<div dir="ltr"> Hello Squid team, First of all, thank you for maintaining such a robust and reliable proxy server. We have been using Squid in production for years in a small network with limited bandwidth (4 Mbps, about 15 users). Current situation: We use delay pools to share bandwidth fairly among users. The "class 1" pool does a decent job of dynamic sharing, but it lacks true QoS features like: 1. Traffic prioritization (e.g., web browsing > social media > downloads) 2. Burst support with borrow (when one class is idle, others can use its bandwidth) 3. Per-user fair queuing (not just per-connection) What we had to do: We ended up using multiple delay pools (one for web, one for downloads) as a workaround. But this is not true prioritization - it's just traffic separation. Suggestion: Would it be possible to add HFSC (Hierarchical Fair Service Curve) or similar queuing disciplines to delay pools? Even a simplified version with: - Priority levels (high/medium/low) - Bandwidth borrowing between classes - Fair queuing per client IP This would make Squid's QoS capabilities comparable to what pfSense, MikroTik, and Linux tc offer at the network level, but at the application layer where Squid operates. I understand this is not a trivial request, but even a basic implementation would be a game-changer for small networks with limited bandwidth. Thank you for considering this suggestion. Keep up the great work! Best regards, Yoan Manuel García Lleó Network Administrator </div>