Today's GNU Radio hits hard limits when it comes to a few things that are absolutely crucial for modern communication stacks: It doesn't make any guarantees on latency, and its architecture doesn't allow for tight integration with hardware accelerators. And whilst most communications are packet-based, packeted data is a second-class citizen in the kingdom of sample streams that is GNU Radio.

In this talk, we'll discuss why that is the case, and what can be remedied within the current framework, and what not. We'll try to assess what usage paradigms are worth keeping for the future of GNU Radio, and what needs to change.