Tags: offline

17

sparkline

Going offline at Indie Web Camp Düsseldorf

I’ve just come back from a ten-day trip to Germany. The trip kicked off with Indie Web Camp Düsseldorf over the course of a weekend.

IndieWebCamp Düsseldorf 2017

Once again the wonderful people at Sipgate hosted us in their beautiful building, and once again myself and Aaron helped facilitate the two days.

IndieWebCamp Düsseldorf 2017

Saturday was the BarCamp-like discussion day. Plenty of interesting topics were covered. I led a session on service workers, and that’s also what I decided to work on for the second day—that’s when the talking is done and we get down to making.

IndieWebCamp Düsseldorf 2017 IndieWebCamp Düsseldorf 2017 IndieWebCamp Düsseldorf 2017 IndieWebCamp Düsseldorf 2017

I like what Ethan is doing on his offline page. He shows a list of pages that have been cached, but instead of just listing URLs, he shows a title and description for each page.

I’ve already got a separate cache for pages that gets added to as the user browses around my site. I needed to figure out a way to store the metadata for those pages so that I could then display it on the offline page. I came up with a workable solution, and interestingly, it involved no changes to the service worker script at all.

When you visit any blog post, I put metadata about the page into localStorage (after first checking that there’s an active service worker):

if (navigator.serviceWorker && navigator.serviceWorker.controller) {
  window.addEventListener('load', function() {
    var data = {
      "title": "A minority report on artificial intelligence",
      "description": "Revisiting Spielberg’s films after a decade and a half.",
      "published": "May 7th, 2017",
      "timestamp": "1494171049"
    };
    localStorage.setItem(
      window.location.href,
      JSON.stringify(data)
    );
  });
}

In my case, I’m outputting the metadata from the server, but you could just as easily grab some from the DOM like this:

var data = {
  "title": document.querySelector("title").innerText,
  "description": document.querySelector("meta[name='description']").getAttribute("contents")
}

Meanwhile in my service worker, when you visit that same page, it gets added to a cache called “pages”. Both localStorage and the cache API are using URLs as keys. I take advantage of that on my offline page.

The nice thing about writing JavaScript on my offline page is that I know the page will only be seen by modern browsers that support service workers, so I can use all sorts of fancy from ES6, or whatever we’re calling it now.

I start by looping through the keys of the “pages” cache (that’s right—the cache API isn’t just for service workers; you can access it from any script). Then I check to see if there is a corresponding localStorage key with the same string (a URL). If there is, I pull the metadata out of local storage and add it to an array called browsingHistory:

const browsingHistory = [];
caches.open('pages')
.then( cache => {
  cache.keys()
  .then(keys => {
    keys.forEach( request => {
      let data = JSON.parse(localStorage.getItem(request.url));
        if (data) {
          data['url'] = request.url;
          browsingHistory.push(data);
      }
    });

Then I sort the list of pages in reverse chronological order:

browsingHistory.sort( (a,b) => {
  return b.timestamp - a.timestamp;
});

Now I loop through each page in the browsing history list and construct a link to each URL, complete with title and description:

let markup = '';
browsingHistory.forEach( data => {
  markup += `
<h2><a href="${ data.url }">${ data.title }</a></h2>
<p>${ data.description }</p>
<p class="meta">${ data.published }</p>
`;
});

Finally I dump the constructed markup into a waiting div in the page with an ID of “history”:

let container = document.getElementById('history');
container.insertAdjacentHTML('beforeend', markup);

All those steps need to be wrapped inside the then clause attached to caches.open("pages") because the cache API is asynchronous.

There you have it. Now if you’re browsing adactio.com and your network connection drops (or my server goes offline), you can choose from a list of pages you’ve previously visited.

The current situation isn’t ideal though. I’ve got a clean-up operation in my service worker to limit the number of items stored in my “pages” cache. The cache never gets bigger than 35 items. But there’s no corresponding clean-up of metadata stored in localStorage. So there could be a lot more bits of metadata in local storage than there are pages in the cache. It’s not harmful, but it’s a bit wasteful.

I can’t do a clean-up of localStorage from my service worker because service workers can’t access localStorage. There’s a very good reason for that: the localStorage API is synchronous, and everything that happens in a service worker needs to be asynchronous.

Service workers can access indexedDB: it’s asynchronous. I could use indexedDB instead of localStorage, but I’m not a masochist. My best bet would be to use the localForage library, which wraps indexedDB in the simple syntax of localStorage.

Maybe I’ll do that at the next Homebrew Website Club here in Brighton.

Progressive Web App questions

I got a nice email recently from Colin van Eenige. He wrote:

For my graduation project I’m researching the development of Progressive Web Apps and found your offline book called resilient web design. I was very impressed by the implementation of the website and it really was a nice experience.

I’m very interested in your vision on progressive web apps and what capabilities are waiting for us regarding offline content. Would it be fine if I’d send you some questions?

I said that would be fine, although I couldn’t promise a swift response. He sent me four questions. I finally got ‘round to sending my answers…

1. https://resilientwebdesign.com/ is an offline web book (progressive web app). What was the primary reason make it available like this (besides the other formats)?

Well, given the subject matter, it felt right that the canonical version of the book should be not just online, but made with the building blocks of the web. The other formats are all nice to have, but the HTML version feels (to me) like the “real” book.

Interestingly, it wasn’t too much trouble for people to generate other formats from the HTML (ePub, MOBI, PDF), whereas I think trying to go in the other direction would be trickier.

As for the offline part, that felt like a natural fit. I had already done that with a previous book of mine, HTML5 For Web Designers, which I put online a year or two after its print publication. In that case, I used AppCache for the offline functionality. AppCache is horrible, but this use case might be one of the few where it works well: a static book that’s never going to change. Cache invalidation is one of the worst parts of using AppCache so by not having any kinds of updates at all, I dodged that bullet.

But when it came time for Resilient Web Design, a service worker was definitely the right technology. Still, I’ve got AppCache in there as well for the browsers that don’t yet support service workers.

2. What effect you you think Progressive Web Apps will have on content consuming and do you think these will take over the purpose of some Native Apps?

The biggest effect that service workers could have is to change the expectations that people have about using the web, especially on mobile devices. Right now, people associate the web on mobile with long waits and horrible spammy overlays. Service workers can help solve that first part.

If people then start adding sites to their home screen, that will be a great sign that the web is really holding its own. But I don’t think we should get too optimistic about that: for a user, there’s no difference between a prompt on their screen saying “add to home screen” and a prompt on their screen saying “download our app”—they’re equally likely to be dismissed because we’ve trained people to dismiss anything that covers up the content they actually came for.

It’s entirely possible that websites could start taking over much of the functionality that previously was only possible in a native app. But I think that inertia and habit will keep people using native apps for quite some time.

The big exception is in markets where storage space on devices is in short supply. That’s where the decision to install a native app isn’t taken likely (given the choice between your family photos and an app, most people will reject the app). The web can truly shine here if we build lightweight, performant services.

Even in that situation, I’m still not sure how many people will end up adding those sites to their home screen (it might feel so similar to installing a native app that there may be some residual worry about storage space) but I don’t think that’s too much of a problem: if people get to a site via search or typing, that’s fine.

I worry that the messaging around “progressive web apps” is perhaps over-fetishising the home screen. I don’t think that’s the real battleground. The real battleground is in people’s heads; how they perceive the web and how they perceive native.

After all, if the average number of native apps installed in a month is zero, then that’s not exactly a hard target to match. :-)

3. What is your vision regarding Progressive Web Apps?

For me, progressive web apps don’t feel like a separate thing from making websites. I worry that the marketing of them might inflate expectations or confuse people. I like the idea that they’re simply websites that have taken their vitamins.

So my vision for progressive web apps is the same as my vision for the web: something that people use every day for all sorts of tasks.

I find it really discouraging that progressive web apps are becoming conflated with single page apps and the app shell model. Those architectural decisions have nothing to do with service workers, HTTPS, and manifest files. Yet I keep seeing the concepts used interchangeably. It would be a real shame if people chose not to use these great technologies just because they don’t classify what they’re building as an “app.”

If anything, it’s good ol’ fashioned content sites (newspapers, wikipedia, blogs, and yes, books) that can really benefit from the turbo boost of service worker+HTTPS+manifest.

I was at a conference recently where someone was given a talk encouraging people to build progressive web apps but discouraging people from doing it for their own personal sites. That’s a horrible, elitist attitude. I worry that this attitude is being codified in the term “progressive web app”.

4. What is the biggest learning you’ve had since working on Progressive Web Apps?

Well, like I said, I think that some people are focusing a bit too much on the home screen and not enough on the benefits that service workers can provide to just about any website.

My biggest learning is that these technologies aren’t for a specific subset of services, but can benefit just about anything that’s on the web. I mean, just using a service worker to explicitly cache static assets like CSS, JS, and some images is a no-brainer for almost any project.

So there you go—I’m very excited about the capabilities of these technologies, but very worried about how they’re being “sold”. I’m particularly nervous that in the rush to emulate native apps, we end up losing the very thing that makes the web so powerful: URLs.

Small steps

The new Clearleft website is live! Huzzah!

