Harden the server against anonymous load and small leaks
- github
- github/openchronology/timeline-app
- commit
- main
- github/athanclark
- Request bodies: each route anyone may call reads only what it needs. A timeline query takes at most 1 MiB and 64 KiB (a comparison's plugins and the rest), the desktop and social sign-in starts and the desktop poll 1 KiB. A body whose Content-Length is over the limit is refused before it is read, and JSON nested more than 128 levels deep is refused before parsing. Saves and imports by signed-in accounts keep 16 MiB. - Queries without an account are limited to 600 a minute per address in each process, counted before the body is read. - Separation views: only the filter's tags a timeline has count (and a relationship filter's entity only if it has it), so filters naming what it lacks build nothing new: the side without them is the timeline's own index, and the side with them one shared empty view. Builds are limited to 20 a minute per address without an account and 60 per account, and run two at a time per process with up to 16 waiting (503 beyond). Views already built are read without either, and each separation's plan is kept per revision. - Per-address limits count an IPv6 client by its /64 network, and an IPv4-mapped address as IPv4. - Proof of work: the salt must be exactly as issued, with fixed-width fields and an expiry no later than a new challenge's, so an answer cannot be split differently between salt and number to count again. - Resolving a pasted address answers a missing timeline and one the reader may not open with the same 404. - The LDAP filter takes the escaped username through a replacement function, so `$` patterns in a name stay as typed. Co-Authored-By: Claude Opus 5.5 Claude-Session: https://claude.ai/code/session_011fFY8eruEi1xAkkMBni13B https://github.com/openchronology/timeline-app/commit/988e83e20df9bd860e43d0f6e1d594b45b74f1d1
Open Harden the server against anonymous load and small leaks in the timeline