In the previous blog post in this series (Using WebSockets – Part 1: Minding the Gates), Ronnie Mitra talked about the promise of the WebSocket protocol, as well as some security aspects. In this post, I’ll talk about some of the details of the protocol and what they mean for those planning their own WS implementations.
The first thing to keep in mind is that WebSocket is a high-level protocol with its own registered schemes (WS: and WSS:). The specification describes it as: “… intended to be as close to just exposing raw TCP to script as possible.” This is very different from HTTP, which is “… an application-level protocol for distributed, collaborative, hypermedia information systems.”
That’s good and bad news. It means you have almost the full range of TCP at your disposal. It also means you have none of the established constraints and conventions of the more detailed and focused HTTP specification. This has implications for both design and implementation of WS solutions.
However, since many installed browsers do not yet natively support the WS protocol, these checks and balances are not always employed. Instead, WebSockets implementations often take advantage of browser workarounds and fallbacks, in order to support the real-time communications the WS protocol was designed to provide.
It’s also important to remember the specification states: “While this protocol is intended to be used by scripts in web pages, it can also be used directly by hosts [which] can therefore send fake ‘Origin’ header fields, misleading the server.” Implementations that will receive requests from non-browser clients should include additional checks to ensure these requests are valid.
Finally, as the protocol was designed to support real-time communications, it won’t scale in the same way HTTP does. Since the server will keep connections open to all active clients in order to track and broadcast content, servers will need to maintain (or persist) information about each connected client (including knowing when that client is no longer connected!)
If your current HTTP implementations rely on server-based session state, you may not see much difference in the scaling limits of WS. Remember though, the Web’s scaling success is largely based on HTTP’s ability to handle client requests without requiring server-persisted data. Also, some software and implementation patterns designed for HTTP will not work for WS.
Implementing WS is not for the faint-of-heart: it’s not yet widely supported on installed browsers; it uses a different implementation model; it takes more effort/resources to scale it up as services become popular. However, there are some good libraries for coding WS solutions and it can be relatively easy to get started on implementing WebSockets.
But be ready. If you experience great success, you’re likely to have a challenge on your hands!