Many people have been working very hard on it and it’s all looking rather nice. But, as I said before, the site launch isn’t the end—it’s just the beginning.

There are some obvious next steps: fixing bugs, adding content, tweaking copy, and, oh yeah, that whole “testing with real users” thing. But there’s also an opportunity to have some fun on the front end. Now that the site is out there in the wild, there’s a real incentive to improve its performance.

Off the top of my head, these are some areas where I think we can play around:

  • Font loading. Right now the site is just using @font-face. A smart font-loading strategy—at least for the body copy—could really help improve the perceived performance.
  • Responsive images. A long-term solution will require some wrangling on the back end, but I reckon we can come up with some way of generating different sized images to reference in srcset.
  • Service worker. It’s a no-brainer. Now that the Clearleft site is (finally!) running on HTTPS, having a simple service worker to cache static assets like CSS, JavaScript and some images seems like the obvious next step. The question is: what other offline shenanigans could we get up to?

I’m looking forward to tinkering with some of those technologies. Each one should make an incremental improvement to the site’s performance. There are already some steps on the back-end that are making a big difference: upgrading to PHP7 and using HTTP2.

Now the real fun begins.

Making Resilient Web Design work offline

I’ve written before about taking an online book offline, documenting the process behind the web version of HTML5 For Web Designers. A book is quite a static thing so it’s safe to take a fairly aggressive offline-first approach. In fact, a static unchanging book is one of the few situations that AppCache works for. Of course a service worker is better, but until AppCache is removed from browsers (and until service worker is supported across the board), I’m using both. I wouldn’t recommend that for most sites though—for most sites, use a service worker to enhance it, and avoid AppCache like the plague.

For Resilient Web Design, I took a similar approach to HTML5 For Web Designers but I knew that there was a good chance that some of the content would be getting tweaked at least for a while. So while the approach is still cache-first, I decided to keep the cache fairly fresh.

Here’s my service worker. It starts with the usual stuff: when the service worker is installed, there’s a list of static assets to cache. In this case, that list is literally everything; all the HTML, CSS, JavaScript, and images for the whole site. Again, this is a pattern that works well for a book, but wouldn’t be right for other kinds of websites.

The real heavy lifting happens with the fetch event. This is where the logic sits for what the service worker should do everytime there’s a request for a resource. I’ve documented the logic with comments:

// Look in the cache first, fall back to the network
  // CACHE
  // Did we find the file in the cache?
      // If so, fetch a fresh copy from the network in the background
      // NETWORK
          // Stash the fresh copy in the cache
  // NETWORK
  // If the file wasn't in the cache, make a network request
      // Stash a fresh copy in the cache in the background
  // OFFLINE
  // If the request is for an image, show an offline placeholder
  // If the request is for a page, show an offline message

So my order of preference is:

  1. Try the cache first,
  2. Try the network second,
  3. Fallback to a placeholder as a last resort.

Leaving aside that third part, regardless of whether the response is served straight from the cache or from the network, the cache gets a top-up. If the response is being served from the cache, there’s an additional network request made to get a fresh copy of the resource that was just served. This means that the user might be seeing a slightly stale version of a file, but they’ll get the fresher version next time round.

Again, I think this acceptable for a book where the tweaks and changes should be fairly minor, but I definitely wouldn’t want to do it on a more dynamic site where the freshness matters more.

Here’s what it usually likes like when a file is served up from the cache:

caches.match(request)
  .then( responseFromCache => {
  // Did we find the file in the cache?
  if (responseFromCache) {
      return responseFromCache;
  }

I’ve introduced an extra step where the fresher version is fetched from the network. This is where the code can look a bit confusing: the network request is happening in the background after the cached file has already been returned, but the code appears before the return statement:

caches.match(request)
  .then( responseFromCache => {
  // Did we find the file in the cache?
  if (responseFromCache) {
      // If so, fetch a fresh copy from the network in the background
      event.waitUntil(
          // NETWORK
          fetch(request)
          .then( responseFromFetch => {
              // Stash the fresh copy in the cache
              caches.open(staticCacheName)
              .then( cache => {
                  cache.put(request, responseFromFetch);
              });
          })
      );
      return responseFromCache;
  }

It’s asynchronous, see? So even though all that network code appears before the return statement, it’s pretty much guaranteed to complete after the cache response has been returned. You can verify this by putting in some console.log statements:

caches.match(request)
.then( responseFromCache => {
  if (responseFromCache) {
      event.waitUntil(
          fetch(request)
          .then( responseFromFetch => {
              console.log('Got a response from the network.');
              caches.open(staticCacheName)
              .then( cache => {
                  cache.put(request, responseFromFetch);
              });
          })
      );
      console.log('Got a response from the cache.');
      return responseFromCache;
  }

Those log statements will appear in this order:

Got a response from the cache.
Got a response from the network.

That’s the opposite order in which they appear in the code. Everything inside the event.waitUntil part is asynchronous.

Here’s the catch: this kind of asynchronous waitUntil hasn’t landed in all the browsers yet. The code I’ve written will fail.

But never fear! Jake has written a polyfill. All I need to do is include that at the start of my serviceworker.js file and I’m good to go:

// Import Jake's polyfill for async waitUntil
importScripts('/js/async-waituntil.js');

I’m also using it when a file isn’t found in the cache, and is returned from the network instead. Here’s what the usual network code looks like:

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

I want to also store that response in the cache, but I want to do it asynchronously—I don’t care how long it takes to put the file in the cache as long as the user gets the response straight away.

Technically, I’m not putting the response in the cache; I’m putting a copy of the response in the cache (it’s a stream, so I need to clone it if I want to do more than one thing with it).

fetch(request)
  .then( responseFromFetch => {
    // Stash a fresh copy in the cache in the background
    let responseCopy = responseFromFetch.clone();
    event.waitUntil(
      caches.open(staticCacheName)
      .then( cache => {
          cache.put(request, responseCopy);
      })
    );
    return responseFromFetch;
  })

That all seems to be working well in browsers that support service workers. For legacy browsers, like Mobile Safari, there’s the much blunter caveman logic of an AppCache manifest.

Here’s the JavaScript that decides whether a browser gets the service worker or the AppCache:

if ('serviceWorker' in navigator) {
  // If service workers are supported
  navigator.serviceWorker.register('/serviceworker.js');
} else if ('applicationCache' in window) {
  // Otherwise inject an iframe to use appcache
  var iframe = document.createElement('iframe');
  iframe.setAttribute('src', '/appcache.html');
  iframe.setAttribute('style', 'width: 0; height: 0; border: 0');
  document.querySelector('footer').appendChild(iframe);
}

Either way, people are making full use of the offline nature of the book and that makes me very happy indeed.

Choice

Laurie Voss has written a thoughtful article called Web development has two flavors of graceful degradation in response to Nolan Lawson’s recent article. But I’m afraid I don’t agree with Laurie’s central premise:

…web app development and web site development are so different now that they probably shouldn’t be called the same thing anymore.

This is an idea I keep returning to, and each time I do, I find that it just isn’t that simple. There are very few web thangs that are purely interactive without any content, and there are also very few web thangs that are purely passive without any interaction. Instead, it’s a spectrum. Quite often, the position on that spectrum changes according to the needs of the user at any particular time—are Twitter and Flicker web sites while I’m viewing text and images, but then transmogrify into web apps the moment I want add, update, or delete a piece of text or an image?

In any case, the more interesting question than “is something a web site or a web app?” is the question “why?” Why does it matter? In my experience, the answer to that question generally comes down to the kind of architectural approach that a developer will take.

That’s exactly what Laurie dives into in his post. For web apps, use one architectural approach—for web sites, use a different architectural approach. To summarise:

  • in a web app, front-load everything and rely on client-side JavaScript for all subsequent interaction,
  • in a web site, optimise for many page loads, and make sure you don’t rely on client-side JavaScript.

I’m oversimplifying here, but the general idea is:

  • build web apps with the single page app architecture,
  • build web sites with progressive enhancement.

That’s sensible advice, but I’m worried that it could lead to a tautological definition of what constitutes a web app:

  1. This is a web app so it’s built as a single page app.
  2. Why do you define it as a web app?
  3. Because it’s built as a single page app.

The underlying question of what makes something a web app is bypassed by the architectural considerations …but the architectural considerations should be based on that underlying question. Laurie says:

If you are developing an app, the user ideally loads the app exactly once — whether it’s over a slow connection or not.

And similarly:

But if you are developing a web site consisting of many discrete pages, the act of loading goes from a single event to the most common event.

I completely agree that the architectural approach of single page apps is better suited to some kinds of web thangs more than others. It’s a poor architectural choice for a content-based site like nasa.gov, for example. Progressive enhancement would make more sense there.

But I don’t think that the architectural choices need to be in opposition. It’s entirely possible to reconcile the two. It’s not always easy—and the further along that spectrum you are, the tougher it gets—but it’s doable. You can begin with progressive enhancement, and then build up to a single page app architecture for more capable browsers.

