Tags: performance

39

sparkline

Request mapping

The Request Map Generator is a terrific tool. It’s made by Simon Hearne and uses the WebPageTest API.

You pop in a URL, it fetches the page and maps out all the subsequent requests in a nifty interactive diagram of circles, showing how many requests third-party scripts are themselves generating. I’ve found it to be a very effective way of showing the impact of third-party scripts to people who aren’t interested in looking at waterfall diagrams.

I was wondering… Wouldn’t it be great if this were built into browsers?

We already have a “Network” tab in our developer tools. The purpose of this tab is to show requests coming in. The browser already has all the information it needs to make a diagram of requests in the same that the request map generator does.

In Firefox, there’s a little clock icon in the bottom left corner of the “Network” tab. Clicking that shows a pie-chart view of requests. That’s useful, but I’d love it if there were the option to also see the connected circles that the request map generator shows.

Just a thought.

The Weight of the WWWorld is Up to Us by Patty Toland

It’s Patty Toland’s first time at An Event Apart! She’s from the fantabulous Filament Group. They’re dedicated to making the web work for everyone.

A few years ago, a good friend of Patty’s had a medical diagnosis that required everyone to pull together. Another friend shared an article about how not to say the wrong thing. This is ring theory. In a moment of crisis, the person involved is in the centre. You need to understand where you are in this ring structure, and only ever help and comfort inwards and dump concerns and problems outwards.

At the same time, Patty spent time with her family at the beach. Everyone reads the same books together. There was a book about a platoon leader in Vietnam. 80% of the story was literally a litany of stuff—what everyone was carrying. This was peppered with the psychic and emotional loads that they were carrying.

A month later there was a lot of coverage of Syrian refugees arriving in Europe. People were outraged to see refugees carrying smartphones as though that somehow showed they weren’t in a desperate situation. But smartphones are absolutely a necessity in that situation, and most of the phones were less expensive, lower-end devices. Refugeeinfo.eu was a useful site for people in crisis, but the navigation was designed to require JavaScript.

When people thing about mobile, they think about freedom and mobility. But with that JavaScript decision, the developers piled baggage on to the users.

There was a common assertion that slow networks were a third-world challenge. Remember Facebook’s network challenges? They always talked about new markets in India and Africa. The implication is that this isn’t our problem in, say, Omaha or New York.

Pew Research provided a lot of data back then that showed that this thinking was wrong. Use of cell phones, especially smartphones and tablets, escalated dramatically in the United States. There was a trend towards mobile-only usage. This was in low-income households—about one third of the population. Among 5,400 panelists, 15% did not have a JavaScript-enabled device.

Pew Research provided updated data this year. The research shows an increase in those trends. Half of the population access the web primarily on mobile. The cost of a broadband subscription is too expensive for many people. Sometimes broadband access simply isn’t available.

There’s a term called “the homework gap.” Two thirds of teachers assign broadband-dependent homework, while one third of students have no access to broadband.

At most 37% of people have unlimited data. Most people run out of data on a frequent basis.

Speed also varies wildly. 4G doesn’t really mean anything. The data is all over the place.

This shows that network issues are definitely not just a third world challenge.

On the 25th anniversary of the web, Tim Berners-Lee said the web’s potential was only just beginning to be glimpsed. Everyone has a role to play to ensure that the web serves all of humanity. In his contract for the web, Tim outlined what governments, companies, and users need to do. This reminded Patty of ring theory. The user is at the centre. Designers and developers are in the next circle out. Then there’s the circle of companies. Then there are platforms, browsers, and frameworks. Finally there’s the outer circle of governments.

Are we helping in or dumping in? If you look at the data for the average web page size (2 megabytes), we are definitely dumping in. The size of third-party JavaScript has octupled.

There’s no way for a user to know before clicking a link how big and bloated the page is going to be. Even if they abandon the page load, they’ve still used (and wasted) a lot of data.

Third party scripts—like ads—are really bad at dumping in (to use the ring theory model). The best practices for ads suggest that up to 100 additional HTTP requests is totally acceptable. Unbelievable! It doesn’t matter how performant you’ve made a site when this crap gets piled on top of it.

In 2018, the internet’s data centres alone may already have had the same carbon footprint as all global air travel. This will probably triple in the next seven years. The amount of carbon it takes to train a single AI algorithm is more than the entire life cycle of a car. Then there’s fucking Bitcoin. A single Bitcoin transaction could power 21 US households. It is designed to use—specifically, waste—more and more energy over time.

What should we be doing?

Accessibility should be at the heart of what we build. Plan, test, educate, and advocate. If advocacy doesn’t work, fear can be a motivator. There’s an increase in accessibility lawsuits.

Our websites should be as light as possible. Ask, measure, monitor, and optimise. RequestMap is a great tool for visualising requests. You can see the size and scale of third-party requests. You can also see when images are far, far bigger than they need to be.

Take a critical guide to everything and pare everything down. Set perforance budgets—file size budgets, for example. Optimise images, subset custom fonts, lazyload images and videos, get third-party tools out of the critical path (or out completely), and seek out lighter frameworks.

Test on real devices that real people are using. See Alex Russell’s data on the differences between the kind of devices we use and typical low-end devices. We literally need to stop people in JavaScript.

Push the boundaries. See the amazing work that Adrian Holovaty did with Soundslice. He had to make on-the-fly sheet music generation work on old iPads that musicians like to use. He recommends keeping old devices around to see how poorly your product is working on it.

If you have some power, then your job is to empower somebody else.

—Toni Morrison

Opening up the AMP cache

I have a proposal that I think might alleviate some of the animosity around Google AMP. You can jump straight to the proposal or get some of the back story first…

The AMP format

Google AMP is exactly the kind of framework I’d like to get behind. Unlike most front-end frameworks, its components take a declarative approach—no knowledge of JavaScript required. I think Lea’s excellent Mavo is the only other major framework that takes this inclusive approach. All the configuration happens in markup, and all the styling happens in CSS. Excellent!

But I cannot get behind AMP.

Instead of competing on its own merits, AMP is unfairly propped up by the search engine of its parent company, Google. That makes it very hard to evaluate whether AMP is being used on its own merits. Instead, the evidence suggests that most publishers of AMP pages are doing so because they feel they have to, rather than because they want to. That’s a real shame, because as a library of web components, AMP seems pretty good. But there’s just no way to evaluate AMP-the-format without taking into account AMP-the-ecosystem.

The AMP ecosystem

Google AMP ostensibly exists to make the web faster. Initially the focus was specifically on mobile performance, but that distinction has since fallen by the wayside. The idea is that by using AMP’s web components, your pages will be speedy. Though, as Andy Davies points out, this isn’t always the case:

This is where I get confused… https://independent.co.uk only have an AMP site yet it’s performance is awful from a user perspective - isn’t AMP supposed to prevent this?

See also: Google AMP lowered our page speed, and there’s no choice but to use it:

According to Google’s own Page Speed Insights audit (which Google recommends to check your performance), the AMP version of articles got an average performance score of 87. The non-AMP versions? 95.

Publishers who already have fast web pages—like The Guardian—are still compelled to make AMP versions of their stories because of the search benefits reserved for AMP. As Terence Eden reported from a meeting of the AMP advisory committee:

We heard, several times, that publishers don’t like AMP. They feel forced to use it because otherwise they don’t get into Google’s news carousel — right at the top of the search results.

Some people felt aggrieved that all the hard work they’d done to speed up their sites was for nothing.

The Google AMP team are at pains to point out that AMP is not a ranking factor in search. That’s true. But it is unfairly privileged in other ways. Only AMP pages can appear in the Top Stories carousel …which appears above any other search results. As I’ve said before:

Now, if you were to ask any right-thinking person whether they think having their page appear right at the top of a list of search results would be considered preferential treatment, I think they would say hell, yes! This is the only reason why The Guardian, for instance, even have AMP versions of their content—it’s not for the performance benefits (their non-AMP pages are faster); it’s for that prime real estate in the carousel.

From A letter about Google AMP:

Content that “opts in” to AMP and the associated hosting within Google’s domain is granted preferential search promotion, including (for news articles) a position above all other results.

That’s not the only way that AMP pages get preferential treatment. It turns out that the secret to the speed of AMP pages isn’t the web components. It’s the prerendering.

The AMP cache

If you’ve ever seen an AMP page in a list of search results, you’ll have noticed the little lightning icon. If you’ve ever tapped on that search result, you’ll have noticed that the page loads blazingly fast!

That’s not down to AMP-the-format, alas. That’s down to the fact that the page has been prerendered by Google before you even went to it. If any page were prerendered that way, it would load blazingly fast. But currently, this privilege is reserved for AMP pages only.

If, after tapping through to that AMP page, you looked at the address bar of your browser, you might have noticed something odd. Even though you might have thought you were visiting The Washington Post, or The New York Times, the URL of the (blazingly fast) page you’re looking at is still under Google’s domain. That’s because Google hosts any AMP pages that it prerenders.

Google calls this “the AMP cache”, but it would be better described as “AMP hosting”. The web page sent down the wire is hosted on Google’s domain.

Here’s that AMP letter again:

When a user navigates from Google to a piece of content Google has recommended, they are, unwittingly, remaining within Google’s ecosystem.

Through gritted teeth, I will refer to this as “the AMP cache”, because that’s what everyone else calls it. But make no mistake, Google is hosting—not caching—these pages.

But why host the pages on a Google domain? Why not prerender the original URLs?

Prerendering and privacy

Scott summed up the situation with AMP nicely:

The pitch I think site owners are hearing is: let us host your pages on our domain and we’ll promote them in search results AND preload them so they feel “instant.” To opt-in, build pages using this component syntax.

But perhaps we could de-couple the AMP format from the AMP cache.

That’s what Terence suggests:

My recommendation is that Google stop requiring that organisations use Google’s proprietary mark-up in order to benefit from Google’s promotion.

The AMP letter, too:

Instead of granting premium placement in search results only to AMP, provide the same perks to all pages that meet an objective, neutral performance criterion such as Speed Index.

Scott reiterates:

It’s been said before but it would be so good for the web if pages with a Lighthouse score over say, 90 could get into that top search result area, even if they’re not built using Google’s AMP framework. Feels wrong to have to rebuild/reproduce an already-fast site just for SEO.

This was also what I was calling for. But then Malte pointed out something that stumped me. Privacy.

Here’s the problem…

Let’s say Google do indeed prerender already-fast pages when they’re listed in search results. You, a search user, type something into Google. A list of results come back. Google begins pre-rendering some of them. But you don’t end up clicking through to those pages. Nonetheless, the servers those pages are hosted on have received a GET request coming from a Google search. Those publishers now know that a particular (cookied?) user could have clicked through to their site. That’s very different from knowing when someone has actually arrived at a particular site.

And that’s why Google host all the AMP pages that they prerender. Given the privacy implications of prerendering non-Google URLs, I must admit that I see their point.

Still, it’s a real shame to miss out on the speed benefit of prerendering:

Prerendering AMP documents leads to substantial improvements in page load times. Page load time can be measured in different ways, but they consistently show that prerendering lets users see the content they want faster. For now, only AMP can provide the privacy preserving prerendering needed for this speed benefit.

A modest proposal

Why is Google’s AMP cache just for AMP pages? (Y’know, apart from the obvious answer that it’s in the name.)

What if Google were allowed to host non-AMP pages? Google search could then prerender those pages just like it currently does for AMP pages. There would be no privacy leaks; everything would happen on the same domain—google.com or ampproject.org or whatever—just as currently happens with AMP pages.

Don’t get me wrong: I’m not suggesting that Google should make a 1:1 model of the web just to prerender search results. I think that the implementation would need to have two important requirements:

  1. Hosting needs to be opt-in.
  2. Only fast pages should be prerendered.

Opting in

Currently, by publishing a page using the AMP format, publishers give implicit approval to Google to host that page on Google’s servers and serve up this Google-hosted version from search results. This has always struck me as being legally iffy. I’ve looked in the AMP documentation to try to find any explicit granting of hosting permission (e.g. “By linking to this JavaScript file, you hereby give Google the right to serve up our copies of your content.”), but no luck. So even with the current situation, I think a clear opt-in for hosting would be beneficial.

This could be a meta element. Maybe something like:

<meta name="caches-allowed" content="google">

This would have the nice benefit of allowing comma-separated values:

<meta name="caches-allowed" content="google, yandex">

(The name is just a strawman, by the way—I’m not suggesting that this is what the final implementation would actually look like.)

If not a meta element, then perhaps this could be part of robots.txt? Although my feeling is that this needs to happen on a document-by-document basis rather than site-wide.

