Independent AT Proto non-archival relay service
Charlotte Som
e35ec83c0c
constant listening spam is so noisy (because we keep getting UnexpectedEOFs) |
||
---|---|---|
src | ||
.editorconfig | ||
.gitignore | ||
Cargo.lock | ||
Cargo.toml | ||
flake.lock | ||
flake.nix | ||
README.md |
cerulea-relay
Realtime non-archival relay for third-party AT Proto PDSes.
In the interest of cost control, we are scaling down the network:
- Only PDSes with fewer than 1000 repos are crawled
- We do no backfilling, only current events are relayed to consumers
- Stale data (≈ 24hrs?) is purged from the database [not doing this yet]
The idea is that we can have apps with much larger limits if we scale down the volume of the network.
- Large block sizes
- Large record size limit
- therefore: Large text field in post records, large uploads
- etcetcetc
todo
- metrics / tracing / observability shit
- history:
- store indexedAt values
- purge based on ttl
- takedowns
- more comprehensive commit validation
- how good can we make this without having the actual repos?