Today we saw a new, interesting service called Does it work on Edge? which allows you to enter a URL, and get that URL rendered in Microsoft Edge. It also gives you a report in case there are issues with your HTML or CSS that are troublesome for Edge (much like Microsoft’s own service does). In most cases, this will be browser-specific code like prefixed CSS. All in all this is a great service, one of many that make our lives as developers very easy.

If you release something on the web, you get feedback. When I tweeted enthusiastically about the service, one of the answers was by @jlbruno, who was concerned about the form not being keyboard accessible.

The reason for this is simple: the form on the site itself is none insofar there is no submit button of any kind. The button in the page is a anchor pointing nowhere and the input element itself has a keypress event attached to it (even inline):

click for bigger

There’s also another anchor that points nowhere that is a loading message with a display of none. Once you click the first one, this one gets a display of block and replaces the original link visually. This is great UX - telling you something is going on – but it only really works when I can see it. It also gives me a link element that does nothing.

Once the complaint got heard, the developers of the site took action and added an autofocus attribute to the input field, and proudly announcing that now the form is keyboard accessible.

Now, I am not having a go here at the developers of the site. I am more concerned that this is pretty much the state of web development we have right now:

• The visual outcome of our tools is the most important aspect – make it look good across all platforms, no matter how.
• As developers, we most likely are abled individuals with great computers and fast connections. Our machines execute JavaScript reliably and we use a trackpad or mouse.
• When something goes wrong, we don’t analyse what the issue is, but instead we look for a tool that solves the issue for us – the fancier that tool is, the better

How can this be keyboard accessible?

In this case, the whole construct is far too complex for the job at hand. If you want to create something like this and make it accessible to keyboard and mouse users alike, the course of action is simple:

• Use a form elment with an input element and a submit button

Use the REST URL of your service (which I very much assume this product has) as the action and re-render the page when it is done.

If you want to get fancy and not reload the page, but keep all in place assign a submit handler to the form element, call preventDefault() and do all the JS magic you want to do:

• You can still have a keypress handler on the input element if you want to interact with the entries while they happen. If you look at the code on the page now, all it does is check for the enter button. Hitting the enter button in a form with a submit button or a button element submits the form – this whole code never has to be written, simply by understanding how forms work.
• You can change the value of a submit button when the submit handler kicks in (or the innerHTML of the button) and make it inactive. This way you can show a loading message and you prevent duplicate form submissions

What’s wrong with autofocus?

Visually and functionally on a browser that was lucky enough to not encounter a JavaScript error until now, the autofocus solution does look like it does the job. However, what it does is shift the focus of the document to the input field once the page has loaded. A screenreader user thusly would never ever learn what the site is about as you skip the header and all the information. As the input element also lacks a label, there isn’t even any information as to what the user is supposed to enter here. You sent that user into a corner without any means of knowing what’s going on. Furthermore, keyboard users are primed and ready to start navigating around the page as soon as it loads. By hijacking the keyboard navigation and automatically sending it to your field you confuse people. Imagine pressing the TV listings button on a TV and instead it just sends you to the poodle grooming channel every time you do it.

The web is obese enough!

So here’s my plea in this: let’s break that pattern of working on the web. Our products don’t get better when we use fancier code. They get better when they are easier to use for everybody. The fascinating bit here is that by understanding how HTML works and what it does in browsers, we can avoid writing a lot of code that looks great but breaks very easily.

There is no shortage of articles lamenting how the web is too slow, too complex and too big on the wire compared to native apps. We can blame tools for that or we could do something about it. And maybe not looking for a readymade solution or the first result of Stackoverflow is the right way to do that.

Trust me, writing code for the web is much more rewarding when it is your code and you learned something while you implemented it.

Let’s stop adding more when doing the right thing is enough.

Last week I ran the panel and the web components/modules breakout session of the excellent Edge Conference in London, England and I think I did quite a terrible job. The reason was that the topic is too large and too fragmented and broken to be taken on as a bundle.

If you want to see the mess that is the standardisation effort around web components right now in all its ugliness, Wilson Page wrote a great post on that on Mozilla Hacks. Make sure to also read the comments – lots of good stuff there.

Web Components are a great idea. Modules are a great idea. Together, they bring us hours and hours of fun debating where what should be done to create a well-performing, easy to maintain and all around extensible complex app for the web. Along the way we can throw around lots of tools and ideas like NPM and ES6 imports or – as Alex Russell said it on the panel: “tooling will save you”.