Many people will, quite rightly, never want Google—or anyone else—to host and serve up their content. That’s why it’s so important that this behaviour needs to be opt-in. It’s kind of appalling that the current hosting of AMP pages is opt-in-by-proxy-sort-of.

Criteria for prerendering

Which pages should be blessed with hosting and prerendering? The fast ones. That’s sorta the whole point of AMP. But right now, there’s a lot of resentment by people with already-fast websites who quite rightly feel they shouldn’t have to use the AMP format to benefit from the AMP ecosystem.

Page speed is already a ranking factor. It doesn’t seem like too much of a stretch to extend its benefits to hosting and prerendering. As mentioned above, there are already a few possible metrics to use:

  • Page Speed Index
  • Lighthouse
  • Web Page Test

Ah, but what if a page has good score when it’s indexed, but then gets worse afterwards? Not a problem! The version of the page that’s measured is the same version of the page that gets hosted and prerendered. Google can confidently say “This page is fast!” After all, they’re the ones serving up the page.

That does raise the question of how often Google should check back with the original URL to see if it has changed/worsened/improved. The answer to that question is however long it currently takes to check back in on AMP pages:

Each time a user accesses AMP content from the cache, the content is automatically updated, and the updated version is served to the next user once the content has been cached.

Issues

This proposal does not solve the problem with the address bar. You’d still find yourself looking at a page from The Washington Post or The New York Times (or adactio.com) but seeing a completely different URL in your browser. That’s not good, for all the reasons outlined in the AMP letter.

In fact, this proposal could potentially make the situation worse. It would allow even more sites to be impersonated by Google’s URLs. Where currently only AMP pages are bad actors in terms of URL confusion, opening up the AMP cache would allow equal opportunity URL confusion.

What I’m suggesting is definitely not a long-term solution. The long-term solutions currently being investigated are technically tricky and will take quite a while to come to fruition—web packages and signed exchanges. In the meantime, what I’m proposing is a stopgap solution that’s technically a lot simpler. But it won’t solve all the problems with AMP.

This proposal solves one problem—AMP pages being unfairly privileged in search results—but does nothing to solve the other, perhaps more serious problem: the erosion of site identity.

Measuring

Currently, Google can assess whether a page should be hosted and prerendered by checking to see if it’s a valid AMP page. That test would need to be widened to include a different measurement of performance, but those measurements already exist.

I can see how this assessment might not be as quick as checking for AMP validity. That might affect whether non-AMP pages could be measured quickly enough to end up in the Top Stories carousel, which is, by its nature, time-sensitive. But search results are not necessarily as time-sensitive. Let’s start there.

Assets

Currently, AMP pages can be prerendered without fetching anything other than the markup of the AMP page itself. All the CSS is inline. There are no initial requests for other kinds of content like images. That’s because there are no img elements on the page: authors must use amp-img instead. The image itself isn’t loaded until the user is on the page.

If the AMP cache were to be opened up to non-AMP pages, then any content required for prerendering would also need to be hosted on that same domain. Otherwise, there’s privacy leakage.

This definitely introduces an extra level of complexity. Paths to assets within the markup might need to be re-written to point to the Google-hosted equivalents. There would almost certainly need to be a limit on the number of assets allowed. Though, for performance, that’s no bad thing.

Make no mistake, figuring out what to do about assets—style sheets, scripts, and images—is very challenging indeed. Luckily, there are very smart people on the Google AMP team. If that brainpower were to focus on this problem, I am confident they could solve it.

Summary

  1. Prerendering of non-Google URLs is problematic for privacy reasons, so Google needs to be able to host pages in order to prerender them.
  2. Currently, that’s only done for pages using the AMP format.
  3. The AMP cache—and with it, prerendering—should be decoupled from the AMP format, and opened up to other fast web pages.

There will be technical challenges, but hopefully nothing insurmountable.

I honestly can’t see what Google have to lose here. If their goal is genuinely to reward fast pages, then opening up their AMP cache to fast non-AMP pages will actively encourage people to make fast web pages (without having to switch over to the AMP format).

I’ve deliberately kept the details vague—what the opt-in should look like; what the speed measurement should be; how to handle assets—I’m sure smarter folks than me can figure that stuff out.

I would really like to know what other people think about this proposal. Obviously, I’d love to hear from members of the Google AMP team. But I’d also love to hear from publishers. And I’d very much like to know what people in the web performance community think about this. (Write a blog post and send me a webmention.)

What am I missing here? What haven’t I thought of? What are the potential pitfalls (and are they any worse than the current acrimonious situation with Google AMP)?

I would really love it if someone with a fast website were in a position to say, “Hey Google, I’m giving you permission to host this page so that it can be prerendered.”

I would really love it if someone with a slow website could say, “Oh, shit! We’d better make our existing website faster or Google won’t host our pages for prerendering.”

And I would dearly love to finally be able to embrace AMP-the-format with a clear conscience. But as long as prerendering is joined at the hip to the AMP format, the injustice of the situation only harms the AMP project.

Google, open up the AMP cache.

Server Timing

Harry wrote a really good article all about the performance measurement Time To First Byte. Time To First Byte: What It Is and Why It Matters:

While a good TTFB doesn’t necessarily mean you will have a fast website, a bad TTFB almost certainly guarantees a slow one.

Time To First Byte has been the chink in my armour over at thesession.org, especially on the home page. Every time I ran Lighthouse, or some other performance testing tool, I’d get a high score …with some points deducted for taking too long to get that first byte from the server.

Harry’s proposed solution is to set up some Server Timing headers:

With a little bit of extra work spent implementing the Server Timing API, we can begin to measure and surface intricate timings to the front-end, allowing web developers to identify and debug potential bottlenecks previously obscured from view.

I rememberd that Drew wrote an excellent article on Smashing Magazine last year called Measuring Performance With Server Timing:

The job of Server Timing is not to help you actually time activity on your server. You’ll need to do the timing yourself using whatever toolset your backend platform makes available to you. Rather, the purpose of Server Timing is to specify how those measurements can be communicated to the browser.

He even provides some PHP code, which I was able to take wholesale and drop into the codebase for thesession.org. Then I was able to put start/stop points in my code for measuring how long some operations were taking. Then I could output the results of these measurements into Server Timing headers that I could inspect in the “Network” tab of a browser’s dev tools (Chrome is particularly good for displaying Server Timing, so I used that while I was conducting this experiment).

I started with overall database requests. Sure enough, that was where most of the time in time-to-first-byte was being spent.

Then I got more granular. I put start/stop points around specific database calls. By doing this, I was able to zero in on which operations were particularly costly. Once I had done that, I had to figure out how to make the database calls go faster.

Spoiler: I did it by adding an extra index on one particular table. It’s almost always indexes, in my experience, that make the biggest difference to database performance.

I don’t know why it took me so long to get around to messing with Server Timing headers. It has paid off in spades. I wish I had done it sooner.

And now thesession.org is positively zipping along!

Navigation preloads in service workers

There’s a feature in service workers called navigation preloads. It’s relatively recent, so it isn’t supported in every browser, but it’s still well worth using.

Here’s the problem it solves…

If someone makes a return visit to your site, and the service worker you installed on their machine isn’t active yet, the service worker boots up, and then executes its instructions. If those instructions say “fetch the page from the network”, then you’re basically telling the browser to do what it would’ve done anyway if there were no service worker installed. The only difference is that there’s been a slight delay because the service worker had to boot up first.

  1. The service worker activates.
  2. The service worker fetches the file.
  3. The service worker does something with the response.

It’s not a massive performance hit, but it’s still a bit annoying. It would be better if the service worker could boot up and still be requesting the page at the same time, like it would do if no service worker were present. That’s where navigation preloads come in.

  1. The service worker activates while simultaneously requesting the file.
  2. The service worker does something with the response.

Navigation preloads—like the name suggests—are only initiated when someone navigates to a URL on your site, either by following a link, or a bookmark, or by typing a URL directly into a browser. Navigation preloads don’t apply to requests made by a web page for things like images, style sheets, and scripts. By the time a request is made for one of those, the service worker is already up and running.

To enable navigation preloads, call the enable() method on registration.navigationPreload during the activate event in your service worker script. But first do a little feature detection to make sure registration.navigationPreload exists in this browser:

if (registration.navigationPreload) {
  addEventListener('activate', activateEvent => {
    activateEvent.waitUntil(
      registration.navigationPreload.enable()
    );
  });
}

If you’ve already got event listeners on the activate event, that’s absolutely fine: addEventListener isn’t exclusive—you can use it to assign multiple tasks to the same event.

Now you need to make use of navigation preloads when you’re responding to fetch events. So if your strategy is to look in the cache first, there’s probably no point enabling navigation preloads. But if your default strategy is to fetch a page from the network, this will help.

Let’s say your current strategy for handling page requests looks like this:

addEventListener('fetch', fetchEvent => {
  const request = fetchEvent.request;
  if (request.headers.get('Accept').includes('text/html')) {
    fetchEvent.respondWith(
      fetch(request)
      .then( responseFromFetch => {
        // maybe cache this response for later here.
        return responseFromFetch;
      })
      .catch( fetchError => {
        return caches.match(request)
        .then( responseFromCache => {
          return responseFromCache || caches.match('/offline');
        });
      })
    );
  }
});

That’s a fairly standard strategy: try the network first; if that doesn’t work, try the cache; as a last resort, show an offline page.

It’s that first step (“try the network first”) that can benefit from navigation preloads. If a preload request is already in flight, you’ll want to use that instead of firing off a new fetch request. Otherwise you’re making two requests for the same file.

To find out if a preload request is underway, you can check for the existence of the preloadResponse promise, which will be made available as a property of the fetch event you’re handling:

fetchEvent.preloadResponse

If that exists, you’ll want to use it instead of fetch(request).

if (fetchEvent.preloadResponse) {
  // do something with fetchEvent.preloadResponse
} else {
  // do something with fetch(request)
}

You could structure your code like this:

addEventListener('fetch', fetchEvent => {
  const request = fetchEvent.request;
  if (request.headers.get('Accept').includes('text/html')) {
    if (fetchEvent.preloadResponse) {
      fetchEvent.respondWith(
        fetchEvent.preloadResponse
        .then( responseFromPreload => {
          // maybe cache this response for later here.
          return responseFromPreload;
        })
        .catch( preloadError => {
          return caches.match(request)
          .then( responseFromCache => {
            return responseFromCache || caches.match('/offline');
          });
        })
      );
    } else {
      fetchEvent.respondWith(
        fetch(request)
        .then( responseFromFetch => {
          // maybe cache this response for later here.
          return responseFromFetch;
        })
        .catch( fetchError => {
          return caches.match(request)
          .then( responseFromCache => {
            return responseFromCache || caches.match('/offline');
          });
        })
      );
    }
  }
});

But that’s not very DRY. Your logic is identical, regardless of whether the response is coming from fetch(request) or from fetchEvent.preloadResponse. It would be better if you could minimise the amount of duplication.

One way of doing that is to abstract away the promise you’re going to use into a variable. Let’s call it retrieve. If a preload is underway, we’ll assign it to that variable:

let retrieve;
if (fetchEvent.preloadResponse) {
  retrieve = fetchEvent.preloadResponse;
}

If there is no preload happening (or this browser doesn’t support it), assign a regular fetch request to the retrieve variable:

let retrieve;
if (fetchEvent.preloadResponse) {
  retrieve = fetchEvent.preloadResponse;
} else {
  retrieve = fetch(request);
}

If you like, you can squash that into a ternary operator:

const retrieve = fetchEvent.preloadResponse ? fetchEvent.preloadResponse : fetch(request);

Use whichever syntax you find more readable.

Now you can apply the same logic, regardless of whether retrieve is a preload navigation or a fetch request:

addEventListener('fetch', fetchEvent => {
  const request = fetchEvent.request;
  if (request.headers.get('Accept').includes('text/html')) {
    const retrieve = fetchEvent.preloadResponse ? fetchEvent.preloadResponse : fetch(request);
    fetchEvent.respondWith(
      retrieve
      .then( responseFromRetrieve => {
        // maybe cache this response for later here.
       return responseFromRetrieve;
      })
      .catch( fetchError => {
        return caches.match(request)
        .then( responseFromCache => {
          return responseFromCache || caches.match('/offline');
        });
      })
    );
  }
});

I think that’s the least invasive way to update your existing service worker script to take advantage of navigation preloads.

Like I said, preload navigations can give a bit of a performance boost if you’re using a network-first strategy. That’s what I’m doing here on adactio.com and on thesession.org so I’ve updated their service workers to take advantage of navigation preloads. But on Resilient Web Design, which uses a cache-first strategy, there wouldn’t be much point enabling navigation preloads.

