instead of blaming the microphone 2026-11-10 16:11:17 -08:00
Sameh Khamis 8ff22f1a12 Merge main into feat/shipwrecks
Main added the News layer at the same places: both layers are kept in the
layer names, panel order (News stays last of Events), standalone sources,
voice manifest, credits, DATA_SOURCES and the CURRENT-STATE table. News took
share-link token 00 first, so Shipwrecks moves to 01 (ledger, registry,
CURRENT-STATE and the token tests). The catalog now has 40 layers and the
share-link registry 38. Shipwreck strings are added to the new Hebrew
catalog; the other catalogs keep both sides' strings.
2026-11-10 14:03:39 -08:00
Sameh Khamis 7704de5075 fix(shipwrecks): the page expires each source's wrecks by its own clock, from the age the proxy reports
The page compared the proxy's fetchedAt with its own clock, so a skewed
browser clock cleared or kept wrecks wrongly; a snapshot without a timestamp
never expired; and one old source cleared the other's fresh wrecks. The proxy
now reports each source's age on its own clock and its seven-day limit; the
page turns them into per-source deadlines when the snapshot arrives (seven
days from arrival when no age is given) and, on a failed refresh, removes only
the sources past theirs. A refused short refresh is logged.
2026-11-10 13:38:31 -08:00
Sameh Khamis 66fba3da47 fix(shipwrecks): a source refresh with less than half of its last good wrecks is refused until the old set expires
A 200 answer that was valid but cut short (30 Wikidata wrecks instead of
about 1,450) replaced the last good set for a day. It now counts as a failed
refresh: the old records are served as stale and the source is retried after
15 minutes; once the old set is past seven days, the new one is taken.
2026-11-10 13:36:18 -08:00