I think that’s going to get easier as frameworks adopt a more mixed approach. Almost all the major libraries are working on server-side rendering as a default. Ember is leading the way with FastBoot, and Angular Universal is following. Neither of them are doing it for reasons of progressive enhancement—they’re doing it for performance and SEO—but the upshot is that you can more easily build a web app that simultaneously uses progressive enhancement and a single-page app model.

I guess my point is that I don’t think we should get too locked into the idea of web apps and web sites requiring fundamentally different approaches, especially with the changes in the technologies we used to build them.

We’ve made the mistake in the past of framing problems as “either/or”, when in fact, the correct solution was “both!”:

  • you can either have a desktop site or a mobile site,
  • you can either have rich interactivity or accessibility,
  • you can either have a single page app or progressive enhancement.

We don’t have to choose. It might take more work, but we can have our web cake and eat it.

The false dichotomy that I’m most concerned about is the pernicious idea that offline functionality is somehow in opposition to progressive enhancement. Given the design of service workers, I find this proposition baffling.

This remark by Tom is the very definition of a false dichotomy:

People who say your site should work without JavaScript are actually hurting the people they think they’re helping.

He was also linking to Nolan’s article, which could indeed be read as saying that you should for offline instead of building with progressive enhancement. But I don’t think that’s what Nolan is saying (at least, I sincerely hope not). I think that Nolan is saying that we should prioritise the offline scenario over scenarios where JavaScript fails or isn’t available. That’s a completely reasonable thing to say. But the idea that we should build for the offline scenario instead of scenarios where JavaScript fails is absurdly reductionist. We don’t have to choose!

But I can certainly understand how developers might come to be believe that building a progressive web app is at odds with progressive enhancement. Having made a bunch of progressive web apps—Huffduffer, The Session, this site, I can testify that service workers work superbly as a layer on top of an existing site, but all the messaging around progressive web apps seems to fixated on the idea of the app-shell model (a small tweak to the single page app model, where a little bit of interface is available on the initial page load instead of requiring JavaScript for absolutely everything). Again, it’s entirely possible to reconcile the app-shell approach with server rendering and progressive enhancement, but nobody seems to be talking about that. Instead, all of the examples and demos are built with an assumption about JavaScript availability.

Assumptions are the problem. Whether it’s assumptions about screen size, assumptions about being able-bodied, assumptions about network connectivity, or assumptions about browser capabilities, I don’t think any assumptions are a safe bet. Now you might quite reasonably say that we have to make some assumptions when we’re building on the web, and you’d be right. But I think we should still aim to keep them to a minimum.

Tom’s tweet included a screenshot of this part of Nolan’s article:

As Benedict Evans has noted, the next billion people who are poised to come online will be using the internet almost exclusively through smartphones. And if Google’s plans with Android One are any indication, then we have a fairly good idea of what kind of devices the “next billion” will be using:

  • They’ll mostly be running Android.
  • They’ll have decent specs (1GB RAM, quad-core processors).
  • They’ll have an evergreen browser and WebView (Android 5+).
  • What they won’t have, however, is a reliable internet connection.

Those seem like a reasonable set of assumptions. But even there, things aren’t so simple. Will people really be using “an evergreen browser and WebView”? Millions of people use proxy browsers like Opera Mini, which means you can’t guarantee JavaScript availability beyond the initial page load. UC Browser—which can also run in proxy mode—is now the second most popular mobile browser in the world.

That’s just one nit-picky example, but what I’m getting at here is that it really isn’t safe to make any assumptions. When we must make assumptions, let’s try to make them a last resort.

And just to be clear here, I’m not saying that just because we can’t make assumptions about devices or browsers doesn’t mean that we can’t build rich interactive web apps that work offline. I’m saying that we can build rich interactive web apps that work offline and also work when JavaScript fails or isn’t supported.

You don’t have to choose between progressive enhancement and a single page app/progressive web app/app shell/other things with the word “app”.

Progressive enhancement is an architectural approach to building on the web. You don’t have to use it, but please try to remember that it is your choice to make. You can choose to build a web app using progressive enhancement or not—there is nothing inherent in the nature of the thing you’re building that precludes progressive enhancement.

Personally, I find progressive enhancement a sensible way to counteract any assumptions I might inadvertently make. Progressive enhancement increases the chances that the web site (or web app) I’m building is resilient to the kind of scenarios that I never would’ve predicted or anticipated.

That’s why I choose to use progressive enhancement …and build progressive web apps.

Backdoor Service Workers

When I was moderating that panel at the Progressive Web App dev Summit, I brought up this point about twenty minutes in:

Alex, in your talk yesterday you were showing the AMP demo there with the Washington Post. You click through and there’s the Washington Post AMP thing, and it was able to install the Service Worker with that custom element. But I was looking at the URL bar …and that wasn’t the Washington Post. It was on the CDN from AMP. So I talked to Paul Backaus from the AMP team, and he explained that it’s an iframe, and using an iframe you can install a Service Worker from somewhere else.

Alex and Emily explained that, duh, that’s the way iframes work. It makes sense when you think about it—an iframe is pretty much the same as any other browser window. Still, it feels like it might violate the principle of least surprise.

Let’s say you followed my tongue-in-cheek advice to build a progressive web app store. Your homepage might have the latest 10 or 20 progressive web apps. You could also include 10 or 20 iframes so that those sites are “pre-installed” for the person viewing your page.

Enough theory. Here’s a practical example…

Suppose you’ve never visited the website for my book, html5forwebdesigners.com (if you have visited it, and you want to play along with this experiment, go to your browser settings and delete anything stored by that domain).

You happen to visit my website adactio.com. There’s a little blurb buried down on the home page that says “Read my book” with a link through to html5forwebdesigners.com. I’ve added this markup after the link:

<iframe src="https://html5forwebdesigners.com/iframe.html" style="width: 0; height: 0; border: 0">
</iframe>

That hidden iframe pulls in an empty page with a script element:

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<title>HTML5 For Web Designers</title>
<script>
if ('serviceWorker' in navigator) {
  navigator.serviceWorker.register('/serviceworker.js');
}
</script>
</head>
</html>

That registers the Service Worker on my book’s site which then proceeds to install all the assets it needs to render the entire site offline.

There you have it. Without ever visiting the domain html5forwebdesigners.com, the site has been pre-loaded onto your device because you visited the domain adactio.com.

A few caveats:

  1. I had to relax the Content Security Policy for html5forwebdesigners.com to allow the iframe to be embedded on adactio.com:

    Header always set Access-Control-Allow-Origin: "https://adactio.com"
    
  2. If your browser’s settings has “Block third-party cookies and site data” selected in the preferences, the iframe-invoked Service Worker won’t install:

    Uncaught (in promise) DOMException: Failed to register a ServiceWorker: The user denied permission to use Service Worker.
    

The example I’ve put together here is relatively harmless. But it’s possible to imagine more extreme scenarios. Imagine there’s a publishing company that has 50 websites for 50 different publications. Each one of them could have an empty page waiting to be embedded via iframe from the other 49 sites. You only need to visit one page on one of those 50 sites to have 50 Service Workers spun up and caching assets in the background.

There’s the potential here for a tragedy of the commons. I hope we’ll be sensible about how we use this power.

Just don’t tell the advertising industry about this.

The Progressive Web App Dev Summit

I was in Amsterdam again at the start of last week for the Progressive Web App Dev Summit, organised by Google. Most of the talks were given by Google employees, but not all—this wasn’t just a European version of Google I/O. Representatives from Opera, Mozilla, Samsung, and Microsoft were also there, and there were quite a few case studies from independent companies. That was very gratifying to see.

Almost all the talks were related to progressive web apps. I say, “almost all” because there were occasional outliers. There was a talk on web components, which don’t have anything directly to do with progressive web apps (and I hope there won’t be any attempts to suggest otherwise), and another on rendering performance that had good advice for anyone building any kind of website. Most of the talks were about the building blocks of progressive web apps: HTTPS, Service Workers, push notifications, and all that jazz.

I was very pleased to see that there was a move away from the suggesting that single-page apps with the app-shell architecture model were the only way of building progressive web apps.

There were lots of great examples of progressively enhancing existing sites into progressive web apps. Jeff Posnick’s talk was a step-by-step walkthrough of doing exactly that. Reading through the agenda, I was really happy to see this message repeated again and again:

In this session we’ll take an online-only site and turn it into a fully network-resilient, offline-first installable progressive web app. We’ll also break out of the app shell and look at approaches that better-suit traditional server-driven sites.

Progressive Web Apps should work everywhere for every user. But what happens when the technology and API’s are not available for in your users browser? In this talk we will show you how you can think about and build sites that work everywhere.

Progressive Web Apps should load fast, work great offline, and progressively enhance to a better experience in modern browsers.

How do you put the “progressive” into your current web app?

You can (and should!) build for the latest and greatest browsers, but through a collection of fallbacks and progressive enhancements you can bring a lot tomorrow’s web to yesterday’s browsers.

I think this is a really smart move. It’s a lot easier to sell people on incremental changes than it is to convince them to rip everything out and start from scratch (another reason why I’m dubious about any association between web components and progressive web apps—but I’ll save that for another post).

The other angle that I really liked was the emphasis on emerging markets, not just wealthy westerners. Tal Oppenheimer’s talk Building for Billions was superb, and Alex kicked the whole thing off with some great facts and figures on mobile usage.