Jeff Posnick made this point in his write-up of bringing service workers to Google search:

Adding a service worker to your web app means inserting an additional piece of JavaScript that needs to be loaded and executed before your web app gets responses to its requests. If those responses end up coming from a local cache rather than from the network, then the overhead of running the service worker is usually negligible in comparison to the performance win from going cache-first. But if you know that your service worker always has to consult the network when handling navigation requests, using navigation preload is a crucial performance win.

Oh, and those browsers that don’t yet support navigation preloads? No problem. It’s a progressive enhancement. Everything still works just like it did before. And having a service worker on your site in the first place is itself a progressive enhancement. So enabling navigation preloads is like a progressive enhancement within a progressive enhancement. It’s progressive enhancements all the way down!

By the way, if all of this service worker stuff sounds like gibberish, but you wish you understood it, I think my book, Going Offline, will prove quite valuable.

Timing out

Service workers are great for creating a good user experience when someone is offline. Heck, the book I wrote about service workers is literally called Going Offline.

But in some ways, the offline experience is relatively easy to handle. It’s a binary situation; either you’re online or you’re offline. What’s more challenging—and probably more common—is the situation that Jake calls Lie-Fi. That’s when technically you’ve got a network connection …but it’s a shitty connection, like one bar of mobile signal. In that situation, because there’s technically a connection, the user gets a slow frustrating experience. Whatever code you’ve got in your service worker for handling offline situations will never get triggered. When you’re handling fetch events inside a service worker, there’s no automatic time-out.

But you can make one.

That’s what I’ve done recently here on adactio.com. Before showing you what I added to my service worker script to make that happen, let me walk you through my existing strategy for handling offline situations.

Service worker strategies

Alright, so in my service worker script, I’ve got a block of code for handling requests from fetch events:

addEventListener('fetch', fetchEvent => {
        const request = fetchEvent.request;
    // Do something with this request.
});

I’ve got two strategies in my code. One is for dealing with requests for pages:

if (request.headers.get('Accept').includes('text/html')) {
    // Code for handling page requests.
}

By adding an else clause I can have a different strategy for dealing with requests for anything else—images, style sheets, scripts, and so on:

if (request.headers.get('Accept').includes('text/html')) {
    // Code for handling page requests.
} else {
    // Code for handling everthing else.
}

For page requests, I’m going to try to go the network first:

fetchEvent.respondWith(
    fetch(request)
    .then( responseFromFetch => {
        return responseFromFetch;
    })

My logic is:

When someone requests a page, try to fetch it from the network.

If that doesn’t work, we’re in an offline situation. That triggers the catch clause. That’s where I have my offline strategy: show a custom offline page that I’ve previously cached (during the install event):

.catch( fetchError => {
    return caches.match('/offline');
})

Now my logic has been expanded to this:

When someone requests a page, try to fetch it from the network, but if that doesn’t work, show a custom offline page instead.

So my overall code for dealing with requests for pages looks like this:

if (request.headers.get('Accept').includes('text/html')) {
    fetchEvent.respondWith(
        fetch(request)
        .then( responseFromFetch => {
            return responseFromFetch;
        })
        .catch( fetchError => {
            return caches.match('/offline');
        })
    );
}

Now I can fill in the else statement that handles everything else—images, style sheets, scripts, and so on. Here my strategy is different. I’m looking in my caches first, and I only fetch the file from network if the file can’t be found in any cache:

caches.match(request)
.then( responseFromCache => {
    return responseFromCache || fetch(request);
})

Here’s all that fetch-handling code put together:

addEventListener('fetch', fetchEvent => {
    const request = fetchEvent.request;
    if (request.headers.get('Accept').includes('text/html')) {
        fetchEvent.respondWith(
            fetch(request)
            .then( responseFromFetch => {
                return responseFromFetch;
            })
            .catch( fetchError => {
                return caches.match('/offline');
            })
        );
    } else {
        caches.match(request)
        .then( responseFromCache => {
            return responseFromCache || fetch(request);
        })
    }
});

Good.

Cache as you go

Now I want to introduce an extra step in the part of the code where I deal with requests for pages. Whenever I fetch a page from the network, I’m going to take the opportunity to squirrel it away in a cache. I’m calling that cache “pages”. I’m imaginative like that.

fetchEvent.respondWith(
    fetch(request)
    .then( responseFromFetch => {
        const copy = responseFromFetch.clone();
        try {
            fetchEvent.waitUntil(
                caches.open('pages')
                .then( pagesCache => {
                    return pagesCache.put(request, copy);
                })
            )
        } catch(error) {
            console.error(error);
        }
        return responseFromFetch;
    })

You’ll notice that I can’t put the response itself (responseFromCache) into the cache. That’s a stream that I only get to use once. Instead I need to make a copy:

const copy = responseFromFetch.clone();

That’s what gets put in the pages cache:

fetchEvent.waitUntil(
    caches.open('pages')
    .then( pagesCache => {
        return pagesCache.put(request, copy);
    })
)

Now my logic for page requests has an extra piece to it:

When someone requests a page, try to fetch it from the network and store a copy in a cache, but if that doesn’t work, show a custom offline page instead.

Here’s my updated fetch-handling code:

addEventListener('fetch', fetchEvent => {
    const request = fetchEvent.request;
    if (request.headers.get('Accept').includes('text/html')) {
        fetchEvent.respondWith(
            fetch(request)
            .then( responseFromFetch => {
                const copy = responseFromFetch.clone();
                try {
                    fetchEvent.waitUntil(
                        caches.open('pages')
                        .then( pagesCache => {
                            return pagesCache.put(request, copy);
                        })
                    )
                } catch(error) {
                    console.error(error);
                }
                return responseFromFetch;
            })
            .catch( fetchError => {
                return caches.match('/offline');
            })
        );
    } else {
        caches.match(request)
        .then( responseFromCache => {
            return responseFromCache || fetch(request);
        })
    }
});

I call this the cache-as-you-go pattern. The more pages someone views on my site, the more pages they’ll have cached.

Now that there’s an ever-growing cache of previously visited pages, I can update my offline fallback. Currently, I reach straight for the custom offline page:

.catch( fetchError => {
    return caches.match('/offline');
})

But now I can try looking for a cached copy of the requested page first:

.catch( fetchError => {
    caches.match(request)
    .then( responseFromCache => {
        return responseFromCache || caches.match('/offline');
    })
});

Now my offline logic is expanded:

When someone requests a page, try to fetch it from the network and store a copy in a cache, but if that doesn’t work, first look for an existing copy in a cache, and otherwise show a custom offline page instead.

I can also access this ever-growing cache of pages from my custom offline page to show people which pages they can revisit, even if there’s no internet connection.

So far, so good. Everything I’ve outlined so far is a good robust strategy for handling offline situations. Now I’m going to deal with the lie-fi situation, and it’s that cache-as-you-go strategy that sets me up nicely.

Timing out

I want to throw this addition into my logic:

When someone requests a page, try to fetch it from the network and store a copy in a cache, but if that doesn’t work, first look for an existing copy in a cache, and otherwise show a custom offline page instead (but if the request is taking too long, try to show a cached version of the page).

The first thing I’m going to do is rewrite my code a bit. If the fetch event is for a page, I’m going to respond with a promise:

if (request.headers.get('Accept').includes('text/html')) {
    fetchEvent.respondWith(
        new Promise( resolveWithResponse => {
            // Code for handling page requests.
        })
    );
}

Promises are kind of weird things to get your head around. They’re tailor-made for doing things asynchronously. You can set up two parameters; a success condition and a failure condition. If the success condition is executed, then we say the promise has resolved. If the failure condition is executed, then the promise rejects.

In my re-written code, I’m calling the success condition resolveWithResponse (and I haven’t bothered with a failure condition, tsk, tsk). I’m going to use resolveWithResponse in my promise everywhere that I used to have a return statement:

addEventListener('fetch', fetchEvent => {
    const request = fetchEvent.request;
    if (request.headers.get('Accept').includes('text/html')) {
        fetchEvent.respondWith(
            new Promise( resolveWithResponse => {
                fetch(request)
                .then( responseFromFetch => {
                    const copy = responseFromFetch.clone();
                    try {
                        fetchEvent.waitUntil(
                            caches.open('pages')
                            then( pagesCache => {
                                return pagesCache.put(request, copy);
                            })
                        )
                    } catch(error) {
                        console.error(error);
                    }
                    resolveWithResponse(responseFromFetch);
                })
                .catch( fetchError => {
                    caches.match(request)
                    .then( responseFromCache => {
                        resolveWithResponse(
                            responseFromCache || caches.match('/offline')
                        );
                    })
                })
            })
        );
    } else {
        caches.match(request)
        .then( responseFromCache => {
            return responseFromCache || fetch(request);
        })
    }
});

By itself, rewriting my code as a promise doesn’t change anything. Everything’s working the same as it did before. But now I can introduce the time-out logic. I’m going to put this inside my promise:

const timer = setTimeout( () => {
    caches.match(request)
    .then( responseFromCache => {
        if (responseFromCache) {
            resolveWithResponse(responseFromCache);
        }
    })
}, 3000);

If a request takes three seconds (3000 milliseconds), then that code will execute. At that point, the promise attempts to resolve with a response from the cache instead of waiting for the network. If there is a cached response, that’s what the user now gets. If there isn’t, then the wait continues for the network.

The last thing left for me to do is cancel the countdown to timing out if a network response does return within three seconds. So I put this in the then clause that’s triggered by a successful network response:

clearTimeout(timer);

I also add the clearTimeout statement to the catch clause that handles offline situations. Here’s the final code:

addEventListener('fetch', fetchEvent => {
    const request = fetchEvent.request;
    if (request.headers.get('Accept').includes('text/html')) {
        fetchEvent.respondWith(
            new Promise( resolveWithResponse => {
                const timer = setTimeout( () => {
                    caches.match(request)
                    .then( responseFromCache => {
                        if (responseFromCache) {
                            resolveWithResponse(responseFromCache);
                        }
                    })
                }, 3000);
                fetch(request)
                .then( responseFromFetch => {
                    clearTimeout(timer);
                    const copy = responseFromFetch.clone();
                    try {
                        fetchEvent.waitUntil(
                            caches.open('pages')
                            then( pagesCache => {
                                return pagesCache.put(request, copy);
                            })
                        )
                    } catch(error) {
                        console.error(error);
                    }
                    resolveWithResponse(responseFromFetch);
                })
                .catch( fetchError => {
                    clearTimeout(timer);
                    caches.match(request)
                    .then( responseFromCache => {
                        resolveWithResponse(
                            responseFromCache || caches.match('/offline')
                        );
                    })
                })
            })
        );
    } else {
        caches.match(request)
        .then( responseFromCache => {
            return responseFromCache || fetch(request)
        })
    }
});

That’s the JavaScript translation of this logic:

When someone requests a page, try to fetch it from the network and store a copy in a cache, but if that doesn’t work, first look for an existing copy in a cache, and otherwise show a custom offline page instead (but if the request is taking too long, try to show a cached version of the page).

For everything else, try finding a cached version first, otherwise fetch it from the network.

Pros and cons

As with all service worker enhancements to a website, this strategy will do absolutely nothing for first-time visitors. If you’ve never visited my site before, you’ve got nothing cached. But the more you return to the site, the more your cache is primed for speedy retrieval.

I think that serving up a cached copy of a page when the network connection is flaky is a pretty good strategy …most of the time. If we’re talking about a blog post on this site, then sure, there won’t be much that the reader is missing out on—a fixed typo or ten; maybe some additional webmentions at the end of a post. But if we’re talking about the home page, then a reader with a flaky network connection might think there’s nothing new to read when they’re served up a stale version.

What I’d really like is some way to know—on the client side—whether or not the currently-loaded page came from a cache or from a network. Then I could add some kind of interface element that says, “Hey, this page might be stale—click here if you want to check for a fresher version.” I’d also need some way in the service worker to identify any requests originating from that interface element and make sure they always go out to the network.

I think that should be doable somehow. If you can think of a way to do it, please share it. Write a blog post and send me the link.

But even without the option to over-ride the time-out, I’m glad that I’m at least doing something to handle the lie-fi situation. Perhaps I should write a sequel to Going Offline called Still Online But Only In Theory Because The Connection Sucks.

