Polaris from MIT could substantially speed up web pages

The Web is in a sort of a rat race. While Internet connections get faster, web sites are getting more and more complicated, balancing, even negating, those speed improvements. A few tech companies, particularly Google, strive to introduce technologies as well as best practices to help speed up the Web, but those solutions are usually limited to specific browsers or specific circumstances. A new project from MIT called "Polaris", however, is aiming for a browser-agnostic method that could make web pages load as much as 34% faster.

One of the main bottlenecks of loading web pages is in how browsers have to fetch the data needed to assemble a single page. The browser doesn't know before hand which pieces it actually needs, like which images, Javascript, HTML, etc. Instead, it has to at least fetch one piece before it can know what else is needed. Rinse and repeat multiple times for even a single page and you've got a formula for inefficiency and latency.

Polaris, in contrast, tries to build a sort of map of a page's dependencies. This data can then be used by the browser to fetch as many as many files as needed to display a page, minimizing the round trips that must be made. One great feature of Polaris is that it is written in Javascript, albeit one that needs to run on the server side. This means it'll work with any browser, no modification necessary.

The program might bring up questions of security, since Polaris would inevitably know the whole website in order to function properly. The tests that yielded a 34% increase in speed were performed on 200 sites, especially popular ones. Of course, the Web is made up of hundreds more pages than that. The paper that details Polaris's functions will be presented at USENIX conference later this week. It could very well catch on, perhaps with a few safeguards, to truly make web surfing a smoother and faster experience.

VIA: Gizmodo