In my mind, these two threads are very much related. Progressive enhancement allows us to have our progressive web app cake and eat it too: we can make websites that can be accessed on devices with limited storage and slow networks, while at the same time ensuring those same sites take advantage of all the newest features in the latest and greatest browsers. I talked to a lot of Google devs about ways to measure the quality of a progressive web app, and I’m coming to the conclusion that a truly high-quality site is one that can still be accessed by a proxy browser like Opera Mini, while providing a turbo-charged experience in the latest version of Chrome. If you think that sounds naive or unrealistic, then I think you might want to dive deeper into all the technologies that make progressive web apps so powerful—responsive design, Service Workers, a manifest file, HTTPS, push notifications; all of those features can and should be used in a layered fashion.

Speaking of Opera, Andreas kind of stole the show, demoing the latest interface experiments in Opera Mobile.

That ambient badging that Alex was talking about? Opera is doing it. The importance of being able to access URLs that I’ve been ranting about? Opera is doing it.

Then we had the idea to somehow connect it to the “pull-to-refresh” spinner, as a secondary gesture to the left or right.

Nice! I’m looking forward to seeing what other browsers come up with it. It’s genuinely exciting to see all these different browser makers in complete agreement on which standards they want to support, while at the same time differentiating their products by competing on user experience. Microsoft recently announced that progressive web apps will be indexed in their app store just like native apps—a really interesting move.

The Progressive Web App Dev Summit wrapped up with a closing panel, that I had the honour of hosting. I thought it was very brave of Paul to ask me to host this, considering my strident criticism of Google’s missteps.

Initially there were going to be six people on the panel. Then it became eight. Then I blinked and it suddenly became twelve. Less of a panel, more of a jury. Half the panelists were from Google and the other half were from Opera, Microsoft, Mozilla, and Samsung. Some of those representatives were a bit too media-trained for my liking: Ali from Microsoft tried to just give a spiel, and Alex Komoroske from Google wouldn’t give me a straight answer about whether he wants Android Instant apps to succeed—Jake was a bit more honest. I should have channelled my inner Paxman a bit more.

Needless to say, nobody from Apple was at the event. No surprise there. They’ve already promised to come to the next event. There won’t be an Apple representative on stage, obviously—that would be asking too much, wouldn’t it? But at least it looks like they’re finally making an effort to engage with the wider developer community.

All in all, the Progressive Web App Dev Summit was good fun. I found the event quite inspiring, although the sausage festiness of the attendees was depressing. It would be good if the marketing for these events reached a wider audience—I met a lot of developers who only found out about it a week or two before the event.

I really hope that people will come away with the message that they can get started with progressive web apps right now without having to re-architect their whole site. Right now the barrier to entry is having your site running on HTTPS. Once you’ve got that up and running, it’s pretty much a no-brainer to add a manifest file and a basic Service Worker—to boost performance if nothing else. From there, you’re in a great position to incrementally add more and more features—an offline-first approach with your Service Worker, perhaps? Or maybe start dabbling in push notifications. The great thing about all of these technologies (with the glaring exception of web components in their current state) is that you don’t need to bet the farm on any of them. Try them out. Use them as enhancements. You’ve literally got nothing to lose …and your users have everything to gain.

Taking an online book offline

Application Cache is—as Jake so infamously described—not a good API. It was specced and shipped before developers had a chance to figure out what they really needed, and so AppCache turned out to be frustrating at best and downright dangerous in some situations. Its over-zealous caching combined with its byzantine cache invalidation ensured it was never going to become a mainstream technology.

There are very few use-cases for AppCache, but I think I hit upon one of them. Six years ago, A Book Apart published HTML5 For Web Designers. A year and a half later, I put the book online. The contents are never going to change. There’s a second edition of the book out now but if you want to read all the extra bits that Rachel added, you’re going to have to buy the book. The website for the original book is static and unchanging. That’s what made it such a good candidate for using AppCache. I could just set it and forget.

Except that’s no longer true. AppCache is being deprecated and browsers are starting to withdraw support. Chrome is already making sure that AppCache—like geolocation—no longer works on sites that aren’t served over HTTPS. That’s for the best. In retrospect, those APIs should never have been allowed over unsecured HTTP.

I mentioned that I spent the weekend switching all my book websites over to HTTPS, so AppCache should continue to work …for now. It’s only a matter of time before AppCache is removed completely from many of the browsers that currently support it.

Seeing as I’ve got the HTML5 For Web Designers site running on HTTPS now, I might as well go all out and make it a progressive web app. By far the biggest barrier to making a progressive web app is that first step of setting up HTTPS. It’s gotten cheaper—thanks to Let’s Encrypt Certbot—but it still involves mucking around in the command line with root access; I never wanted to become a sysadmin. But once that’s finally all set up, the other technological building blocks—a Service Worker and a manifest file—are relatively easy.

In this case, the Service Worker is using a straightforward bit of logic:

  • On installation, cache absolutely everything: HTML, CSS, images.
  • When anything is requested, grab it from the cache.
  • If it isn’t in the cache, try the network.
  • If the network doesn’t work, show an offline page (or image).

Basically I’m reproducing AppCache’s overzealous approach. It works for this site because the content is never going to change. I hope that this time, I really can just set it and forget it. I want the site to be an historical artefact, available at the same URL for at least my lifetime. I don’t want to have to maintain it or revisit it every few years to swap out one API for another.

Which brings me back to the way AppCache is being deprecated…

The Firefox team are very eager to ditch AppCache as soon as possible. On the one hand, that’s commendable. They’re rightly proud of shipping Service Workers and they want to encourage people to use the better technology instead. But it sure stings for the suckers (like me) who actually went and built stuff using AppCache.

In a weird way, I think this rush to deprecate AppCache might actually hurt the adoption of Service Workers. Let me explain…

At last year’s Edge Conference, Nolan Lawson gave a great presentation on storing data in the browser. He enumerated the many ways—past and present—that we could store data locally: WebSQL, Local Storage, IndexedDB …the list goes on. He also posed the question: why aren’t more people using insert-name-of-latest-API-here? To me it seemed obvious why more people weren’t diving into using the latest and greatest option for local data storage. It was because they had been burned before. The developers who rushed into trying previous solutions end up being mocked for their choice. “Still using that ol’ thing? Pffftt!”

You can see that same attitude on display from Mozilla as they push towards removing AppCache. Like in a comment that refers to developers using AppCache in production as “the angry hordes”. Reminds me of something Tom said:

In that same Mozilla thread, Soledad echoes Tom’s point:

As a member of the devrel team: I think that this should be better addressed in a blog post that someone from the team responsible for switching AppCache off should write, so everyone can understand the reasons and ask questions to those people.

I’d rather warn people beforehand, pointing them to that post and help them with migration paths than apply emergency mitigation strategies when a lot of people find their stuff stopped working in the newer Firefox…

Bravo! That same approach should have also been taken by the Chrome team when it came to their thread about punishing display:browser in manifest files. There was absolutely no communication with developers about this major decision. I only found out about it because Paul happened to mention it to me.

I was genuinely shocked by this:

Withholding the “add to home screen” prompt like that has a whiff of blackmail about it.

I can confirm that smell. When I was making the manifest file for HTML5 For Web Designers, I really wanted to put display: browser because I want people to be able to copy and paste URLs (for the book, for individual chapters, and for sections within chapters). But knowing that if I did that, Android users would never see the “add to home screen” prompt made me question that decision. I felt strong-armed into declaring display: standalone. And no, I’m not mollified by hand-waving reassurances that the Chrome team will figure out some solution for this. Figure out the solution first, then punish the saps like me who want to use display: browser to allow people to share URLs.

Anyway, the website for HTML5 For Web Designers is now using AppCache and Service Workers. The AppCache part will probably be needed for quite a while yet to provide offline support on iOS. Apple are really dragging their heels on Service Worker support, with at least one WebKit engineer actively looking for reasons not to implement it.

There’s a lot of talk about making apps work offline, but I think it’s just as important that we consider making information work offline. Books are a great example of this. To use the tired transport tropes, the website for a book is something you might genuinely want to access when you’re on a plane, or in the underground, or out at sea.

I really, really like progressive web apps. But I also think it’s important that we don’t fall into the trap of just trying to imitate native apps on the web. I love the idea of taking the best of the web—like information being permanently available at a URL—and marrying that up with the best of native—like offline access. I also like the idea of taking the best of books—a tome of thought—and marrying it up with the best of the web—hypertext.

I’d love to see more experimentation around online/offline hypertext/books. For now, you can visit HTML5 For Web Designers, add it to your home screen, and revisit it whenever and wherever you like.

A web for everyone

I gave the closing talk at the Render conference in Oxford a few weeks back. It was a very smoothly-run event, the spiritual successor to jQuery UK.

In amongst the mix of talks there were a few emerging themes. Animation was covered from a few different angles by Val and Sara. Bruce, Jake, Ola, and I talked about Service Workers and offline functionality. But there were also some differences of opinion.

In her great talk—I’m Offline, Cool! Now What?—Ola outlined the many and varied offline use cases that drove the creation and philosophy of Hoodie. She described all the reasons why people need the web; for communication, for access to information, for empowerment, and for love. “Hell, yes!” I thought.