Mobile Planet by Luke Wroblewski

It’s the afternoon of day two of An Event Apart in Seattle. The mighty green one, Luke Wroblewski, is here to deliver a talk called Mobile Planet:

With 3.5 billion active smartphones on Earth, we’re now faced with the challenges and opportunities of designing planet-scale software. Through a data-informed, big-picture walk-through of our mobile planet, Luke will dig into how people use computing devices today and how the design of our products needs to adapt to this reality. He’ll cover key issues like app on-boarding and performance in enough detail to give you clear ways to improve first time and subsequent use of your mobile apps and sites.

Luke has been working on figuring out hardware and software for years. He looks at a lot of data. The more we understand how people use technology in their daily lives, the better we can design for them.

Earth is the third planet from the sun, and the only place that we know of that harbours life. Our population is at about 7.7 billion people. There are about 5.6 billion people in our addressable market (people over 14 years old). There are already 5 billion mobile subscribers in there. That’s interesting, but which of those devices are modern smartphones? There are about 3.6 billion active smartphones. Compare that to about 1.3 billion active personal computers—the vast majority of them Windows devices (about 1.2 billion). Over the next four or five years, we’ll have about 5 billion smartphone users and a global population of 8 billion.

The point is that we can reach a significant proportion of the human species. The diversity of our species makes it challenging to design for everyone.

Let’s take a closer look at these 3.6 billion active smartphones. About 25% of them are iOS devices. 75% of them are Android. Bear in mind that these are active devices—what’s actually being used. That’s different to shipping devices. Apple ships 15% of smartphone, and Android ships 85%, but the iOS devices tend to have longer lifespans (around 2 years for Android; around 4 years for iOS).

The UK has 82% smartphone penetration. Compare that to India, where it’s 27%. There’s room to grow.

Everywhere you look, the growth of these devices has led to a shift of digital things overtaking analogue. Shopping, advertising, music, you name it. We’ve seen enough of these transitions happen, that we should be prepared for it.

So there are lots of smartphones, with basically two major operating systems. But how are people using these devices?

In the US, adults spend about 2.3 - 3.5 hours per day on their mobile devices. Let’s call it an even 3 hours. That’s a lot of time. Where does that time come from? Interestingly, as time spent on mobile devices has surged, time spent on other media has only slowly declined. So mobile is additive. It’s contributing to more time spent on the internet rather than taking it away from existing screen time.

Next question: what the hell are people doing during those 3 hours per day on smartphones? Native apps get about 169 minutes of time compared to only 11 minutes on the web. There are about 2 million native apps on Apple, and about 2 million native apps on Android. But although people have a lot of apps, people only use about half them. Remember folks, downloads does not equal usage. Most apps don’t make it past the first opening. Only a third make it past being opened ten times.

Because people spend so much time and energy on these apps, and given the abysmal abandonment, people start freaking out about “engagement.” So what do they reach for? Push notifications. Either that or onboarding.

Push notifications. The worst. I mean, they do succeed in getting your attention: push notifications do increase the amount of time spent in your app …but there’s a human cost.

Let’s look at app onboarding. Take Flickr, for example. It walks through some of the features and benefits of the service. But it doesn’t actually help you much. It’s a list of marketing slogans. So why do people reach for onboarding?

If you just drop people into an interface and talk to them about it, they’ll say things like “I don’t know what to do. I’m lost.” The Intuit team heard this from people using their app. They reached for onboarding to solve the problem. They created guided tutorials and intro tours. Turns out that nobody would read these screens and everyone would try to skip them. What the hell, people!?

So they try in-context help, with a cute cartoon robot to explain the features. Or they scribble Einstein’s equations over the interface. Test this. People respond with “Please make it stop.”

They decided to try something simpler: one tip that calls out a good first step. That worked.

Vevo used to have an intro tour. Most people were swiping through without reading. They experimented with not running the tour. They got a 10% increase in log-ins and a 6% increase in sign-ups.

Vevo got rid of their tour, but left the sign-in/registration step. You can’t remove that, right?

Well, Hotel Tonight experimented with not doing registration. Signing up was confusing people—it’s Hotels Tonight, not Accounts Today. When they got rid of accounts, they saw a 15% increase in conversions.

Ruthlessly edit.

Google Photos used to have an in-depth on-boarding experience. First they got rid of the animation. Then the start-up screen. Then the animated tutorial. Each time they removed something, conversion went up. All that was left from the original onboarding was a half screen with one option to turn on auto-backup.

Get to your product value as fast as possible. Of course that requires you to know what your core value is. And that’s not easy to figure out.

Google Maps went through a similar reduction, removing intro screens and explanations. Now they just drop you into the map.

It’s not “get rid of everything”. It’s “get rid of everything that gets in the way of the core user action.”

Going back to the Intuit example, that’s exactly what they did in the end. That one initial tip was for the core action.

But it’s worth discussing how to present this kind of thing. If you have to overlay a tooltip for an important UI feature, maybe that UI feature should have a clearer affordance. People treat overlays as annoyances. People ignore or dismiss overlays when they’re focused on a task. It’s like an instinct to get rid of them. So if you put something useful or valuable there, it’s gone.

The core part of your application should feel like the core part of your application. It’s tough because stakeholders want to make things “pop.” We throw contrast, colour, and animation at things. But when something sticks out from the UI, people ignore it. Integrate the core action into the product UI. When elements feel foreign to a product UI, they are at best ignored, or at worst dismissed.

These is why cohesive design matters. It’s not about consistency. It’s about feeling integrated. In many cases, consistency can be counter-productive.

Some principles for successful onboarding:

  1. Get to to the product value as fast as possible. Grubhub needs your address. Pinterest needs your interests.
  2. Get rid of everything that doesn’t lead to that product value. Ruthlessly edit. Remove all friction that distracts the user from experiencing product value.
  3. Don’t be afraid to educate contextually. But do so with integrated UI.

Luke talked a lot about what’s happening in mobile apps, and mentioned that the mobile web only gets 11 minutes to the native’s 169. But let’s dive into this, because people sometimes think that a “mobile strategy” comes down to picking between these two. 50% of those 169 minutes are spent in your most used app (Facebook). 78% of the time is spent in the top 5 apps. Now the mobile web doesn’t look so bad. It turns out you can get people to a mobile web experience much, much faster than to a native one. The audience size is much, much, much higher on the web (although people will do more in a dedicated native app). So strategically both are useful—the web can attract people to native.

Back to our planet, and those 3 hours of usage on smartphones every day. People unlock their phones around 80 times a day. The average time people sleep is about 8 hours. So for every 12 waking minutes, you’re unlocking your phone. Given this frequency, it’s unsurprising that most sessions are very short—most under 30 seconds.

Given that, if things are slow, you’re going to really, really, really hate it. Waiting for slow pages to load is what really pisses people off.

The cognitive load and stress of waiting for slow pages is worse than waiting in line in a store, or watching a horror movie. That’s an industry that’s all about stressing people out by design! But experiencing mobile delays is more stressful! Probably because people aren’t watching horror movies every 12 minutes.

Because mobile delays are such a big deal, many mobile apps reach for loading spinners. But Luke saw that adding a spinner to his product increased complaints of slow loading times. Of course! The spinner is explicitly telling people, “Hey, we’re slow.”

So the switched to skeleton screens. This should feel like something is always happening. Focus on the progress, not the progress indicator. Occupied time feels shorter than unoccupied time.

A lot of people have implemented skeleton screen, but without the progressive loading. Swapping out a skeleton screen to a completely different UI all at once doesn’t help. The skeleton screens should represent the real content.

This is a lot of work; figuring how to prioritise what to load first. Luke isn’t talking about the techical side here, but the user’s experience. Investing in getting this right makes a lot of sense.

Let’s look a little closer at this number: people interacting with their phones 80 times a day. The average user touches the device 2,617 times a day. A power user touches the device over 5,000 times a day. Most touches are within one app.

90% of the touches are dealing with one thumb. Young people tend to operate with one hand. For older people, it’s more like 60%.

This is why your interface targets need to work for the thumb.

On phones, 90% of the time you’re dealing with portrait mode. Things at the top of the screen on larger devices are hard to reach. Core actions gravitate to the bottom of the screen.

Opera Touch is a new browser designed specifically for one-handed use. The Palm Pre’s WebOS was also about one-hand usage. Now that’s how iOS and Android work: swiping up from the bottom.

So mobile usage is:

  • One-handed/thumb.
  • In portrait mode on large screens.
  • Design accordingly.

What’s next? What do we need to be aware of so we don’t get caught with our pants down?

We can use the product lifecycle chart to figure this out. There’s an emergent phase, then a growth phase, then consolidation in a mature market, and then that gets disrupted and becomes a declining market.

  • Mobile devices—hand computers—are in a mature consolidated market.
  • Desktop and laptop computers are in a declining market.
  • Wrist computers and voice computers are in a growth market.

Small screens get used more frequently, but for shorter periods of time than large screens. Wrist and voice computers are figuring out what their core offerings are.

In the emergent category, it’s all about exploration. We have no idea how things will turn out. We just don’t know. But we do know that we are now designing for lots and lots of different devices.

For today, though, focusing on mobile is still a pretty good idea.

To summarise:

  • It’s a mobile planet.
  • Understanding real world usage helps you design.
  • Prep for what’s next

Move Fast and Don’t Break Things by Scott Jehl

Scott Jehl is speaking at An Event Apart in Seattle—yay! His talk is called Move Fast and Don’t Break Things:

Performance is a high priority for any site of scale today, but it can be easier to make a site fast than to keep it that way. As a site’s features and design evolves, its performance is often threatened for a number of reasons, making it hard to ensure fast, resilient access to services. In this session, Scott will draw from real-world examples where business goals and other priorities have conflicted with page performance, and share some strategies and practices that have helped major sites overcome those challenges to defend their speed without compromises.

The title is a riff on the “move fast and break things” motto, which comes from a more naive time on the web. But Scott finds part of it relatable. Things break. We want to move fast without breaking things.

This is a performance talk, which is another kind of moving fast. Scott starts with a brief history of not breaking websites. He’s been chipping away at websites for 20 years now. Remember Positioning Is Everything? How about Quirksmode? That one's still around.

In the early days, building a website that was "not broken" was difficult, but it was difficult for different reasons. We were focused on consistency. We had deal with differences between browsers. There were two ways of dealing with browsers: browser detection and feature detection.

The feature-based approach was more sustainable but harder. It fits nicely with the practice of progressive enhancement. It's a good mindset for dealing with the explosion of devices that kicked off later. Touch screens made us rethink our mouse and hover-centric matters. That made us realise how much keyboard-driven access mattered all along.

Browsers exploded too. And our data networks changed. With this explosion of considerations, it was clear that our early ideas of “not broken” didn’t work. Our notion of what constituted “not broken” was itself broken. Consistency just doesn’t cut it.

But there was a comforting part to this too. It turned out that progressive enhancement was there to help …even though we didn’t know what new devices were going to appear. This is a recurring theme throughout Scott’s career. So given all these benefits of progressive enhancement, it shouldn’t be surprising that it turns out to be really good for performance too. If you practice progressive enhancement, you’re kind of a performance expert already.

People started talking about new performance metrics that we should care about. We’ve got new tools, like Page Speed Insights. It gives tangible advice on how to test things. Web Page Test is another great tool. Once you prove you’re a human, Web Page Test will give you loads of details on how a page loaded. And you get this great visual timeline.

This is where we can start to discuss the metrics we want to focus on. Traditionally, we focused on file size, which still matters. But for goal-setting, we want to focus on user-perceived metrics.

First Meaningful Content. It’s about how soon appears to be useful to a user. Progressive enhancement is a perfect match for this! When you first make request to a website, it’s usually for a web page. But to render that page, it might need to request more files like CSS or JavaScript. All of this adds up. From a user perspective, if the HTML is downloaded, but the browser can’t render it, that’s broken.

The average time for this on the web right now is around six seconds. That’s broken. The render blockers are the problem here.

Consider assets like scripts. Can you get the browser to load them without holding up the rendering of the page? If you can add async or defer to a script element in the head, you should do that. Sometimes that’s not an option though.

For CSS, it’s tricky. We’ve delivered the HTML that we need but we’ve got to wait for the CSS before rendering it. So what can you bundle into that initial payload?

You can user server push. This is a new technology that comes with HTTP2. H2, as it’s called, is very performance-focused. Just turning on H2 will probably make your site faster. Server push allows the server to send files to the browser before the browser has even asked for them. You can do this with directives in Apache, for example. You could push CSS whenever an HTML file is requested. But we need to be careful not to go too far. You don’t want to send too much.

Server push is great in moderation. But it is new, and it may not even be supported by your server.

