Extract duplicated retry logic from OverpassService and RoutingService
into a shared resilience framework in service_policy.dart:
- ResiliencePolicy: configurable retries, backoff, and HTTP timeout
- executeWithFallback: retry loop with primary→fallback endpoint chain
- ErrorDisposition enum: abort / fallback / retry classification
- ServicePolicy + ServicePolicyResolver: per-service compliance rules
(rate limits, caching, concurrency) for OSMF and third-party services
- ServiceRateLimiter: async semaphore-based concurrency and rate control
OverpassService now hits overpass.deflock.org first, falls back to
overpass-api.de. RoutingService hits api.dontgetflocked.com first,
falls back to alprwatch.org. Both use per-service error classifiers
to determine retry vs fallback vs abort behavior.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Route calculation to alprwatch API fails with HTTP 400 because
built-in profiles include empty tag values (e.g. camera:mount: '')
that get serialized into the request body and rejected by the API.
Add routing_service_test.dart with 5 tests:
- Empty tags filtered from request (reproduces the bug)
- Successful route parsing
- HTTP error handling
- Network error wrapping
- API-level error surfacing
Add node_profile_test.dart with 4 tests:
- toJson/fromJson round-trip
- getDefaults returns expected profiles
- Empty tag values exist in defaults (documents bug origin)
- Equality based on id
Tests require RoutingService to accept an injectable http.Client,
which will be added in the next commit along with the fix.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
