Boris Smus

interaction engineering

The ebb of the web

Tech pundits like to lament that the web has no viable future, while web idealists hold that in fact the web is totally fine, with a "too big to fail" sort of attitude.

At the root of this disagreement are poorly defined terms. The web can mean many different things to different people. Though it started from a pretty abstract notion of a series of interlinked documents, it has now evolved to refer to a very specific technology stack of hyperlinked HTML documents styled with CSS, enhanced with JavaScript, all served on top of HTTP. In light of an increasing movement away from desktop-style computing, we've seen a big shift away from the web in mobile platforms.

Let's take apart this gob of web technology in light of the increasingly complex landscape of computing and try to make sense of what the web is and where it's going.

A framework for webiness

Webiness: how far down the rabbit hole?

I want to introduce the concept of webiness, a framework for evaluating how deeply an application embraces "the web":

Games are typical examples of apps that are pure native and not webby at all (barring high score servers, etc). Because many are playable offline, with no added benefit of having an internet connection, there is obviously no room for web. They are built entirely on native APIs, benefiting from being as close to the hardware as possible for performance reasons.

Another class of native apps are Twitter, Facebook and the like, which essentially act as specialized browsers. They largely use HTTP to access RESTful endpoints that serve up JSON, which is rendered by the native clients. Because a specialized browser is designed for a specific use case in mind (eg. interacting with Twitter), it can be streamlined for that purpose and doesn't need to deal with processing the web's presentation layer (HTML, CSS, JavaScript). Therefore it presents a more focused experience and benefits from being faster at start-up and during interaction, offline support, and a generally better experience.

Embedded browsers (also called hybrid apps) embed a webview into the user interface of the page to use the web's presentation layer for part of the user interface. How much of the user interface is created using web technologies varies widely. This approach is beneficial because it allows web developers to be featured in app stores, and also gives them access to native APIs that are not available on the open web.

Lastly, browser-based pages and apps are ones where even the code itself is fetched over HTTP. The app is written in a cross-browser way to ensure that regardless of which browser loads the page, the user is presented with a reasonable experience. The idea of this is to be able to write once, deploy everywhere. Like the other types of apps on our spectrum, these also use HTTP to interact with the server. All of the content is rendered using HTML/CSS/JavaScript. Additionally, the HTML is often hyperlinked.

The web as greatest common divisor

In mathematics, the greatest common divisor (GCD) between multiple digits is the largest positive integer that divides the numbers without a remainder. Observe that the GCD of a set of numbers (S) is by definition less than or equal to each number. Furthermore, if the numbers in S are not multiples of one another, the inequality becomes strict. So foreach n in S, gcd(S) < n.

In the webbiest case above, where we write once for all browsers, the web user interface becomes the GCD for all existing interfaces. It is guaranteed to be slower, less featureful, etc, than each individual native platform, but when the web was conceived, the benefits outweighed the costs. As the web evolved in the 90s, native platforms evolved alongside it. Computing at the time was very desktop-centric, requiring a physical keyboard, mouse, and relatively large display at, let's say 1024x768 pixels. At best, variance was between operating systems. The computer geeks were on the Linux fringe. The art and music geeks used Macs. But the hardware was pretty much the same.

Because of this uniformity, it was easy to standardize on a set of input and outputs that would work reasonably well across a bunch of existing computer configurations and operating systems. Computer hardware was all very similar, just off by some factor. And the GCD of this orderly set was pretty large: gcd(200, 300, 400) = 100.

We're not in Kansas anymore

Contrast this uniformity to today, when our mots du jour are "mobile first", or even "mobile only". But even these notions are becoming passé, as our day-to-day tech encounters start including wearable sensors for health, chips embedded in your appliances, shoes, and computers on your face. Even with the most conservative notion of what mobile means - small screens and touch input - we've really thrown a wrench into the big-screen, mouse-and-keyboard web. Scott Jenson is absolutely right in remarking in an Edge conf panel, that the web hasn't even recovered from the fact that screens have gotten smaller. With today's extended notion of computing, the GCD of all of the devices and use cases the web is trying to support becomes very small: gcd(100, 200, 300, 50, 99, 198, 33) = 1.

At the same Edge conf panel on the future of the web, somebody asked the question of how the web would work on hardware without a display. Answers from the panel were incoherent, but it's unclear how this would be built into today's frankenweb, which is already a snowball of many, often redundant technologies. And this is largely because of the notion of THE WEB as a single platform. This is both its greatest strength, and ultimately its tragic flaw, as the legacy of the early 90s causes the singular web to cave in on itself, as we are experiencing today.

We can no longer have a one-web-for-all approach. We need to focus on having many different webs, each specializing on a particular subset of our universe of devices. Taking our set above, we can split it in two subsets, S1 = {100, 200, 300, 50}, and S2 = {99, 198, 33}. Imagine S1 are the desktop-like devices, and S2 are the phone-like devices. Now, we have pretty okay GCDs: gcd(S1) = 50, and gcd(S2) = 33. Our worst case GCD is now 33, which is a lot better than 1!