Another option is to inline CSS (well, actually Scott, this is technically embedding CSS). It’s great for first render, but isn’t it wasteful for caching? Scott has a clever pattern that uses the Cache API to grab the contents of the inlined CSS and put a copy of its contents into the cache. Then it’s ready to be served up by a service worker.

By the way, this isn’t just for CSS. You could grab the contents of inlined SVGs and create cached versions for later use.

So inlining CSS is good, but again, in moderation. You don’t want to embed anything bigger than 15 or 20 kilobytes. You might want separate out the critical CSS and only embed that on first render. You don’t need to go through your CSS by hand to figure out what’s critical—there are tools that to do this that integrate with your build process. Embed that critical CSS into the head of your document, and also start preloading the full CSS. Here’s a clever technique that turns a preload link into a stylesheet link:

<link rel="preload" href="site.css" as="style" onload="this.rel='stylesheet'">

Also include this:

<noscript><link rel="stylesheet" href="site.css"></noscript>

You can also optimise for return visits. It’s all about the cache.

In the past, we might’ve used a cookie to distinguish a returning visitor from a first-time visitor. But cookies kind of suck. Here’s something that Scott has been thinking about: service workers can intercept outgoing requests. A service worker could send a header that matches the current build of CSS. On the server, we can check for this header. If it’s not the latest CSS, we can server push the latest version, or inline it.

The neat thing about service workers is that they have to install before they take over. Scott makes use of this install event to put your important assets into a cache. Only once that is done to we start adding that extra header to requests.

Watch out for an article on the Filament Group blog on this technique!

With performance, more weight doesn’t have to mean more wait. You can have a heavy page that still appears to load quickly by altering the prioritisation of what loads first.

Web pages are very heavy now. There’s a real cost to every byte. Tim’s WhatDoesMySiteCost.com shows that the CNN home page costs almost fifty cents to load for someone in America!

Time to interactive. This is is the time before a user can use what’s on the screen. The issue is almost always with JavaScript. The page looks usable, but you can’t use it yet.

Addy Osmani suggests we should get to interactive in under five seconds on a 3G network on a median mobile device. Your iPhone is not a median mobile device. A typical phone takes six seconds to process a megabyte of JavaScript after it has downloaded. So even if the network is fast, the time to interactive can still be very long.

This all comes down to our industry’s increasing reliance on JavaScript just to render content. There seems to be pendulum shifts between client-side and server-side rendering. It’s been great to see libraries like Vue and Ember embrace server-side rendering.

But even with server-side rendering, there’s still usually a rehydration step where all the JavaScript gets parsed and that really affects time to interaction.

Code splitting can help. Webpack can do this. That helps with first-party JavaScript, but what about third-party JavaScript?

Scott believes easier to make a fast website than to keep a fast website. And that’s down to all the third-party scripts that people throw in: analytics, ads, tracking. They can wreak havoc on all your hard work.

These scripts apparently contribute to the business model, so it can be hard for us to make the case for removing them. Tools like SpeedCurve can help people stay informed on the impact of these scripts. It allows you to set up performance budgets and it shows you when pages go over budget. When that happens, we have leverage to step in and push back.

Assuming you lose that battle, what else can we do?

These days, lots of A/B testing and personalisation happens on the client side. The tooling is easy to use. But they are costly!

A typical problematic pattern is this: the server sends one version of the page, and once the page is loaded, the whole page gets replaced with a different layout targeted at the user. This leads to a terrifying new metric that Scott calls Second Meaningful Content.

Assuming we can’t remove the madness, what can we do? We could at least not do this for first-time visits. We could load the scripts asyncronously. We can preload the scripts at the top of the page. But ideally we want to move these things to the server. Server-side A/B testing and personalisation have existed for a while now.

Scott has been experimenting with a middleware solution. There’s this idea of server workers that Cloudflare is offering. You can manipulate the page that gets sent from the server to the browser—all the things you would do for an A/B test. Scott is doing this by using comments in the HTML to demarcate which portions of the page should be filtered for testing. The server worker then deletes a block for some users, and deletes a different block for other users. Scott has written about this approach.

The point here isn’t about using Cloudflare. The broader point is that it’s much faster to do these things on the server. We need to defend our user’s time.

Another issue, other than third-party scripts, is the page weight on home pages and landing pages. Marketing teams love to fill these things with enticing rich imagery and carousels. They’re really difficult to keep performant because they change all the time. Sometimes we’re not even in control of the source code of these pages.

We can advocate for new best practices like responsive images. The srcset attribute on the img element; the picture element for when you need more control. These are great tools. What’s not so great is writing the markup. It’s confusing! Ideally we’d have a CMS drive this, but a lot of the time, landing pages fall outside of the purview of the CMS.

Scott has been using Vue.js to make a responsive image builder—a form that people can paste their URLs into, which spits out the markup to use. Anything we can do by creating tools like these really helps to defend the performance of a site.

Another thing we can do is lazy loading. Focus on the assets. The BBC homepage uses some lazy loading for images—they blink into view as your scroll down the page. They use LazySizes, which you can find on Github. You use data- attributes to list your image sources. Scott realises that LazySizes is not progressive enhancement. He wouldn’t recommend using it on all images, just some images further down the page.

But thankfully, we won’t need these workarounds soon. Soon we’ll have lazy loading in browsers. There’s a lazyload attribute that we’ll be able to set on img and iframe elements:

<img src=".." alt="..." lazyload="on">

It’s not implemented yet, but it’s coming in Chrome. It might be that this behaviour even becomes the default way of loading images in browsers.

If you dig under the hood of the implementation coming in Chrome, it actually loads all the images, but the ones being lazyloaded are only sent partially with a 206 response header. That gives enough information for the browser to lay out the page without loading the whole image initially.

To wrap up, Scott takes comfort from the fact that there are resilient patterns out there to help us. And remember, it is our job to defend the user’s experience.

Prototypes and production

When we do front-end development at Clearleft, we’re usually delivering production code, often in the form of a component library. That means our priorities are performance, accessibility, robustness, and other markers of quality when it comes to web development.

But every so often, we use the materials of front-end development—HTML, CSS, and JavaScript—to produce something that isn’t intended for production. I’m talking about prototyping.

There are plenty of non-code prototyping tools out there, and our designers often reach for them to communicate subtleties like motion design. But when it comes to testing a prototype with real users, it’s hard to beat the flexibility of HTML, CSS, and JavaScript. Load it up in a browser and away you go.

We do a lot of design sprints, where time is of the essence. The prototype we produce on the penultimate day of the sprint definitely won’t be production quality, but it will be good enough to test.

What’s interesting is that—when it comes to prototyping—our usual front-end priorities can and should go out the window. The priority now is speed. If that means sacrificing semantics or performance, then so be it. If I’m building a prototype and I find myself thinking “now, what’s the right class name for this component?”, then I know I’m in the wrong mindset. That question might be valid for production code, but it’s a waste of time for prototypes.

So these two kinds of work require very different attitudes. For production work, quality is key. For prototyping, making something quickly is what matters.

Whereas I would think long and hard about the performance impacts of third-party libraries and frameworks on a public project, I won’t give it a second thought when it comes to a prototype. Throw all the JavaScript frameworks and CSS libraries you want at it (although I would argue that in-browser technologies like CSS Grid have made CSS libraries like Bootstrap less necessary, even for prototyping).

Alternating between production projects and prototyping projects can be quite fun, if a little disorienting. It’s almost like I have to flip a switch in my brain to change tracks.

When a prototype is successful, works great, and tests well, there’s a real temptation to use the prototype code as the basis for the final product. Don’t do this! I’ve made that mistake in the past and it always ends badly. I ended up spending far more time trying to wrangle prototype code to a production level than if I had just started from a clean slate.

Build prototypes to test ideas, designs, interactions, and interfaces …and then throw the code away. The value of a prototype is in answering questions and testing hypotheses. Don’t fall for the sunk cost fallacy when it’s time to switch over into production mode.

Of course it should go without saying that you should never, ever release prototype code into production.

And yet…

More and more live sites seem to be built with a prototyping mindset. Weighty JavaScript frameworks are used regardless of appropriateness. Accessibility, if it’s even considered at all, is relegated to an afterthought. Fragile architectures are employed that rely on first loading and then executing JavaScript in order to render basic content. Developer experience is prioritised over user experience.

Heydon recently highlighted an article that offered this tip for aspiring web developers:

As for HTML, there’s not much to learn right away and you can kind of learn as you go, but before making your first templates, know the difference between in-line elements like span and how they differ from block ones like div.

That’s perfectly reasonable advice …if you’re building a prototype. But if you’re building something for public consumption, you have a duty of care to the end users.

Optimise without a face

I’ve been playing around with the newly-released Squoosh, the spiritual successor to Jake’s SVGOMG. You can drag images into the browser window, and eyeball the changes that any optimisations might make.

On a project that Cassie is working on, it worked really well for optimising some JPEGs. But there were a few images that would require a bit more fine-grained control of the optimisations. Specifically, pictures with human faces in them.

I’ve written about this before. If there’s a human face in image, I open that image in a graphics editing tool like Photoshop, select everything but the face, and add a bit of blur. Because humans are hard-wired to focus on faces, we’ll notice any jaggy artifacts on a face, but we’re far less likely to notice jagginess in background imagery: walls, materials, clothing, etc.

On the face of it (hah!), a browser-based tool like Squoosh wouldn’t be able to optimise for faces, but then Cassie pointed out something really interesting…

When we were both at FFConf on Friday, there was a great talk by Eleanor Haproff on machine learning with JavaScript. It turns out there are plenty of smart toolkits out there, and one of them is facial recognition. So I wonder if it’s possible to build an in-browser tool with this workflow:

  • Drag or upload an image into the browser window,
  • A facial recognition algorithm finds any faces in the image,
  • Those portions of the image remain crisp,
  • The rest of the image gets a slight blur,
  • Download the optimised image.

Maybe the selecting/blurring part would need canvas? I don’t know.

Anyway, I thought this was a brilliant bit of synthesis from Cassie, and now I’ve got two questions:

  1. Does this exist yet? And, if not,
  2. Does anyone want to try building it?

A framework for web performance

Here at Clearleft, we’ve recently been doing some front-end consultancy. That prompted me to jot down thoughts on design principles and performance:

We continued with some more performance work this week. Having already covered some of the nitty-gritty performance tactics like font-loading, image optimisation, etc., we wanted to take a step back and formulate an ongoing strategy for performance.

When it comes to web performance, the eternal question is “What should we measure?” The answer to that question will determine where you then concentrate your efforts—whatever it is your measuring, that’s what you’ll be looking to improve.

I started by drawing a distinction between measurements of quantities and measurements of time. Quantities are quite easy to measure. You can measure these quantities using nothing more than browser dev tools:

  • overall file size (page weight + assets), and
  • number of requests.

I think it’s good to measure these quantities, and I think it’s good to have a performance budget for them. But I also think they’re table stakes. They don’t actually tell you much about the impact that performance is having on the user experience. For that, we need to enumerate moments in time:

  • time to first byte,
  • time to first render,
  • time to first meaningful paint, and
  • time to first meaningful interaction.

There’s one more moment in time, which is the time until DOM content is loaded. But I’m not sure that has a direct effect on how performance is perceived, so it feels like it belongs more in the category of quantities than time.

Next, we listed out all the factors that could affect each of the moments in time. For example, the time to first byte depends on the speed of the network that the user is on. It also depends on how speedily your server (or Content Delivery Network) can return a response. Meanwhile, time to first render is affected by the speed of the user’s network, but it’s also affected by how many blocking elements are on the critical path.

By listing all the factors out, we can draw a distinction between the factors that are outside of our control, and the factors that we can do something about. So while we might not be able to do anything about the speed of the user’s network, we might well be able to optimise the speed at which our server returns a response, or we might be able to defer some assets that are currently blocking the critical path.

Factors
1st byte
  • server speed
  • network speed
1st render
  • network speed
  • critical path assets
1st meaningful paint
  • network speed
  • font-loading strategy
  • image optimisation
1st meaningful interaction
  • network speed
  • device processing power
  • JavaScript size

So far, everything in our list of performance-affecting factors is related to the first visit. It’s worth drawing up a second list to document all the factors for subsequent visits. This will look the same as the list for first visits, but with the crucial difference that caching now becomes a factor.