It does. But that was always the case. When browsers didn’t support CSS, we had Dreamweaver to create horribly nested tables that achieved the same effect. There is always a way to make browsers do what we want them to do. In the past, we did a lot of convoluted things client-side with libraries. With the advent of node and others we now have even more environments to innovate and release “not for production ready” impressive and clever solutions.

When it comes to componentising the web, the rabbit hole is deep and also a maze. Many developers don’t have time to even start digging and use libraries like Polymer or React instead and call it a day and that the “de facto standard” (a term that makes my toenails crawl up – layout tables were a “de facto standard”, so was Flash video).

React did a genius thing: by virtualising the DOM, it avoided a lot of the problems with browsers. But it also means that you forfeit all the good things the DOM gives you in terms of accessibility and semantics/declarative code. It simply is easier to write a than to create a fragment for it or write it in JavaScript.

Of course, either are easy for us clever and amazing developers, but the fact is that the web is not for developers. It is a publishing platform, and we are moving away from that concept at a ridiculous pace.

And whilst React gives us all the goodness of Web Components now, it is also a library by a commercial company. That it is open source, doesn’t make much of a difference. YUI showed that a truckload of innovation can go into “maintenance mode” very quickly when a company’s direction changes. I have high hopes for React, but I am also worried about dependencies on a single company.

Let’s rewind and talk about Web Components

Let’s do away with modules and imports for now, as I think this is a totally different discussion.

I always loved the idea of Web Components – allowing me to write widgets in the browser that work with it rather than against it is an incredible idea. Years of widget frameworks trying to get the correct performance out of a browser whilst empowering maintainers would come to a fruitful climax. Yes, please, give me a way to write my own controls, inherit from existing ones and share my independent components with other developers.

However, in four years, we haven’t got much to show.. When we asked the very captive and elite audience of EdgeConf about Web Components, nobody raised their hand that they are using them in real products. People either used React or Polymer as there is still no way to use Web Components in production otherwise. When we tried to find examples in the wild, the meager harvest was GitHub’s time element. I do hope that this was not all we wrote and many a company is ready to go with Web Components. But most discussions I had ended up the same way: people are interested, tried them out once and had to bail out because of lack of browser support.

Web Components are a chicken and egg problem where we are currently trying to define the chicken and have many a different idea what an egg could be. Meanwhile, people go to chicken-meat based fast food places to get quick results. And others increasingly mention that we should hide the chicken and just give people the eggs leaving the chicken farming to those who also know how to build a hen-house. OK, I might have taken that metaphor a bit far.

We all agreed that XHTML2 sucked, was overly complicated, and defined without the input of web developers. I get the weird feeling that Web Components and modules are going in the same direction.

In 2012 I wrote a longer post as an immediate response to Google’s big announcement of the foundation of the web platform following Alex Russell’s presentation at Fronteers 11 showing off what Web Components could do. In it I kind of lamented the lack of clean web code and the focus on developer convenience over clarity. Last year, I listed a few dangers of web components. Today, I am not too proud to admit that I lost sight of what is going on. And I am not alone. As Wilson’s post on Mozilla Hacks shows, the current state is messy to say the least.

We need to enable web developers to use “vanilla” web components

What we need is a base to start from. In the browser and in a browser that users have and doesn’t ask them to turn on a flag. Without that, Web Components are doomed to become a “too complex” standard that nobody implements but instead relies on libraries.

During the breakout session, one of the interesting proposals was to turn Bootstrap components into web components and start with that. Tread the cowpath of what people use and make it available to see how it performs.

Of course, this is a big gamble and it means consensus across browser makers. But we had that with HTML5. Maybe there is a chance for harmony amongst competitors for the sake of an extensible and modularised web that is not dependent on ES6 availability across browsers. We’re probably better off with implementing one sci-fi idea at a time.

I wished I could be more excited or positive about this. But it left me with a sour taste in my mouth to see that EdgeConf, that hot-house of web innovation and think-tank of many very intelligent people were as confused as I was.

I’d love to see a “let’s turn it on and see what happens” instead of “but, wait, this could happen”. Of course, it isn’t that simple – and the Mozilla Hacks post explains this well – but a boy can dream, right? Remember when using HTML5 video was just a dream?