Oh, Websockets! Never has a technology so unambiguously demonstrated that we abused the web to become a new desktop where we run (mostly) cross-platform pieces of minified blackbox JS and need the same types of abstractions like you'd find in a real OS, like, say, sockets.
That being said, it was better than imperfect emulations of the same concept that pre-dated it by ~6+ years (like long polling). I'd even go so far as to say that the rise of websockets and the excitement over streaming real-time data between webapps greatly contributed to the traction of Node.js, which was the runtime for socket.io.
Socket.io offered one consistent API that just worked, and internally it wrapped websocket or its workarounds, depending on where the code ran. It was beautiful. I remember many tutorials about Node.js starting off with socket.io.
Now, with HTTP2, Websocket is sort-of abandoned -- not because HTTP2 is that good, but no one is excited about WS anymore. But once QUIC becomes mainstream and HTTP2's imperfect attempt to reinvent a transport protocol over an existing transport protocol is subsumed into HTTP2-over-QUIC, there will be zero need for WS.
It's true that WebSockets is nothing new, and the title did make it sound like the next new shiny thing. I wouldn't say that it's abandoned though - its adoption is still spreading from where I see, and it's very useful during build/development too, i.e., live-reload on code change.
"..that we abused the web to become a new desktop"
Oh, Websockets! Never has a technology so unambiguously demonstrated that we abused the web to become a new desktop where we run (mostly) cross-platform pieces of minified blackbox JS and need the same types of abstractions like you'd find in a real OS, like, say, sockets.
That being said, it was better than imperfect emulations of the same concept that pre-dated it by ~6+ years (like long polling). I'd even go so far as to say that the rise of websockets and the excitement over streaming real-time data between webapps greatly contributed to the traction of Node.js, which was the runtime for socket.io.
Socket.io offered one consistent API that just worked, and internally it wrapped websocket or its workarounds, depending on where the code ran. It was beautiful. I remember many tutorials about Node.js starting off with socket.io.
Now, with HTTP2, Websocket is sort-of abandoned -- not because HTTP2 is that good, but no one is excited about WS anymore. But once QUIC becomes mainstream and HTTP2's imperfect attempt to reinvent a transport protocol over an existing transport protocol is subsumed into HTTP2-over-QUIC, there will be zero need for WS.