0xMiden / 0xMiden/note-transport-service

No rate limiting or per-client request quotas

未關閉
#117 2 則留言 0 個 reaction 已指派 0 人 在 GitHub 檢視
enhancement production-readiness
主要語言
Rust
星號
3
分支
10
平均合併
2 小時 23 分鐘
30 天內合併 PR
4

描述

Severity: critical (for an internet-facing deployment).

### Summary

There is no per-IP, per-client, or per-RPC rate limiting. The only throttle is a global 4096 in-flight-request limit (`GlobalConcurrencyLimitLayer::new`, `crates/node/src/node/grpc/mod.rs:121`). Remote peer address is never inspected. A single client can saturate the write, read, and stream paths at line rate.

This is the concrete mechanism behind placeholder #44 (spam protection).

### Recommendation

- Add per-IP rate limiting (tower middleware, or document a mandatory upstream proxy/WAF) with separate budgets for writes (`SendNote`) vs reads (`FetchNotes`/`StreamNotes`).
- Add a global write-rate cap.
- Reject with `RESOURCE_EXHAUSTED` rather than queueing (see the concurrency/timeout-ordering note in the papercuts issue).

Related: #44, and the auth-model issue (rate limits are only meaningful once "who is a client" is defined).

---
Part of #114.

貢獻指南

開啟貢獻指南

評估

這個 Issue 還沒有評估資料。

把新 issue 寄到你的電子郵件信箱

精選適合新手參與的 GitHub issue 摘要。