But then she said:

So since when is helping people to fulfil a basic need, progressive enhancement?

And even more forcefully:

This is why I think, putting offline first in the progressive enhancement slot is pure bullshit.

Strong words indeed! And I have to say I was a little puzzled by them.

Ola had demonstrated again and again just how fragile the network could be. That is absolutely correct. All too often, we make the assumption that people using our sites have a decent network connection. That’s not a safe assumption to make.

But the suggested solution—to rely on technologies like local storage, Service Workers, or other APIs—assumes a certain level of JavaScript capabilities in the devices and browsers out there. That’s an unsafe assumption to make.

I remember discussing this with Alex from Hoodie a while back. I was confused by the cognitive dissonance I was observing. It seems to me that, laudable as Hoodie’s offline-first goals are, they are swapping out one unstable dependency—the network—for a different unstable dependency—a set of JavaScript APIs.

(I remember Alex pointed out that Hoodie was intended primarily for web apps rather than web sites, and my response—predictably enough—was to say “Define web app”.)

I think I understand why Ola reacted so strongly to the suggestion that offline functionality should be added as an enhancement. I’ve seen the same reaction when I’ve said that beautiful typography on the web is an enhancement. I think that when I say something is an enhancement, what people hear is that something is just an enhancement. It sounds belittling. That’s not my intention, but I can understand how it could come across like that. Perhaps this is one reason why some people have a real issue with the term “progressive enhancement”.

I wish we could make offline functionality a requirement. But the reality is that not everyone is using a browser that supports the necessary technology. I wish we could make beautiful typography a requirement. But, again, the reality is that there will always be some browsers or devices that won’t be capable of executing that typography. Accepting these facets of reality might seem like admissions of defeat, but I actually find it quite liberating.

In her brilliant talk at Render, Ashley G. Williams channeled Carl Sagan, quoting from his book The Demon-Haunted World:

It is far better to grasp the universe as it really is than to persist in delusion, however satisfying and reassuring.

That’s how I feel we should approach building for the web. Let’s accept that network connections are unevenly distributed. Let’s also accept that browser features are unevenly distributed. Pretending that millions of Opera Mini users don’t exist isn’t a viable strategy. They too are people who want to communicate, to access information, to be empowered, and to love.

Pointing out that you can’t always rely on client-side JavaScript shouldn’t be taken as an admonishment. It’s an opportunity.

Karolina Szczur wrote a wonderful piece on Ev’s blog called The Web Isn’t Uniform. She noticed how many sites—Facebook, AirBnB, Basecamp—failed to even render some useful information if the JavaScript fails to load. It’s a situation that many of us—with our fast connections, capable browsers, and modern devices—might never even notice.

It’s a privilege to be able to use breaking edge technologies and devices, but let’s not forget basic accessibility and progressive enhancement. Ultimately, we’re building for the users, not for our own tastes or preferences.

Karolina asks that we, as makers of the web, have a little more empathy. If the comments on her article are anything to go by, that’s a tall order. All the usual tropes are rolled out—there’s the misunderstanding that progressive enhancement means making sure everything works without JavaScript (it doesn’t; it’s about the core functionality), and the evergreen argument that as soon as you’re building a web “app”, that best practices, good engineering, and empathy can go out the window…

I strongly disagree that this has anything to do at all about empathy. Instead, it’s all about resources and priorities. Making a JS app is already hard enough, duplicating all that work so that it also works without JS is quite often just not practical.—Sacha Greif

But requiring that a site be functional when JavaScript is disabled, may not be a valid requirement anymore. HTML and CSS were originally created and designed for documents, not applications. Many websites these days should be considered apps rather then docs.—Dan Shappir

What you’re suggesting is that all these companies should write all their software twice, once in javascript and again in good ol’ html with forms, to cater to that point-whatever-percentage that has decided to break their own web browser by turning one of the three fundamental web technologies off. In what universe is this a reasonable request?—Erlend Halvorsen

JavaScript is as important as HTML. This is modern internet. If someone doesn’t have JavaScript, they should not be using the new applications that were possible because of JavaScript.—HarshaL

I am a web developer. I build web applications not web sites. What you say may be true for web sites with static pages displaying images and text.—R. Fancsiki

Ah, Medium! Where the opinions of self-entitled dudes flow like rain from the tech heavens.

While they were so busy defending the lack of basic functionality in all the examples that Karolina listed, they failed to notice the most important development:

Let’s build a web that works for everyone. That doesn’t mean everyone has to have the same experience. Let’s accept that there are all sorts of people out there accessing the web with all sorts of browsers on all sorts of devices.

What a fantastic opportunity!

Handling redirects with a Service Worker

When I wrote about implementing my first Service Worker, I finished with this plea:

And remember, please share your code and your gotchas: it’s early days for Service Workers so every implementation counts.

Well, I ran into a gotcha that was really frustrating but thanks to the generosity of others, I was able to sort it out.

It was all because of an issue in Chrome. Here’s the problem…

Let’s say you’ve got a Service Worker running that takes care of any requests to your site. Now on that site, you’ve got a URL that receives POST data, does something with it, and then redirects to another URL. That’s a fairly common situation—it’s how I handle webmentions here on adactio.com, and it’s how I handle most add/edit/delete actions over on The Session to help prevent duplicate form submissions.

Anyway, it turns out that Chrome’s Service Worker implementation would get confused by that. Instead of redirecting, it showed the offline page instead. The fetch wasn’t resolving.

I described the situation to Jake, but rather than just try and explain it in 140 characters, I built a test case.

There’s a Chromium issue filed on this, and it will get fixed, but it in the meantime, it was really bugging me recently when I was rolling out a new feature on The Session. Matthew pointed out that the Chromium bug report also contained a workaround that he’s been using on traintimes.org.uk. Adrian also posted his expanded workaround in there too. That turned out to be exactly what I needed.

I think the problem is that the redirect means that a body is included in the GET request, which is what’s throwing the Service Worker. So I need to create a duplicate request without the body:

request = new Request(url, {
    method: 'GET',
    headers: request.headers,
    mode: request.mode == 'navigate' ? 'cors' : request.mode,
    credentials: request.credentials,
    redirect: request.redirect
});

So here’s what I had in my Service Worker before:

// For HTML requests, try the network first, fall back to the cache, finally the offline page
if (request.headers.get('Accept').indexOf('text/html') !== -1) {
    event.respondWith(
        fetch(request)
            .then( response => {
                // NETWORK
                // Stash a copy of this page in the pages cache
                let copy = response.clone();
                stashInCache(pagesCacheName, request, copy);
                return response;
            })
            .catch( () => {
                // CACHE or FALLBACK
                return caches.match(request)
                    .then( response => response || caches.match('/offline') );
                })
        );
    return;
}

And here’s what I have now:

// For HTML requests, try the network first, fall back to the cache, finally the offline page
if (request.headers.get('Accept').indexOf('text/html') !== -1) {
    request = new Request(url, {
        method: 'GET',
        headers: request.headers,
        mode: request.mode == 'navigate' ? 'cors' : request.mode,
        credentials: request.credentials,
        redirect: request.redirect
    });
    event.respondWith(
        fetch(request)
            .then( response => {
                // NETWORK
                // Stash a copy of this page in the pages cache
                let copy = response.clone();
                stashInCache(pagesCacheName, request, copy);
                return response;
            })
            .catch( () => {
                // CACHE or FALLBACK
                return caches.match(request)
                    .then( response => response || caches.match('/offline') );
                })
        );
    return;
}

Now the test case is working just fine in Chrome.

On the off-chance that someone out there is struggling with the same issue, I hope that this is useful.

Share what you learn.

Service Worker notes

Here’s a disconnected hodge-podge of things related to Service Workers I’ve noticed recently…

Service Workers have landed in Firefox. When it was first unveiled in a nightly build a few people spotted some weird character issues on sites like mine that are using Service Workers, but that should all be fixed in the next release.

A while back I voted up Service Workers on Microsoft’s roadmap for Edge. I got an email today to say that the roadmap priority is high:

We intend to begin development soon.

We’re getting there.

Here’s a little gotcha that had me tearing my hair out until I tracked down the culprit: don’t use Header append Vary User-Agent in your site’s Apache config file. I don’t know why it snuck in there in the first place, but once I removed it, it fixed a weird issue that Aaron T. Grogg pointed out to me whereby my offline page would get cached, but not my CSS.

I really like this proposal for:

<link rel="serviceworker" href="/serviceworker.js">

It makes sense to me that I should be able to point to the Service Worker of a page in the same way that I point to a style sheet. It makes sense as a rel value too: “the linked resource has the relationship of ‘serviceworker’ to the current document.”

Also, I’m just generally a fan of declarative solutions. This feels like another good example of functionality that starts life in an imperative language (JavaScript) and then becomes declarative over time (see also: :hover, the required attribute, etc.).

Lyza wrote a fantastic article on Smashing Magazine that goes into all the details of her Service Worker. I must admit to feeling extremely gratified when she wrote:

First, I’m hugely indebted to Jeremy Keith for the implementation of service workers on his own website, which served as the starting point for my own code.

