System loads web pages 34 percent faster by fetching files more effectively

Interesting. The paper was released before HTTP/2 was in widespread use. They do show that their approach has significant improvements over SPDY alone...I wonder how the comparison to HTTP/2 alone would fare.

Relevant paper: &quot;Polaris: Faster Page Loads Using Fine-grained Dependency Tracking&quot;, <a href="http:&#x2F;&#x2F;web.mit.edu&#x2F;ravinet&#x2F;www&#x2F;polaris_nsdi16.pdf" rel="nofollow">http:&#x2F;&#x2F;web.mit.edu&#x2F;ravinet&#x2F;www&#x2F;polaris_nsdi16.pdf</a>

but we can already make web pages load 500% faster by not shoveling a ton of shit, not loading scripts from 60 third-party domains (yes stop using CDNs for jQuery&#x2F;js libs, those https connections aren&#x27;t free - they&#x27;re <i>much</i> more expensive than just serving the same script from your existing connection), reducing total requests to &lt; 10, not serving 700kb hero images, <i>1.22MB</i> embedded youtube players [1], 500kb of other js bloat, 200kb of webfonts, 150kb of boostrap css :&#x2F;<p>the internet is faster than ever, browsers&#x2F;javascript is faster than ever, cross-browser compat is better than ever, computers &amp; servers are faster than ever, yet websites are slower than ever. i literally cannot consume the internet without uMatrix &amp; uBlock Origin. and even with these i have to often give up my privacy by selectively allowing a bunch of required shit from third-party CDNs.<p>no website&#x2F;SPA should take &gt; 2s on a fast connection, (or &gt; 4s on 3g) to be fully loaded. it&#x27;s downright embarrassing. we <i>can</i> and <i>must</i> do better. we have everything we need <i>today</i>.<p>[1] <a href="https:&#x2F;&#x2F;s.ytimg.com&#x2F;yts&#x2F;jsbin&#x2F;player-en_US-vfljAVcXG&#x2F;base.js" rel="nofollow">https:&#x2F;&#x2F;s.ytimg.com&#x2F;yts&#x2F;jsbin&#x2F;player-en_US-vfljAVcXG&#x2F;base.js</a>

So, uh, what does it do? I mean, I can&#x27;t even tell if it&#x27;s a server or client side change.

Talk to some people in the porn industry. They will tell you how important fast pages are. You will also be surprised what they have done to achieve this.

From my experience preconnect is a big improvement for connecting to 3rd party domains. Will also mention that once you have JS deferred, CSS on a 3rd party domain (google fonts) can cause some major slowdowns in terms of the start rendering metrics if using HTTP&#x2F;2 on a slow connection; all the bandwidth is used for the primary domain connection and not used for blocking resources, end result being images get downloaded before external blocking CSS.

The web is way better without JS. Rendering engines could in principal do the same (improved) dependency tracking.

I feel Webpack deserves a mention as it resolves the dependencies at build-time and compiles one (or a few chunked&#x2F;entrybased) assets, hence also solving the problem of too many roundtrips

What are you looking to learn about a CDN?<p>There are many ways to accelerate page speed and, like everything else, it&#x27;s a question of costs and benefits. For most things, some level of technical debt is OK and CDNs even for jQuery are good. Of course, good design and setting things up right is always the best - and the other question is where your site traffic comes from.

Is there a comparison with Google AMP?

&gt; What Polaris does is automatically track all of the interactions between objects, which can number in the thousands for a single page. For example, it notes when one object reads the data in another object, or updates a value in another object. It then uses its detailed log of these interactions to create a “dependency graph” for the page.<p>&gt; Mickens offers the analogy of a travelling businessperson. When you visit one city, you sometimes discover more cities you have to visit before going home. If someone gave you the entire list of cities ahead of time, you could plan the fastest possible route. Without the list, though, you have to discover new cities as you go, which results in unnecessary zig-zagging between far-away cities.<p>What a terrible analogy. Finding a topological sorting is O(|V|+|E|), while the traveling salesman problem is NP-complete.