Learning HTTP/2

Chapter 9. What Is Next?

We may have had a 20-year gap between HTTP/1.1 and HTTP/2, but from the looks of the current state of experimentation and research it seems unlikely we will see another decades-long break between versions. There is work going on even as adoption of h2 rises that is likely going to produce the next great thing in the world of internet protocols. To understand where this rapid iteration is coming from and the ideas behind it, it is important to have a bit of a background on the stack HTTP is built on.

TCP or UDP?

This is a debate you hear a lot these days, and though the question is valid it is also misleading. We discussed TCP in Chapter 3 when looking at the underlying motivations behind h2. To quickly recap, TCP is an IP datagram–based protocol that provides an agreed concept of a connection, reliability, and congestion control. These are the elements that allow TCP to be dependable and work consistently in a crowded internet. UDP (User Datagram Protocol), on the other hand, is much more basic. In this protocol datagrams (packets) are individual with no relation to any other UDP packet. There is no “connection,” no guarantee of delivery, and no ability to adapt to different network conditions. It is perfect for building simple protocols like DNS that have small individual queries and small responses.

In order to make an application based on UDP be something that our next generation web can be built upon, we need to implement the idea of a connection, ...