Going through her code, she made this remark:

Note: I use certain ECMAScript6 (or ES2015) features in the sample code for service workers because browsers that support service workers also support these features.

That’s a really good point. I haven’t messed around much with ES6 HipsterScript stuff partly because I haven’t yet set up a transpiler, so it was nice to know that my Service Worker is a “safe space” to try some stuff out in the browser. I refactored my JavaScript to use const, let, and =>.

Jake is looking for feedback on a specific part of Service Worker functionality around URLs. If I can wrap my head around what’s being described, I’ll chime in.

Finally, I had a nice little Service Worker moment earlier today. I was doing some updates on my web server that required a reboot. When I checked in Chrome to see how long adactio.com was down, I was surprised to see that the downtime appeared to be …zero. “That’s odd” I thought, “How can my site still be reachable if the server is …oh!” That’s when I realised I was seeing a cached version of my homepage. My Service Worker was doing it’s thing.

I had been thinking of Service Workers as a tool to help in situations where the user agent goes offline. But of course it’s an equally useful tool for when the server goes offline. This was a nice reminder of that.

Cache-limiting in Service Workers …again

Okay, so remember when I was talking about cache-limiting in Service Workers?

It wasn’t quite working:

The cache-limited seems to be working for pages. But for some reason the images cache has blown past its allotted maximum of 20 (you can see the items in the caches under the “Resources” tab in Chrome under “Cache Storage”).

This is almost certainly because I’m doing something wrong or have completely misunderstood how the caching works.

Sure enough, I was doing something wrong. Thanks to Brandon Rozek and Jonathon Lopes for talking me through the problem.

In a nutshell, I’m mixing up synchronous instructions (like “delete the first item from a cache”) with asynchronous events (pretty much anything to do with fetching and caching with Service Workers).

Instead of trying to clean up a cache at the same time as I’m adding a new item to it, it’s better for me to have clean-up function to run at a different time. So I’ve written that function:

var trimCache = function (cacheName, maxItems) {
    caches.open(cacheName)
        .then(function (cache) {
            cache.keys()
                .then(function (keys) {
                    if (keys.length > maxItems) {
                        cache.delete(keys[0])
                            .then(trimCache(cacheName, maxItems));
                    }
                });
        });
};

But now the question is …when should I run this function? What’s a good event to trigger a clean-up? I don’t think the activate event is going to work. I probably want something like background sync but I don’t think that’s quite ready for primetime yet.

In the meantime, if you can think of a good way of doing a periodic clean-up like this, please let me know.

Anyone? Anyone? Bueller?

In other Service Worker news, I’ve added a basic Service Worker to The Session. It caches static caches—CSS and JavaScript—and keeps another cache of site section index pages topped up. If the network connection drops (or the server goes down), there’s an offline page that gives a few basic options. Nothing too advanced, but better than nothing.

Update: Brandon has been tackling this problem and it looks like he’s found the solution: use the page load event to fire a postMessage payload to the active Service Worker:

window.addEventListener('load', function() {
    if (navigator.serviceWorker.controller) {
        navigator.serviceWorker.controller.postMessage({'command': 'trimCaches'});
    }
});

Then inside the Service Worker, I can listen for that message event and run my cache-trimming function:

self.addEventListener('message', function(event) {
    if (event.data.command == 'trimCaches') {
        trimCache(pagesCacheName, 35);
        trimCache(imagesCacheName, 20);
    }
});

So what happens is you visit a page, and the caching happens as usual. But then, once the page and all its assets are loaded, a message is fired off and the caches get trimmed.

I’ve updated my Service Worker and it looks like it’s working a treat.

Cache-limiting in Service Workers

When I was documenting my first Service Worker I mentioned that every time a user requests a page, I store that page in a cache for later (offline) use:

Right now I’m stashing any HTML pages the user visits into the cache. I don’t think that will get out of control—I imagine most people only ever visit just a handful of pages on my site. But there’s the chance that the cache could get quite bloated. Ideally I’d have some way of keeping the cache nice and lean.

I was thinking: maybe I should have a separate cache for HTML pages, and limit the number in that cache to, say, 20 or 30 items. Every time I push something new into that cache, I could pop the oldest item out.

I could imagine doing something similar for images: keeping a cache of just the most recent 10 or 20.

Well I’ve done that now. Here’s the updated Service Worker code.

I’ve got a function in there called stashInCache that takes a few arguments: which cache to use, the maximum number of items that should be in there, the request (URL), and the response:

var stashInCache = function(cacheName, maxItems, request, response) {
    caches.open(cacheName)
        .then(function (cache) {
            cache.keys()
                .then(function (keys) {
                    if (keys.length < maxItems) {
                        cache.put(request, response);
                    } else {
                        cache.delete(keys[0])
                            .then(function() {
                                cache.put(request, response);
                            });
                    }
                })
        });
};

It looks to see if the current number of items in the cache is less than the specified maximum:

if (keys.length < maxItems)

If so, go ahead and cache the item:

cache.put(request, response);

Otherwise, delete the first item from the cache and then put the item in the cache:

cache.delete(keys[0])
  .then(function() {
    cache.put(request, response);
  });

For HTML requests, I limit the cache to 35 items:

var copy = response.clone();
var cacheName = version + pagesCacheName;
var maxItems = 35;
stashInCache(cacheName, maxItems, request, copy);
return response;

For images, I’m limiting the cache to 20 items:

var copy = response.clone();
var cacheName = version + imagesCacheName;
var maxItems = 20;
stashInCache(cacheName, maxItems, request, copy);
return response;

Here’s my updated Service Worker.

The cache-limited seems to be working for pages. But for some reason the images cache has blown past its allotted maximum of 20 (you can see the items in the caches under the “Resources” tab in Chrome under “Cache Storage”).

This is almost certainly because I’m doing something wrong or have completely misunderstood how the caching works. If you can spot what I’m doing wrong, please let me know.

Home screen

Remy posted a screenshot to Twitter last week.

A screenshot of adactio.com on an Android device showing an Add To Home Screen prompt.

That “Add To Home Screen” dialogue is not something that Remy explicitly requested (though, of course, you can—and should—choose to add adactio.com to your home screen). That prompt appears in Chrome on Android as the result of a fairly simple algorithm based on a few factors:

  1. The website is served over HTTPS. My site is.
  2. The website has a manifest file. Here’s my JSON manifest file.
  3. The website has a Service Worker. Here’s my site’s Service Worker script (although a little birdie told me that the Service Worker script can be as basic as a blank file).
  4. The user visits the website a few times over the course of a few days.

I think that’s a reasonable set of circumstances. I particularly like that there is no way of forcing the prompt to appear.

There are some carrots in there: Want to have the user prompted to add your site to their home screen? Well, then you need to be serving on a secure connection, and you’d better get on board that Service Worker train.

Speaking of which, after I published a walkthrough of my first Service Worker, I got an email bemoaning the lack of browser support:

I was very much interested myself in this topic, until I checked on the “Can I use…” site the availability of this technology. In one word “limited”. Neither Safari nor IOS Safari support it, at least now, so I cannot use it for implementing mobile applications.

I don’t think this is the right way to think about Service Workers. You don’t build your site on top of a Service Worker—you add a Service Worker on top of your existing site. It has been explicitly designed that way: you can’t make it the bedrock of your site’s functionality; you can only add it as an enhancement.

I think that’s really, really smart. It means that you can start implementing Service Workers today and as more and more browsers add support, your site will appear to get better and better. My site worked fine for fifteen years before I added a Service Worker, and on the day I added that Service Worker, it had no ill effect on non-supporting browsers.

Oh, and according to the Webkit five year plan, Service Worker support is on its way. This doesn’t surprise me. I can’t imagine that Apple would let Google upstage them for too long with that nice “add to home screen” flow.

Alas, Mobile Safari’s glacial update cycle means that the earliest we’ll see improvements like Service Workers will probably be September or October of next year. In the age of evergreen browsers, Apple’s feast-or-famine approach to releasing updates is practically indistinguishable from stagnation.

Still, slowly but surely, game-changing technologies are landing in browsers. At the same time, the long-term problems with betting on native apps are starting to become clearer. Native apps are still ahead of what can be accomplished on the web, but it was ever thus:

The web will always be lagging behind some other technology. I’m okay with that. If anything, I see these other technologies as the research and development arm of the web. CD-ROMs, Flash, and now native apps show us what authors want to be able to do on the web. Slowly but surely, those abilities start becoming available in web browsers.

The pace of this standardisation can seem infuriatingly slow. Sometimes it is too slow. But it’s important that we get it right—the web should hold itself to a higher standard. And so the web plays the tortoise while other technologies race ahead as the hare.

It’s interesting to see how the web could take the desirable features of native—offline support, smooth animations, an icon on the home screen—without sacrificing the strengths of the web—linking, responsiveness, the lack of App Store gatekeepers. That kind of future is what Alex is calling progressive apps:

Critically, these apps can deliver an even better user experience than traditional web apps. Because it’s also possible to build this performance in as progressive enhancement, the tangible improvements make it worth building this way regardless of “appy” intent.