First visit factors Repeat visit factors
1st byte
  • server speed
  • network speed
  • server speed
  • network speed
  • caching
1st render
  • network speed
  • critical path assets
  • network speed
  • critical path assets
  • caching
1st meaningful paint
  • network speed
  • font-loading strategy
  • image optimisation
  • network speed
  • font-loading strategy
  • image optimisation
  • caching
1st meaningful interaction
  • network speed
  • device processing power
  • JavaScript size
  • network speed
  • device processing power
  • JavaScript size
  • caching

Alright. Now it’s time to get some numbers for each of the four moments in time. I use Web Page Test for this. Choose a realistic setting, like 3G on an Android from the East coast of the USA. Under advanced settings, be sure to select “First View and Repeat View” so that you can put those numbers in two different columns.

Here are some numbers for adactio.com:

First visit time Repeat visit time
1st byte 1.476 seconds 1.215 seconds
1st render 2.633 seconds 1.930 seconds
1st meaningful paint 2.633 seconds 1.930 seconds
1st meaningful interaction 2.868 seconds 2.083 seconds

I’m getting the same numbers for first render as first meaningful paint. That tells me that there’s no point in trying to optimise my font-loading, for example …which makes total sense, because adactio.com isn’t using any web fonts. But on a different site, you might see a big gap between those numbers.

I am seeing a gap between time to first byte and time to first render. That tells me that I might be able to get some blocking requests off the critical path. Sure enough, I’m currently referencing an external stylesheet in the head of adactio.com—if I were to inline critical styles and defer the loading of that stylesheet, I should be able to narrow that gap.

A straightforward site like adactio.com isn’t going to have much to worry about when it comes to the time to first meaningful interaction, but on other sites, this can be a significant bottleneck. If you’re sending UI elements in the initial HTML, but then waiting for JavaScript to “hydrate” those elements into working, the user can end up in an uncanny valley of tapping on page elements that look fine, but aren’t ready yet.

My point is, you’re going to see very different distributions of numbers depending on the kind of site you’re testing. There’s no one-size-fits-all metric to focus on.

Now that you’ve got numbers for how your site is currently performing, you can create two new columns: one of those is a list of first-visit targets, the other is a list of repeat-visit targets for each moment in time. Try to keep them realistic.

For example, if I could reduce the time to first render on adactio.com by 0.5 seconds, my goals would look like this:

First visit goal Repeat visit goal
1st byte 1.476 seconds 1.215 seconds
1st render 2.133 seconds 1.430 seconds
1st meaningful paint 2.133 seconds 1.430 seconds
1st meaningful interaction 2.368 seconds 1.583 seconds

See how the 0.5 seconds saving cascades down into the other numbers?

Alright! Now I’ve got something to aim for. It might also be worth having an extra column to record which of the moments in time are high priority, which are medium priority, and which are low priority.

Priority
1st byte Medium
1st render High
1st meaningful paint Low
1st meaningful interaction Low

Your goals and priorities may be quite different.

I think this is a fairly useful framework for figuring out where to focus when it comes to web performance. If you’d like to give it a go, I’ve made a web performance chart for you to print out and fill in. Here’s a PDF version if that’s easier for printing. Or you can download the HTML version if you want to edit it.

I have to say, I’m really enjoying the front-end consultancy work we’ve been doing at Clearleft around performance and related technologies, like offline functionality. I’d like to do more of it. If you’d like some help in prioritising performance at your company, please get in touch. Let’s make the web faster together.

The top four web performance challenges

Danielle and I have been doing some front-end consultancy for a local client recently.

We’ve both been enjoying it a lot—it’s exhausting but rewarding work. So if you’d like us to come in and spend a few days with your company’s dev team, please get in touch.

I’ve certainly enjoyed the opportunity to watch Danielle in action, leading a workshop on refactoring React components in a pattern library. She’s incredibly knowledgable in that area.

I’m clueless when it comes to React, but I really enjoy getting down to the nitty-gritty of browser features—HTML, CSS, and JavaScript APIs. Our skillsets complement one another nicely.

This recent work was what prompted my thoughts around the principles of robustness and least power. We spent a day evaluating a continuum of related front-end concerns: semantics, accessibility, performance, and SEO.

When it came to performance, a lot of the work was around figuring out the most suitable metric to prioritise:

  • time to first byte,
  • time to first render,
  • time to first meaningful paint, or
  • time to first meaningful interaction.

And that doesn’t even cover the more easily-measurable numbers like:

  • overall file size,
  • number of requests, or
  • pagespeed insights score.

One outcome was to realise that there’s a tendency (in performance, accessibility, or SEO) to focus on what’s easily measureable, not because it’s necessarily what matters, but precisely because it is easy to measure.

Then we got down to some nuts’n’bolts technology decisions. I took a step back and looked at the state of performance across the web. I thought it would be fun to rank the most troublesome technologies in order of tricksiness. I came up with a top four list.

Here we go, counting down from four to the number one spot…

4. Web fonts

Coming in at number four, it’s web fonts. Sometimes it’s the combined weight of multiple font files that’s the problem, but more often that not, it’s the perceived performance that suffers (mostly because of when the web fonts appear).

Fortunately there’s a straightforward question to ask in this situation: WWZD—What Would Zach Do?

3. Images

At the number three spot, it’s images. There are more of them and they just seem to be getting bigger all the time. And yet, we have more tools at our disposal than ever—better file formats, and excellent browser support for responsive images. Heck, we’re even getting the ability to lazy load images in HTML now.

So, as with web fonts, it feels like the impact of images on performance can be handled, as long as you give them some time and attention.

2. Our JavaScript

Just missing out on making the top spot is the JavaScript that we send down the pipe to our long-suffering users. There’s nothing wrong with the code itself—I’m sure it’s very good. There’s just too damn much of it. And that’s a real performance bottleneck, especially on mobile.

So stop sending so much JavaScript—a solution as simple as Monty Python’s instructions for playing the flute.

1. Other people’s JavaScript

At number one with a bullet, it’s all the crap that someone else tells us to put on our websites. Analytics. Ads. Trackers. Beacons. “It’s just one little script”, they say. And then that one little script calls in another, and another, and another.

It’s so disheartening when you’ve devoted your time and energy into your web font loading strategy, and optimising your images, and unbundling your JavaScript …only to have someone else’s JavaScript just shit all over your nice performance budget.

Here’s the really annoying thing: when I go to performance conferences, or participate in performance discussions, you know who’s nowhere to be found? The people making those third-party scripts.

The narrative around front-end performance is that it’s up to us developers to take responsibility for how our websites perform. But by far the biggest performance impact comes from third-party scripts.

There is a solution to this, but it’s not a technical one. We could refuse to add overweight (and in many cases, unethical) third-party scripts to the sites we build.

I have many, many issues with Google’s AMP project, but I completely acknowledge that it solves a political problem:

No external JavaScript is allowed in an AMP HTML document. This covers third-party libraries, advertising and tracking scripts. This is A-okay with me.

The reasons given for this ban are related to performance and I agree with them completely. Big bloated JavaScript libraries are one of the biggest performance killers on the web.

But how can we take that lesson from AMP and apply it to all our web pages? If we simply refuse to be the one to add those third-party scripts, we get fired, and somebody else comes in who is willing to poison web pages with third-party scripts. There’s nothing to stop companies doing that.

Unless…

Suppose we were to all make a pact that we would stand in solidarity with any of our fellow developers in that sort of situation. A sort of joining-together. A union, if you will.

There is power in a factory, power in the land, power in the hands of the worker, but it all amounts to nothing if together we don’t stand.

There is power in a union.

Name That Script! by Trent Walton

Trent is about to pop his AEA cherry and give a talk at An Event Apart in Boston. I’m going to attempt to liveblog this:

How many third-party scripts are loading on our web pages these days? How can we objectively measure the value of these (advertising, a/b testing, analytics, etc.) scripts—considering their impact on web performance, user experience, and business goals? We’ve learned to scrutinize content hierarchy, browser support, and page speed as part of the design and development process. Similarly, Trent will share recent experiences and explore ways to evaluate and discuss the inclusion of 3rd-party scripts.

Trent is going to speak about third-party scripts, which is funny, because a year ago, he never would’ve thought he’d be talking about this. But he realised he needed to pay more attention to:

any request made to an external URL.

Or how about this:

A resource included with a web page that the site owner doesn’t explicitly control.

When you include a third-party script, the third party can change the contents of that script.

Here are some uses:

  • advertising,
  • A/B testing,
  • analytics,
  • social media,
  • content delivery networks,
  • customer interaction,
  • comments,
  • tag managers,
  • fonts.

You get data from things like analytics and A/B testing. You get income from ads. You get content from CDNs.

But Trent has concerns. First and foremost, the user experience effects of poor performance. Also, there are the privacy implications.

Why does Trent—a designer—care about third party scripts? Well, over the years, the areas that Trent pays attention to has expanded. He’s progressed from image comps to frontend to performance to accessibility to design systems to the command line and now to third parties. But Trent has no impact on those third-party scripts. That’s very different to all those other areas.

Trent mostly builds prototypes. Those then get handed over for integration. Sometimes that means hooking it up to a CMS. Sometimes it means adding in analytics and ads. It gets really complex when you throw in third-party comments, payment systems, and A/B testing tools. Oftentimes, those third-party scripts can outweigh all the gains made beforehand. It happens with no discussion. And yet we spent half a meeting discussing a border radius value.

Delivering a performant, accessible, responsive, scalable website isn’t enough: I also need to consider the impact of third-party scripts.

Trent has spent the last few months learning about third parties so he can be better equiped to discuss them.

UX, performance and privacy impact

We feel the UX impact every day we browse the web (if we turn off our content blockers). The Food Network site has an intersitial asking you to disable your ad blocker. They promise they won’t spawn any pop-up windows. Trent turned his ad blocker off—the page was now 15 megabytes in size. And to top it off …he got a pop up.

Privacy can harder to perceive. We brush aside cookie notifications. What if the wording was “accept trackers” instead of “accept cookies”?

Remarketing is that experience when you’re browsing for a spatula and then every website you visit serves you ads for spatula. That might seem harmless but allowing access to our browsing history has serious privacy implications.

Web builders are on the front lines. It’s up to us to advocate for data protection and privacy like we do for web standards. Don’t wait to be told.

Categories of third parties

Ghostery categories third-party providers: advertising, comments, customer interaction, essential, site analytics, social media. You can dive into each layer and see the specific third-party services on the page you’re viewing.

Analyse and itemise third-party scripts

We have “view source” for learning web development. For third parties, you need some tool to export the data. HAR files (HTTP ARchive) are JSON files that you can create from most browsers’ network request panel in dev tools. But what do you do with a .har file? The site har.tech has plenty of resources for you. That’s where Trent found the Mac app, Charles. It can open .har files. Best of all, you can export to CSV so you can share spreadsheets of the data.

You can visualise third-party requests with Simon Hearne’s excellent Request Map. It’s quite impactful for delivering a visceral reaction in a meeting—so much more effective than just saying “hey, we have a lot of third parties.” Request Map can also export to CSV.

Know industry averages

Trent wanted to know what was “normal.” He decided to analyse HAR files for Alexa’s top 50 US websites. The result was a massive spreadsheet of third-party providers. There were 213 third-party domains (which is not even the same as the number of requests). There was an average of 22 unique third-party domains per site. The usual suspects were everywhere—Google, Amazon, Facebook, Adobe—but there were many others. You can find an alphabetical index on better.fyi/trackers. Often the lesser-known domains turn out to be owned by the bigger domains.

News sites and shopping sites have the most third-party scripts, unsurprisingly.

Understand benefits

Trent realised he needed to listen and understand why third-party scripts are being included. He found out what tag managers do. They’re funnels that allow you to cram even more third-party scripts onto your website. Trent worried that this was a Pandora’s box. The tag manager interface is easy to access and use. But he was told that it’s more like a way of organising your third-party scripts under one dashboard. But still, if you get too focused on the dashboard, you could lose focus of the impact on load times. So don’t blame the tool: it’s all about how it’s used.

Take action

Establish a centre of excellence. Put standards in place—in a cross-discipline way—to define how third-party scripts are evaluated. For example:

  1. Determine the value to the business.
  2. Avoid redundant scripts and services.
  3. Fit within the established performance budget.
  4. Comply with the organistional privacy policy.

Document those decisions, maybe even in your design system.

Also, include third-party scripts within your prototypes to get a more accurate feel for the performance implications.

On a live site, you can regularly audit third-party scripts on a regular basis. Check to see if any are redundant or if they’re exceeding the performance budget. You can monitor performance with tools like Calibre and Speed Curve to cover the time in between audits.

Make your case

Do competitive analysis. Look at other sites in your sector. It’s a compelling way to make a case for change. WPO Stats is very handy for anecdata.

You can gather comparative data with Web Page Test: you can run a full test, and you can run a test with certain third parties blocked. Use the results to kick off a discussion about the impact of those third parties.