The web is dead, long live the web!

I'm pretty sure that HTTP is here to stay. Our desire for content is universal, and that content needs to live somewhere. Regardless of how that content is presented to us, it is likely to be served to us through the cloud, over HTTP in the forseeable future.

What is indisputably being downplayed in the vibrant and variant near future of computing, is the web's presentation layer - HTML, JavaScript and CSS. These comprise the lingua franca of the web, and have no real competition. However, this browser-served presentation layer of the web will become less relevant as fewer things are done through the browser, especially on mobile platforms.

To me, the critical thing is that content be addressable by URL, and cross-linkable in some reasonable way. This is conventionally achieved with HTML's <a> elements, but can also be done with JavaScript (eg. a button that runs javascript:window.location.href = myUrl;, or a <form> that creates a POST request to some other resource. This can even be done without HTML at all. As long as we continue using HTTP, we are guaranteed to have content that is available at a given URL. And as long as we can guarantee that there's some handler for that content, the spirit of the web lives on.

Specialized browsers are a good solution

RSS readers are good examples of specialized browsers focused on presenting a good reading experience to the user. They are a single entry point for all of the interesting things on the internet for the user to read.

The Twitter app on your mobile device is an example of a specialized browser designed for reading and sending tweets. This specialized browser relies on the web's ability to link to various kinds of content addressable by URLs. Tweets typically include an article or an image which are served up using HTTP, and sometimes require HTML to render. Twitter is all about hyperlinked content, without ever using an <a> tag.

Apple also has several projects that are specialized browsers in spirit, though they rarely link out to the wild west of the world wide web. Generally, these specialized browsers focus on giving a great experience for the user aiming to do something specific. Similar in function to an RSS reader, Newsstand is an entry point to the magazines and newspapers you read. Passbook is a specialized browser for tracking event tickets, boarding passes and coupons. From a developer perspective, both of these browsers require you to setup a server and write some iOS code. (Unfortunately you're then stuck in Apple's ecosystem forever.)

By identifying common patterns of functionality, Apple is able to successfully introduce a specialized browser notion that spans across multiple services, unifying it with a consistent user experience. This begins to address the concern that there are too many apps for everything. Of course you don't want separate apps for NYTimes, WSJ, USA Today, LA Times, etc. And of course you don't want separate apps for each event booking service, airline and coupon company.

The problem is endemic to the web community. The standards process behemoth is slow, heavy and change-averse. Its focus is on a monolothic web, THE WEB. Imagine if we focused on use cases in the same way that Apple does, and created specialized sub-webs for various related things. The next big thing is wearable health. Imagine a sub-web for that, where we get to define the way all of these devices can talk to one another. There would be a browser for that sub-web, providing a good experience to see historical data, analyze trends, and plot data over time. This is not the stuff of a web browser, but a completely different beast.

Android's intent filters are a step in the direction of a specialized browser. Intent filters let apps register to handle specific URL patterns. If the URL pattern is opened by the user, she is presented with a dialog of all of the possible handlers for that resource, which may include Chrome, other web browsers, and specialized browsers that subscribe to that URL. Once you're in an Android app, however, you're stuck. Unless the app provides the ability to share content, there's no way to send your state to someone else. In contrast, with a web browser, you just take the URL and send it to your friend and if they have access to that content, they get to see it.

The web, the good parts

When tech pundits hail the ebb of the web, they mean that mobile native apps are eating away at the presentation layer on the mobile web and beyond. After some contemplation, I have made peace with this possible future. HTML is not the best possible way of creating content, nor is CSS a reasonable way of laying out content. JavaScript may be commonly used, but that does not make it a very good language. The presentation layer is optimized for content consumption using pointers and keyboard input, or if you're really adventurous, a smaller touch screen.

Apps are another story. Frameworks like iOS and Android provide a much more modern, cogent way of developing apps for mobile platforms. But unfortunately they aren't linkable or portable across platforms. As I mentioned earlier, there's no need for <a> elements, or anything from the presentation layer to preserve benefits of linkability. But what does the web look like once we've gutted the presentation layer? Here are some characteristics that we should preserve:

  • The ability to take any content that is currently being shown and serialize it into a URL.
  • The ability to open a URL with the right specialized browser of the user's choosing.
  • If no specialized browser is installed, some way of presenting the user with a list of good browsers.

By dropping the notion of THE WEB (singular), and ushering an era of specialized browsers, we can split our universe of devices into subsets and increase the baseline greatest common denominator. Trying to extend the web to work for every possible case will lead to even more feature creep in a web platform that is already keeling over.

The web community should take a look at verticals and spec and build specialized browsers for them. How would a web for boarding passes, concert tickets and coupons look like? How about a web for personal health tracking data? A web for content that should be consumed on an audio-only device?

This has been my slightly edited brain dump on the future of the web. Thanks for reading, I eagerly await your thoughts :)