Flipkart recently launched something along those lines, although it’s somewhat lacking in the “enhancement” department; the core content is delivered via JavaScript—a fragile approach.

What excites me is the prospect of building services that work just fine on low-powered devices with basic browsers, but that also take advantage of all the great possibilities offered by the latest browsers running on the newest devices. Backwards compatible and future friendly.

And if that sounds like a naïve hope, then I humbly suggest that Service Workers are a textbook example of exactly that approach.

My first Service Worker

I’ve made no secret of the fact that I’m really excited about Service Workers. I’m not alone. At the Coldfront conference in Copenhagen, pretty much every talk mentioned Service Workers.

Obviously I’m excited about what Service Workers enable: offline caching, background processes, push notifications, and all sorts of other goodies that allow the web to compete with native. But more than that, I’m really excited about the way that the Service Worker spec has been designed. Instead of being an all-or-nothing technology that you have to bet the farm on, it has been deliberately crafted to be used as an enhancement on top of existing sites (oh, how I wish that web components would follow a similar path).

I’ve got plenty of ideas on how Service Workers could be used to enhance a community site like The Session or the kind of events sites that we produce at Clearleft, but to begin with, I figured it would make sense to use my own personal site as a playground.

To start with, I’ve already conquered the first hurdle: serving my site over HTTPS. Service Workers require a secure connection. But you can play around with running a Service Worker locally if you run a copy of your site on localhost.

That’s how I started experimenting with Service Workers: serving on localhost, and stopping and starting my local Apache server with apachectl stop and apachectl start on the command line.

That reminds of another interesting use case for Service Workers: it’s not just about the user’s network connection failing (say, going into a train tunnel); it’s also about your web server not always being available. Both scenarios are covered equally.

I would never have even attempted to start if it weren’t for the existing examples from people who have been generous enough to share their work:

Also, I knew that Jake was coming to FF Conf so if I got stumped, I could pester him. That’s exactly what ended up happening (thanks, Jake!).

So if you decide to play around with Service Workers, please, please share your experience.

It’s entirely up to you how you use Service Workers. I figured for a personal site like this, it would be nice to:

  1. Explicitly cache resources like CSS, JavaScript, and some images.
  2. Cache the homepage so it can be displayed even when the network connection fails.
  3. For other pages, have a fallback “offline” page to display when the network connection fails.

So now I’ve got a Service Worker up and running on adactio.com. It will only work in Chrome, Android, Opera, and the forthcoming version of Firefox …and that’s just fine. It’s an enhancement. As more and more browsers start supporting it, this Service Worker will become more and more useful.

How very future friendly!

The code

If you’re interested in the nitty-gritty of what my Service Worker is doing, read on. If, on the other hand, code is not your bag, now would be a good time to bow out.

If you want to jump straight to the finished code, here’s a gist. Feel free to take it, break it, copy it, improve it, or do anything else you want with it.

To start with, let’s establish exactly what a Service Worker is. I like this definition by Matt Gaunt:

A service worker is a script that is run by your browser in the background, separate from a web page, opening the door to features which don’t need a web page or user interaction.

register

From inside my site’s global JavaScript file—or I could do this from a script element inside my pages—I’m going to do a quick bit of feature detection for Service Workers. If the browser supports it, then I’m going register my Service Worker by pointing to another JavaScript file, which sits at the root of my site:

if (navigator.serviceWorker) {
  navigator.serviceWorker.register('/serviceworker.js', {
    scope: '/'
  });
}

The serviceworker.js file sits in the root of my site so that it can act on any requests to my domain. If I put it somewhere like /js/serviceworker.js, then it would only be able to act on requests to the /js directory.

Once that file has been loaded, the installation of the Service Worker can begin. That means the script will be installed in the user’s browser …and it will live there even after the user has left my website.

install

I’m making the installation of the Service Worker dependent on a function called updateStaticCache that will populate a cache with the files I want to store:

self.addEventListener('install', function (event) {
  event.waitUntil(updateStaticCache());
});

That updateStaticCache function will be used for storing items in a cache. I’m going to make sure that the cache has a version number in its name, exactly as described in the Guardian’s use case. That way, when I want to update the cache, I only need to update the version number.

var staticCacheName = 'static';
var version = 'v1::';

Here’s the updateStaticCache function that puts the items I want into the cache. I’m storing my JavaScript, my CSS, some images referenced in the CSS, the home page of my site, and a page for displaying when offline.

function updateStaticCache() {
  return caches.open(version + staticCacheName)
    .then(function (cache) {
      return cache.addAll([
        '/path/to/javascript.js',
        '/path/to/stylesheet.css',
        '/path/to/someimage.png',
        '/path/to/someotherimage.png',
        '/',
        '/offline'
      ]);
    });
};

Because those items are part of the return statement for the Promise created by caches.open, the Service Worker won’t install until all of those items are in the cache. So you might want to keep them to a minimum.

You can still put other items in the cache, and not make them part of the return statement. That way, they’ll get added to the cache in their own good time, and the installation of the Service Worker won’t be delayed:

function updateStaticCache() {
  return caches.open(version + staticCacheName)
    .then(function (cache) {
      cache.addAll([
        '/path/to/somefile',
        '/path/to/someotherfile'
      ]);
      return cache.addAll([
        '/path/to/javascript.js',
        '/path/to/stylesheet.css',
        '/path/to/someimage.png',
        '/path/to/someotherimage.png',
        '/',
        '/offline'
      ]);
    });
}

Another option is to use completely different caches, but I’ve decided to just use one cache for now.

activate

When the activate event fires, it’s a good opportunity to clean up any caches that are out of date (by looking for anything that doesn’t match the current version number). I copied this straight from Nicolas’s code:

self.addEventListener('activate', function (event) {
  event.waitUntil(
    caches.keys()
      .then(function (keys) {
        return Promise.all(keys
          .filter(function (key) {
            return key.indexOf(version) !== 0;
          })
          .map(function (key) {
            return caches.delete(key);
          })
        );
      })
  );
});

fetch

The fetch event is fired every time the browser is going to request a file from my site. The magic of Service Worker is that I can intercept that request before it happens and decide what to do with it:

self.addEventListener('fetch', function (event) {
  var request = event.request;
  ...
});

POST requests

For a start, I’m going to just back off from any requests that aren’t GET requests:

if (request.method !== 'GET') {
  event.respondWith(
      fetch(request)
  );
  return;
}

That’s basically just replicating what the browser would do anyway. But even here I could decide to fall back to my offline page if the request doesn’t succeed. I do that using a catch clause appended to the fetch statement:

if (request.method !== 'GET') {
  event.respondWith(
      fetch(request)
          .catch(function () {
              return caches.match('/offline');
          })
  );
  return;
}

HTML requests

I’m going to treat requests for pages differently to requests for files. If the browser is requesting a page, then here’s the order I want:

  1. Try fetching the page from the network first.
  2. If that doesn’t work, try looking for the page in the cache.
  3. If all else fails, show the offline page.

First of all, I need to test to see if the request is for an HTML document. I’m doing this by sniffing the Accept headers, which probably isn’t the safest method:

if (request.headers.get('Accept').indexOf('text/html') !== -1) {

Now I try to fetch the page from the network:

event.respondWith(
  fetch(request)
);

If the network is working fine, this will return the response from the site and I’ll pass that along.

But if that doesn’t work, I’m going to look for a match in the cache. Time for a catch clause:

.catch(function () {
  return caches.match(request);
})

So now the whole event.respondWith statement looks like this:

event.respondWith(
  fetch(request)
    .catch(function () {
      return caches.match(request)
    })
);

Finally, I need to take care of the situation when the page can’t be fetched from the network and it can’t be found in the cache.

Now, I first tried to do this by adding a catch clause to the caches.match statement, like this:

return caches.match(request)
  .catch(function () {
    return caches.match('/offline');
  })

That didn’t work and for the life of me, I couldn’t figure out why. Then Jake set me straight. It turns out that caches.match will always return a response …even if that response is undefined. So a catch clause will never be triggered. Instead I need to return the offline page if the response from the cache is falsey:

return caches.match(request)
  .then(function (response) {
    return response || caches.match('/offline');
  })

With that cleared up, my code for handing HTML requests looks like this:

event.respondWith(
  fetch(request, { credentials: 'include' })
    .catch(function () {
      return caches.match(request)
        .then(function (response) {
          return response || caches.match('/offline');
        })
    })
);

Actually, there’s one more thing I’m doing with HTML requests. If the network request succeeds, I stash the response in the cache.

Well, that’s not exactly true. I stash a copy of the response in the cache. That’s because you’re only allowed to read the value of a response once. So if I want to do anything with it, I have to clone it:

var copy = response.clone();
caches.open(version + staticCacheName)
  .then(function (cache) {
    cache.put(request, copy);
  });

I do that right before returning the actual response. Here’s how it fits together:

if (request.headers.get('Accept').indexOf('text/html') !== -1) {
  event.respondWith(
    fetch(request, { credentials: 'include' })
      .then(function (response) {
        var copy = response.clone();
        caches.open(version + staticCacheName)
          .then(function (cache) {
            cache.put(request, copy);
          });
        return response;
      })
      .catch(function () {
        return caches.match(request)
          .then(function (response) {
            return response || caches.match('/offline');
          })
      })
  );
  return;
}

Okay. So that’s requests for pages taken care of.