Talk it out

Work to maintain an ongoing discussion with the entire team. As Tim Kadlec says:

Everything should have a value, because everything has a cost.

AMPstinction

I’ve come to believe that the goal of any good framework should be to make itself unnecessary.

Brian said it explicitly of his PhoneGap project:

The ultimate purpose of PhoneGap is to cease to exist.

That makes total sense, especially if your code is a polyfill—those solutions are temporary by design. Autoprefixer is another good example of a piece of code that becomes less and less necessary over time.

But I think it’s equally true of any successful framework or library. If the framework becomes popular enough, it will inevitably end up influencing the standards process, thereby becoming dispensible.

jQuery is the classic example of this. There’s very little reason to use jQuery these days because you can accomplish so much with browser-native JavaScript. But the reason why you can accomplish so much without jQuery is because of jQuery. I don’t think we would have querySelector without jQuery. The library proved the need for the feature. The same is true for a whole load of DOM scripting features.

The same process is almost certain to occur with React—it’s a good bet there will be a standardised equivalent to the virtual DOM at some point.

When Google first unveiled AMP, its intentions weren’t clear to me. I hoped that it existed purely to make itself redundant:

As well as publishers creating AMP versions of their pages in order to appease Google, perhaps they will start to ask “Why can’t our regular pages be this fast?” By showing that there is life beyond big bloated invasive web pages, perhaps the AMP project will work as a demo of what the whole web could be.

Alas, as time has passed, that hope shows no signs of being fulfilled. If anything, I’ve noticed publishers using the existence of their AMP pages as a justification for just letting their “regular” pages put on weight.

Worse yet, the messaging from Google around AMP has shifted. Instead of pitching it as a format for creating parallel versions of your web pages, they’re now also extolling the virtues of having your AMP pages be the only version you publish:

In fact, AMP’s evolution has made it a viable solution to build entire websites.

On an episode of the Dev Mode podcast a while back, AMP was a hotly-debated topic. But even those defending AMP were doing so on the understanding that it was more a proof-of-concept than a long-term solution (and also that AMP is just for news stories—something else that Google are keen to change).

But now it’s clear that the Google AMP Project is being marketed more like a framework for the future: a collection of web components that prioritise performance …which is kind of odd, because that’s also what Google’s Polymer project is. The difference being that pages made with Polymer don’t get preferential treatment in Google’s search results. I can’t help but wonder how the Polymer team feels about AMP’s gradual pivot onto their territory.

If the AMP project existed in order to create a web where AMP was no longer needed, I think I could get behind it. But the more it’s positioned as the only viable solution to solving performance, the more uncomfortable I am with it.

Which, by the way, brings me to one of the most pernicious ideas around Google AMP—positioning anyone opposed to it as not caring about web performance. Nothing could be further from the truth. It’s precisely because performance on the web is so important that it deserves a long-term solution, co-created by all of us: not some commandents delivered to us from on-high by one organisation, enforced by preferential treatment by that organisation’s monopoly in search.

It’s the classic logical fallacy:

  1. Performance! Something must be done!
  2. AMP is something.
  3. Now something has been done.

By marketing itself as the only viable solution to the web performance problem, I think the AMP project is doing itself a great disservice. If it positioned itself as an example to be emulated, I would welcome it.

I wish that AMP were being marketed more like a temporary polyfill. And as with any polyfill, I look forward to the day when AMP is no longer necesssary.

I want AMP to become extinct. I genuinely think that the Google AMP team should share that wish.

Beyond Engagement: the Content Performance Quotient by Jeffrey Zeldman

I’m at An Event Apart Seattle (Special Edition). Jeffrey is kicking off the show with a presentation called Beyond Engagement: the Content Performance Quotient. I’m going to jot down some notes during this talk…

First, a story. Jeffrey went to college in Bloomington, Indiana. David Frost—the British journalist—came to talk to them. Frost had a busy schedule, and when he showed up, he seemed a little tipsy. He came up to the podium and said, “Good evening, Wilmington.”

Jeffrey remembers this and knows that Seattle and Portland have a bit of a rivalry, and so Jeffrey thought, the first time he spoke in Portland, it would be funny to say “Good morning, Seattle!” …and that was the last time he spoke in Portland.

Anyway …”Good morning, Portland!”

Jeffrey wants to talk about content. He spends a lot of time in meetings with stakeholders. Those stakeholders always want things to be better, and they always talk about “engagement.” It’s the number one stakeholder request. It’s a metric that makes stakeholders feel comfortable. It’s measurable—the more seconds people give us, the better.

But is that really the right metric?

There are some kinds of sites where engagement is definitely the right metric. Instagram, for example. That’s how they make money. You want to distract yourself. Also, if you have a big content site—beautifully art-directed and photographed—then engagement is what you want. You want people to spend a lot of time there. Or if you have a kids site, or a games site, or a reading site for kids, you want them to be engaged and spend time. A List Apart, too. It’s like the opposite of Stack Overflow, where you Google something and grab the piece of code you need and then get out. But for A List Apart or Smashing Magazine, you want people to read and think and spend their time. Engagement is what you want.

But for most sites—insurance, universities—engagement is not what you want. These sites are more like a customer service desk. You want to help the customer as quickly as possible. If a customer spends 30 minutes on our site, was she engaged …or frustrated? Was it the beautiful typography and copy …or because she couldn’t find what they wanted? If someone spends a long time on an ecommerce site, is it because the products are so good …or because search isn’t working well?

What we need is a metric called speed of usefulness. Jeffrey calls this Content Performance Quotient (CPQ) …because business people love three-letter initialisms. It’s a loose measurement: How quickly can you solve the customer’s problem? It’s the shortest distance between the problem and the solution. Put another way, it’s a measurement of your value to the customer. It’s a new way to evaluate success.

From the customer’s point of view, CPQ is the time it takes the customer to get the information she came for. From the organisation’s point of view, it’s the time it takes for a specific customer to find, receive, and absorb your most important content.

We’re all guilty of neglecting the basics on our sites—just what it is it that we do? We need to remember that we’re all making stuff to make people’s live’s easier. Otherwise we end up with what Jeffrey calls “pretty garbage.” It’s aesthetically coherent and visually well-designed …but if the content is wrong and doesn’t help anyone, it’s garbage. Garbage in a delightfully responsive grid is still garbage.

Let’s think of an example of where people really learned to cut back and really pare down their message. Advertising. In the 1950s, when the Leo Burnett agency started the Marlboro campaign, TV spots were 60 seconds long. An off-camera white man in a suit with a soothing voice would tell you all about the product while the visuals showed you what he was talking about. No irony. Marlboro did a commercial where there was no copy at all until the very end. For 60 seconds they showed you cowboys doing their rugged cowboy things. Men in the 1950s wanted to feel rugged, you see. Leo Burnett aimed the Marlboro cigarettes at those men. And at the end of the 60 second montage of rugged cowboys herding steers, they said “Come to where the flavour is. Come to Marlboro Country.” For the billboard, they cut it back even more. Just “Come to Marlboro Country.” In fact, they eventually went to just “Marlboro.” Jeffrey knows that this campaign worked well, because he started smoking Marlboros as a kid.

Leaving aside the ethical implications of selling cigarettes to eight-graders, let’s think about the genius of those advertisers. Slash your architecture and shrink your content. Constantly ask yourself, “Why do we need this?”

As Jared Spool says, design is the rendering of intent. Every design is intentional. There is some intent—like engagement—driving our design. If there’s no intent behind the design, it will fail, even if what you’re doing is very good. If your design isn’t going somewhere, it’s going nowhere. You’ve got to stay ruthlessly focused on what the customer needs and “kill your darlings” as Hemingway said. Luke Wroblewski really brought this to light when he talked about Mobile First.

To paraphrase David Byrne, how did we get here?

Well, we prioritised meetings over meaning. Those meetings can be full of tension; different stakeholders arguing over what should be on the homepage. And we tried to solve this by giving everyone what they want. Having a good meeting doesn’t necessarily mean having a good meeting. We think of good meetings as conflict-free where everyone emerges happy. But maybe there should be a conflict that gets resolved. Maybe there should be winners and losers.

Behold our mighty CMS! Anyone can add content to the website. Anyone can create the information architecture …because we want to make people happy in meetings. It’s easy to give everyone what they want. It’s harder to do the right thing. Harder for us, but better for the customer and the bottom line.

As Gerry McGovern says:

Great UX professionals are like whistleblowers. They are the voice of the user.

We need to stop designing 2001 sites for a 2018 web.

One example of cutting down content was highlighted in A List Apart where web design was compared to chess: The King vs. Pawn Game of UI Design. Don’t start by going through all the rules. Teach them in context. Teach chess by starting with a checkmate move, reduced down to just three pieces on the board. From there, begin building out. Start with the most important information, and build out from there.

When you strip down the game to its core, everything you learn is a universal principle.

Another example is atomic design: focus relentlessly on the individual interaction. We do it for shopping carts. We can do it for content.

Another example on A List Apart is No More FAQs: Create Purposeful Information for a More Effective User Experience. FAQ problems include:

  • duplicate and contradictory information,
  • lack of discernible content order,
  • repetitive grammatical structure,
  • increased cognitive load, and
  • more content than they need.

Users come to any type of content with a particular purpose in mind, ranging from highly specific (task completion) to general learning (increased knowledge).

The important word there is purpose. We need to eliminate distraction. How do we do that?

One way is the waterfall method. Do a massive content inventory. It’s not recommended (unless maybe you’re doing a massive redesign).

Agile and scrum is another way. Constantly iterate on content. Little by little over time, we make the product better. It’s the best bet if you work in-house.

If you work in an agency, a redesign is an opportunity to start fresh. Take everything off the table and start from scratch. Jeffrey’s friend Fred Gates got an assignment to redesign an online gaming platform for kids to teach them reading and management skills. The organisation didn’t have much money so they said, let’s just do the homepage. Fred challenged himself to put the whole thing on the homepage. The homepage tells the whole story. Jeffrey is using this same method on a site for an insurance company, even though the client has a bigger budget and can afford more than just the homepage. The point is, what Fred did was effective.

So this is what Jeffrey is going to be testing and working on: speed of usefulness.

And for those of you who do need to use engagement as the right metric, Jeffrey covered the two kinds of metrics in an article called We need design that is faster and design that is slower.

For example, “scannability” is good for transactions (CPQ), but bad for thoughtful content (engagement). Our news designs need to slow down the user. Bigger type, typographic hierarchy, and more whitespace. Art direction. Shout out to Derek Powazek who designed Fray.com—each piece was designed based on the content. These days, look at what David Sleight and his crew are doing over at Pro Publica.

Who’s doing it right?

The Washington Post, The New York Times, Pro Publica, Slate, Smashing Magazine, and Vox are all doing this well in different ways. They’re bringing content to the fore.

Readability, Medium, and A List Apart are all using big type to encourage thoughtful reading and engagement.

But for other sites …apply the Content Performance Quotient.

See also:

Minimal viable service worker

I really, really like service workers. They’re one of those technologies that have such clear benefits to users that it seems like a no-brainer to add a service worker to just about any website.

The thing is, every website is different. So the service worker strategy for every website needs to be different too.

Still, I was wondering if it would be possible to create a service worker script that would work for most websites. Here’s the script I came up with.

The logic works like this:

  • If there’s a request for an HTML page, fetch it from the network and store a copy in a cache (but if the network request fails, try looking in the cache instead).
  • For any other files, look for a copy in the cache first but meanwhile fetch a fresh version from the network to update the cache (and if there’s no existing version in the cache, fetch the file from the network and store a copy of it in the cache).

So HTML files are served network-first, while all other files are served cache-first, but in both cases a fresh copy is always put in the cache. The idea is that HTML content will always be fresh (unless there’s a problem with the network), while all other content—images, style sheets, scripts—might be slightly stale, but get refreshed with every request.

My original attempt was riddled with errors. Jake came to my rescue and we revised the script into something that actually worked. In the process, my misunderstanding of how await works led Jake to write a great blog post on await vs return vs return await.

I got there in the end and the script seems solid enough. It’s a fairly simplistic strategy that could work for quite a few sites, but it has some issues…

Service workers don’t perform any automatic cleanup of caches—that’s up to you to do (usually during the activate event). This script doesn’t do any cleanup so the cache might grow and grow and grow. For that reason, I think the script is best suited for fairly small sites.

The strategy also assumes that a file will either be fetched from the network or the cache. There’s no contingency for when both attempts fail. So there’s no fallback offline page, for example.

