Working on a Polymer project recently, I got to experiment with serving the content from Firebase Hosting, which now offers support for HTTP/2 and Server Push. Server Push allows content to be delivered speculatively: content is delivered whether or not the user has it already, and in some cases it might not be required at all.
To take advantage of this ability though, I needed to declare the assets I wanted to be pushed down. In my
firebase.json file I declared:
which instructed Firebase to serve the file with the appropriate
Link header. It's possible to do the same when referencing resources via a
link element too:
<link rel="preload" href="foo.js" as="script">
<link rel="preload" href="bar.css" as="style">
If you inspect this site, you'll see that Gatsby does this for me in production. Coupled with HTTP/2 on Cloudfront, visitors should experience a fairly snappy load. Essentially, all assets for a page are transferred in a single round trip, though it does mean if you're a repeat visitor (hi mum!), I'm probably using bandwidth unneccessarily. Addy Osmani suggests that:
Use Push when you know the precise loading order for resources and have a service worker to intercept requests that would cause cached resources to be pushed again.
I guess that's next on my to-do list with Workbox.
<link rel="prefetch" href="/font.woff2">
Prefetch is another resource hint developers can take advantage of. Prefetch tells the browser that a resource will be needed on the next page, or navigation. Resources marked as such will be treated as a much lower-priority fetch than those with
preload, and it's up to the browser to make the call as to whether it will load it or not. I've seen a pattern where a webfont is set to be pre-fetched so that it doesn't block rendering on the first load, but is available right away when a visitor navigates to a new page.
<link rel="preconnect" href="/page2">
Preconnect can be used to 'warm-up' the connection to a resource in order to "eliminate the costly DNS, TCP, and TLS roundtrips from the critical path of the actual request"
All of these techniques can be used in an attempt to improve the performance of your site but it's a trade-off between second-guessing the behaviour of your guests, while not exhausting their bandwidth unneccessarily – especially for those on metered connections.