File requests

I want to handle requests for files differently to requests for pages. Here’s my list of priorities:

  1. Look for the file in the cache first.
  2. If that doesn’t work, make a network request.
  3. If all else fails, and it’s a request for an image, show a placeholder.

Step one: try getting the file from the cache:

event.respondWith(
  caches.match(request)
);

Step two: if that didn’t work, go out to the network. Now remember, I can’t use a catch clause here, because caches.match will always return something: either a response or undefined. So here’s what I do:

event.respondWith(
  caches.match(request)
    .then(function (response) {
      return response || fetch(request);
    })
);

Now that I’m back to dealing with a fetch statement, I can use a catch clause to take care of the third and final step: if the network request doesn’t succeed, check to see if the request was for an image, and if so, display a placeholder:

.catch(function () {
  if (request.headers.get('Accept').indexOf('image') !== -1) {
    return new Response('<svg>...</svg>',  { headers: { 'Content-Type': 'image/svg+xml' }});
  }
})

I could point to a placeholder image in the cache, but I’ve decided to send an SVG on the fly using a new Response object.

Here’s how the whole thing looks:

event.respondWith(
  caches.match(request)
    .then(function (response) {
      return response || fetch(request)
        .catch(function () {
          if (request.headers.get('Accept').indexOf('image') !== -1) {
            return new Response('<svg>...</svg>', { headers: { 'Content-Type': 'image/svg+xml' }});
          }
        })
    })
);

The overall shape of my code to handle fetch events now looks like this:

self.addEventListener('fetch', function (event) {
  var request = event.request;
  // Non-GET requests
  if (request.method !== 'GET') {
    event.respondWith(
      ... 
    );
    return;
  }
  // HTML requests
  if (request.headers.get('Accept').indexOf('text/html') !== -1) {
    event.respondWith(
      ...
    );
    return;
  }
  // Non-HTML requests
  event.respondWith(
    ...
  );
});

Feel free to peruse the code.

Next steps

The code I’m running now is fine for a first stab, but there’s room for improvement.

Right now I’m stashing any HTML pages the user visits into the cache. I don’t think that will get out of control—I imagine most people only ever visit just a handful of pages on my site. But there’s the chance that the cache could get quite bloated. Ideally I’d have some way of keeping the cache nice and lean.

I was thinking: maybe I should have a separate cache for HTML pages, and limit the number in that cache to, say, 20 or 30 items. Every time I push something new into that cache, I could pop the oldest item out.

I could imagine doing something similar for images: keeping a cache of just the most recent 10 or 20.

If you fancy having a go at coding that up, let me know.

Lessons learned

There were a few gotchas along the way. I already mentioned the fact that caches.match will always return something so you can’t use catch clauses to handle situations where a file isn’t found in the cache.

Something else worth noting is that this:

fetch(request);

…is functionally equivalent to this:

fetch(request)
  .then(function (response) {
    return response;
  });

That’s probably obvious but it took me a while to realise. Likewise:

caches.match(request);

…is the same as:

caches.match(request)
  .then(function (response) {
    return response;
  });

Here’s another thing… you’ll notice that sometimes I’ve used:

fetch(request);

…but sometimes I’ve used:

fetch(request, { credentials: 'include' } );

That’s because, by default, a fetch request doesn’t include cookies. That’s fine if the request is for a static file, but if it’s for a potentially-dynamic HTML page, you probably want to make sure that the Service Worker request is no different from a regular browser request. You can do that by passing through that second (optional) argument.

But probably the trickiest thing is getting your head around the idea of Promises. Writing JavaScript is generally a fairly procedural affair, but once you start dealing with then clauses, you have to come to grips with the fact that the contents of those clauses will return asynchronously. So statements written after the then clause will probably execute before the code inside the clause. It’s kind of hard to explain, but if you find problems with your Service Worker code, check to see if that’s the cause.

And remember, please share your code and your gotchas: it’s early days for Service Workers so every implementation counts.

Updates

I got some very useful feedback from Jake after I published this…

Expires headers

By default, JavaScript files on my server are cached for a month. But a Service Worker script probably shouldn’t be cached at all (or cached for a very, very short time). I’ve updated my .htaccess rules accordingly:

<FilesMatch "serviceworker.js">
  ExpiresDefault "now"
</FilesMatch>
Credentials

If a request is initiated by the browser, I don’t need to say:

fetch(request, { credentials: 'include' } );

It’s enough to just say:

fetch(request);
Scope

I set the scope parameter of my Service Worker to be “/” …but because the Service Worker is sitting in the root directory anyway, I don’t really need to do that. I could just register it with:

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

If, on the other hand, the Service Worker file were sitting in a folder, but I wanted it to act on the whole site, then I would need to specify the scope:

if (navigator.serviceWorker) {
  navigator.serviceWorker.register('/path/to/serviceworker.js', {
    scope: '/'
  });
}

…and I’d also need to send a special header. So it’s probably easiest to just put Service Worker scripts in the root directory.

Ice cold in Copenhagen

I went to Copenhagen last week for the Coldfront conference. It was lovely to be back in Denmark’s capital. I used to go over there ever year when the Reboot conference was running, but that wrapped up a few year’s back so it’s been quite a while since I had the opportunity to savour Copenhagen’s architecture, culture, coffee, food, and beer.

Coldfront was fun. Kenneth has modelled the format of the event on Remy’s Full Frontal conference—one day of a single track of front-end dev talks in a comfy cinema.

Going to a focused conference like this is a great way of getting a short sharp shock of what’s hot—like a State of the Union address for the web. At Coldfront there were some very clear themes around building for resilience, and specifically routing around the damage of inconsistent connectivity. There was a very clear message—from Paul, Alex, and Patrick (blog imminent)—that the network is not always on our side. Making our sites work offline should be much more of a priority than it currently is.

On a related note, the technology that was mentioned the most was Service Workers …and Jake wasn’t even there! Heck, even I mentioned it in glowing terms in my own little presentation. I was admiring the way it has been designed specifically to be used in a progressive enhancement kind of way.

So if I were Mr. McGuire in The Graduate, my line to a web developer equivalent of Dustin Hoffman would be “I want to say one word to you, just one word. Are you listening? …Service Workers.”

Hackfarming Tiny Planner

Towards the end of each year, we Clearlefties head off to a remote location in the countryside for a week of hacking on non-client work. It’s all good unclean fun.

It started two years ago when we made Map Tales. Then last year we worked on the Politmus project. A few months back, it was the turn of Hackfarm 2013.

Hackfarm 2013

This time it was bigger than ever. Rather than having everyone working on one big project all week, it made more sense to split into smaller teams and work on a few different smaller projects. Ant has written a detailed description of what went down.

By the middle of the week, I found myself on a team with James, other James, Graham, and an Andy. We started working on something that Boxman has wanted for a while now: a simple little app for adding steps to a list of things to do.

Here’s what differentiates it from the many other to-do list apps out there: you start by telling it what time you want to be finished by. Then, after you’ve added all your steps, it tells you what time you need to get started. An example use case would be preparing a Sunday roast. You know all the steps involved, and you know what time you want to sit down to eat, so what time do you need start your preparation?

We call it Tiny Planner. It’s not “done” in any meaningful sense of the word, and let’s face it, it probably never will be. What happens at hackdays, stays at hackdays …unfinished. Still, the code is public if anyone fancies doing something with it.

Hackfarm 2013 Hackfarm 2013

What made this project interesting from my perspective, was that it was one of them new-fangled single-page-app thingies. You know the kind: the ones that are made without progressive enhancement, and cease to exist in the absence of JavaScript. Exactly the kind of thing I would normally never work on, in other words.

It was …interesting. I though it would be a good opportunity to evaluate all the various JS-or-it-doesn’t-happen frameworks like Angular, Ember, and Backbone. So I started reading the documentation. I guess I hadn’t realised quite how stupid I am, because I couldn’t make any headway. It was quite dispiriting. So I left Graham to do all the hard JavaScript work and concentrated on the CSS instead. So much for investigating new technologies.

Hackfarm 2013

Partly because the internet connection at Hackfarm was so bad, we decided to reduce the server dependencies as much as possible. In the end, we didn’t need any server at all. All the data is stored in the browser in local storage. A handy side-effect of that is that we could offline everything—this may one of the few legitimate uses of appcache. Mind you, I never did get ‘round to actually adding the appcache component because, well, you know what it’s like with cache-invalidation and all that. (And like I said, the code’s public now so if it ever does get put into a presentable state, someone can add the offline stuff then.)

From a development perspective, it was an interesting experiment all ‘round; dabbling in client-side routing, client-side templating, client-side storage, client-side everything really. But it did feel …weird. There’s something uncanny about building something that doesn’t have proper URLs. It uses web technologies but it doesn’t really feel like it’s part of the web.

Anyway, feel free to play around with Tiny Planner, bearing in mind that it’s not a finished thing.

I should really put together a plan for finishing it. If only there were an app for that.

Hackfarm 2013