I decided to test it in the wild, but I expanded it slightly to fix the fallback issue. The version on the Ampersand 2018 website includes a worst-case-scenario option to show a custom offline page that has been pre-cached. (By the way, if you haven’t got a ticket for Ampersand yet, get a ticket now—it’s going to be superb day of web typography nerdery.)

Anyway, this fairly basic script seems to be delivering some good performance improvements. If you’ve got a site that you think would benefit from this network/caching strategy, and it’s served over HTTPS, then:

  1. Feel free to download the script or copy and paste it into a file called serviceworker.js,
  2. Put that file in the root directory of your website,
  3. Add this in a script element at the bottom of your HTML pages:

if (navigator.serviceWorker && !navigator.serviceWorker.controller) { navigator.serviceWorker.register('/serviceworker.js'); }

You can also use the script as a starting point. You might find issues specific to your particular website. That’s okay—you can tweak and adjust the script to suit your needs.

If this minimal service worker script proves in any way useful to you, thank Jake.

Needs must

I got a follow-up comment to my follow-up post about the follow-up comment I got on my original post about Google Analytics. Keep up.

I made the point that, from a front-end performance perspective, server logs have no impact whereas a JavaScript-based analytics solution must have some impact on the end user. Paul Anthony says:

Google won the analytics war because dropping one line of JS in the footer and handing a tried and tested interface to customers is an obvious no brainer in comparison to setting up an open source option that needs a cron job to parse the files, a database to store the results and doesn’t provide mobile interface.

Good point. Dropping one snippet of JavaScript into your front-end codebase is certainly an easier solution …easier for you, that is. The cost is passed on to your users. This is a classic example of where user needs and developer needs are in opposition. I’ve said it before and I’ll say it again:

Given the choice between making something my problem, and making something the user’s problem, I’ll choose to make it my problem every time.

It’s true that this often means doing more work. That’s why it’s called work. This is literally what our jobs are supposed to entail: we put in the work to make life easier for users. We’re supposed to be saving them time, not passing it along.

The example of Google Analytics is pretty extreme, I’ll grant you. The cost to the user of adding that snippet of JavaScript—if you’ve configured things reasonably well—is pretty small (again, just from a performance perspective; there’s still the cost of allowing Google to track them across domains), and the cost to you of setting up a comparable analytics system based on server logs can indeed be disproportionately high. But this tension between user needs and developer needs is something I see play out again and again.

I’ve often thought the HTML design principle called the priority of constituencies could be adopted by web developers:

In case of conflict, consider users over authors over implementors over specifiers over theoretical purity. In other words costs or difficulties to the user should be given more weight than costs to authors.

In Resilient Web Design, I documented the three-step approach I take when I’m building anything on the web:

  1. Identify core functionality.
  2. Make that functionality available using the simplest possible technology.
  3. Enhance!

Now I’m wondering if I should’ve clarified that second step further. When I talk about choosing “the simplest possible technology”, what I mean is “the simplest possible technology for the user”, not “the simplest possible technology for the developer.”

For example, suppose I were going to build a news website. The core functionality is fairly easy to identify: providing the news. Next comes the step where I choose the simplest possible technology. Now, if I were a developer who had plenty of experience building JavaScript-driven single page apps, I might conclude that the simplest route for me would be to render the news via JavaScript. But that would be a fragile starting point if I’m trying to reach as many people as possible (I might well end up building a swishy JavaScript-driven single page app in step three, but step two should almost certainly be good ol’ HTML).

Time and time again, I see decisions that favour developer convenience over user needs. Don’t get me wrong—as a developer, I absolutely want developer convenience …but not at the expense of user needs.

I know that “empathy” is an over-used word in the world of user experience and design, but with good reason. I think we should try to remind ourselves of why we make our architectural decisions by invoking who those decisions benefit. For example, “This tech stack is best option for our team”, or “This solution is the best for the widest range of users.” Then, given the choice, favour user needs in the decision-making process.

There will always be situations where, given time and budget constraints, we end up choosing solutions that are easier for us, but not the best for our users. And that’s okay, as long as we acknowledge that compromise and strive to do better next time.

But when the best solutions for us as developers become enshrined as the best possible solutions, then we are failing the people we serve.

That doesn’t mean we must become hairshirt-wearing martyrs; developer convenience is important …but not as important as user needs. Start with user needs.

Heisenberg

I wrote about Google Analytics yesterday. As usual, I syndicated the post to Ev’s blog, and I got an interesting response over there. Kelly Burgett set me straight on some of the finer details of how goals work, and finished with this thought:

You mention “delivering a performant, accessible, responsive, scalable website isn’t enough” as if it should be, and I have to disagree. It’s not enough for a business to simply have a great website if you are unable to understand performance of channel marketing, track user demographics and behavior on-site, and optimize your site/brand based on that data. I’ve seen a lot of ugly sites who have done exceptionally well in terms of ROI, simply because they are getting the data they need from the site in order make better business decisions. If your site cannot do that (ie. through data collection, often third party scripts), then your beautifully-designed site can only take you so far.

That makes an excellent case for having analytics. But that’s not necessarily the same as having Google analytics, or even JavaScript-driven analytics at all.

By far the most useful information you get from analytics is around where people have come from, where did they go next, and what kind of device are they using. None of that information requires JavaScript. It’s all available from your server logs.

I don’t want to come across all old-man-yell-at-cloud here, but I’m trying to remember at what point self-hosted software for analysing your log traffic became not good enough.

Here’s the thing: logging on the server has no effect on the user experience. It’s basically free, in terms of performance. Logging via JavaScript, by its very nature, has some cost. Even if its negligible, that’s one more request, and that’s one more bit of processing for the CPU.

All of the data that you can only get via JavaScript (in-page actions, heat maps, etc.) are, in my experience, better handled by dedicated software. To me, that kind of more precise data feels different to analytics in the sense of funnels, conversions, goals and all that stuff.

So in order to get more fine-grained data to analyse, our analytics software has now doubled down on a technology—JavaScript—that has an impact on the end user, where previously the act of observation could be done at a distance.

There are also blind spots that come with JavaScript-based tracking. According to Google Analytics, 0% of your customers don’t have JavaScript. That’s not necessarily true, but there’s literally no way for Google Analytics—which relies on JavaScript—to even do its job in the absence of JavaScript. That can lead to a dangerous situation where you might be led to think that 100% of your potential customers are getting by, when actually a proportion might be struggling, but you’ll never find out about it.

Related: according to Google Analytics, 0% of your customers are using ad-blockers that block requests to Google’s servers. Again, that’s not necessarily a true fact.

So I completely agree than analytics are a good thing to have for your business. But it does not follow that Google Analytics is a good thing for your business. Other options are available.

I feel like the assumption that “analytics = Google Analytics” is like the slippery slope in reverse. If we’re all agreed that analytics are important, then aren’t we also all agreed that JavaScript-based tracking is important?

In a word, no.

This reminds me of the arguments made in favour of intrusive, bloated advertising scripts. All of the arguments focus on the need for advertising—to stay in business, to pay the writers—which are all great reasons for advertising, but have nothing to do with JavaScript, which is at the root of the problem. Everyone I know who uses an ad-blocker—including me—doesn’t use it to stop seeing adverts, but to stop the performance of the page being degraded (and to avoid being tracked across domains).

So let’s not confuse the means with the ends. If you need to have advertising, that doesn’t mean you need to have horribly bloated JavaScript-based advertising. If you need analytics, that doesn’t mean you need an analytics script on your front end.

Analysing analytics

Hell is other people’s JavaScript.

There’s nothing quite so crushing as building a beautifully performant website only to have it infested with a plague of third-party scripts that add to the weight of each page and reduce the responsiveness, making a mockery of your well-considered performance budget.

Trent has been writing about this:

My latest realization is that delivering a performant, accessible, responsive, scalable website isn’t enough: I also need to consider the impact of third-party scripts.

He’s started the process by itemising third-party scripts. Frustratingly though, there’s rarely one single culprit that you can point to—it’s the cumulative effect of “just one more beacon” and “just one more analytics script” and “just one more A/B testing tool” that adds up to a crappy experience that warms your user’s hands by ensuring your site is constantly draining their battery.

Actually, having just said that there’s rarely one single culprit, Adobe Tag Manager is often at the root of third-party problems. That and adverts. It’s like opening the door of your beautifully curated dream home, and inviting a pack of diarrhetic elephants in: “Please, crap wherever you like.”

But even the more well-behaved third-party scripts can get out of hand. Google Analytics is so ubiquitous that it’s hardly even considered in the list of potentially harmful third-party scripts. On the whole, it’s a fairly well-behaved citizen of your site’s population of third-party scripts (y’know, leaving aside the whole surveillance capitalism business model that allows you to use such a useful tool for free in exchange for Google tracking your site’s visitors across the web and selling the insights from that data to advertisers).

The initial analytics script that you—asynchronously—load into your page isn’t very big. But depending on how you’ve configured your Google Analytics account, that might just be the start of a longer chain of downloads and event handlers.

Ed recently gave a lunchtime presentation at Clearleft on using Google Analytics—he professes modesty but he really knows his stuff. He was making sure that everyone knew how to set up goals’n’stuff.

As I understand it, there are two main categories of goals: events and destinations (there are also durations and pages, but they feel similar to destinations). You use events to answer questions like “Did the user click on this button?” or “Did the user click on that search field?”. You use destinations to answer questions like “Did the user arrive at this page?” or “Did the user come from that page?”

You can add as many goals to your site’s analytics as you want. That’s an intoxicating offer. The problem is that there is potentially a cost for each goal you create. It’s an invisible cost. It’s paid by the user in the currency of JavaScript sent down the wire (I wish that the Google Analytics admin interface were more like the old interface for Google Fonts, where each extra file you added literally pushed a needle higher on a dial).

It strikes me that the event-based goals would necessarily require more JavaScript in order to listen out for those clicks and fire off that information. The destination-based goals should be able to get all the information needed from regular page navigations.

So I have a hypothesis. I think that destination-based goals are less harmful to performance than event-based goals. I might well be wrong about that, and if I am, please let me know.

With that hypothesis in mind, and until I learn otherwise, I’ve got two rules of thumb to offer when it comes to using Google Analytics:

  1. Try to keep the number of goals to a minimum.
  2. If you must create a goal, favour destinations over events.

The dConstruct Audio Archive works offline

The dConstruct conference is as old as Clearleft itself. We put on the first event back in 2005, the year of our founding. The last dConstruct was in 2015. It had a good run.

I’m really proud of the three years I ran the show—2012, 2013, and 2014—and I have great memories from each event. I’m inordinately pleased that the individual websites are still online after all these years. I’m equally pleased with the dConstruct audio archive that we put online in 2012. Now that the event itself is no longer running, it truly is an archive—a treasury of voices from the past.

I think that these kinds of online archives are eminently suitable for some offline design. So I’ve added a service worker script to the dConstruct archive.

Caching

To start with, there’s the no-brainer: as soon as someone hits the website, pre-cache static assets like CSS, JavaScript, the logo, and icon images. Now subsequent page loads will be quicker—those assets are taken straight from the cache.

But what about the individual pages? For something like Resilient Web Design—another site that won’t be updated—I pre-cache everything. I could do that with the dConstruct archive. All of the pages with all of the images add up to less than two megabytes; the entire site weighs less than a single page on Wired.com or The Verge.

In the end, I decided to go with a cache-as-you-go strategy. Every time a page or an image is fetched from the network, it is immediately put in a cache. The next time that page or image is requested, the file is served from that cache instead of the network.

Here’s the logic for fetch requests:

  1. First, look to see if the file is in a cache. If it is, great! Serve that.
  2. If the file isn’t in a cache, make a network request and serve the response …but put a copy of a file in the cache.
  3. The next time that file is requested, go to step one.

Save for offline

That caching strategy works great for pages, images, and other assets. But there’s one kind of file on the dConstruct archive that’s a bit different: the audio files. They can be fairly big, so I don’t want to cache those unless the user specifically requests it.

If you end up on the page for a particular talk, and your browser supports service workers, you’ll get an additional UI element in the list of options: a toggle to “save offline” (under the hood, it’s a checkbox). If you activate that option, then the audio file gets put into a cache.

Now if you lose your network connection while browsing the site, you’ll get a custom offline page with the option to listen to any audio files you saved for offline listening. You’ll also see this collection of talks on the homepage, regardless of whether you’ve got an internet connection or not.

So if you’ve got a long plane journey ahead of you, have a browse around the dConstruct archive and select some talks for your offline listening pleasure.

Or just enjoy the speediness of browsing the site.

Turning another website into a Progressive Web App.