PHP Streaming and Output Buffering Explained

Recently, Rails launched a feature called Streaming. Even though streaming is not a new concept, having it built into the framework to let developers build rapidly responding applications easily is definitely a huge step. As a PHP developer, I was wondering whether we can have something similar in our favorite language? The answer is yes - we can easily have streaming in PHP applications with little effort, but in order to get this right we have to become familiar with some underlying concepts.

In this article, we will see what streaming is, what output_buffering is and how to get our desired result under different webservers (Apache, Nginx) and PHP configurations (CGI, mod_php, FastCGI).

Streaming and Output Buffering

Streaming

Streaming in web refers to the mechanism of sending responses in chunks rather than sending them at once. In the traditional HTTP request / response cycle, a response is not transferred to the browser until it is fully prepared which makes users wait. This limitation is a remnant of the weak software and hardware of the past, but everything has changed a lot since then. Network systems and browsers are now powerful enough to handle transmission of content in a fast and efficient way. Imagine trying to watch a video and not being able to watch it until it’s fully downloaded into the player’s buffer. With the help of streaming, we can watch videos or listen to our favorite music quite efficiently as content is being loaded instantly while the rest of the data is being downloaded behind the scenes.

Streaming seems ideal for giving users the perception that a web-app is loading fast. Streaming of HTTP responses, however, is quite different from streaming of media content. Streaming of HTTP responses simply means sending a response in fixed or variable size chunks to the browser while the webserver is preparing the remainder. For example, you want to display a list of Hollywood movies on a single page. If you prepare the full response first and then send it to browser, the end-user will definitely feel the delay. But if you send 100 movies in one chunk and display them in the browser while you are preparing the html for the next 100 movies, the content will feel like it’s loading quite fast.

HTTP responses don’t consist just of renderable items - there are plenty of other things such as response status code, HTTP headers, cookies etc. that are essential parts of a response, but go unnoticed by end-users because they’re never rendered. Without them, our content has no meaning at all. Instead of letting users wait, the bigger sites send the non-renderable information to the browser and once the browser starts receiving it, it starts rotating the loading indicator you’ve almost certainly seen when on slower connections. Stylesheet and JavaScript files remain unchanged for most of the time, so many sites also send them along with non-renderable content chunks, so that the browser starts fetching and executing them while the rest of the response is being prepared. This is quite a powerful technique for creating the illusion of speed. When the content of the body tag is generated, it is sent to the browser, and that content can be sent in chunks again, further propagating the illusion of speed.

Continue reading %PHP Streaming and Output Buffering Explained%

more