I was lucky enough to be invited to attend and speak at Edge Conf London 2014, an assembly of web development superheroes charged with discussing the future of web technology in front of a live audience. I've written up my main take-aways from the event.
Web Components Panel at EdgeConf 3
- Web Components are the custom elements that you've always wanted. If you're after a
tag, web components can give it to you.
- The basics are registering a new element with the DOM, then you can do anything
- Web components are imported with a HTML link element:
- To make these components useful, you need to use the Shadow DOM - this is the DOM inside an element, which is already being used on the web: take a look inside Chrome's dev tools at an element - tickers are
- There are no browsers that support this out of the box yet, so there are two polyfills that you can use: Polymer (Google) and X-Tags (Mozilla)
- The Server/Client rendering trade-off is the concern at the moment. Any JS downloading in a web component will block rendering unless specified as async. You can also compress and minify web components and their resources, which is a necessary step to get anywhere near good performance. We're adding new tools, but the old techniques still apply.
- Responsive components, that have media queries related to their own size, aren't possible because they could be influenced by the parent too, which will get the rendering engine into infinite loops.
- On semantics and accessibility: they are still important, but with ARIA roles not caring what element they are on, you can make anything appear like anything, so the argument that web components are bad for semantics is kinda moot.
- On SEO, the usual rules still apply, you've got to make your content accessible and not hide it, but the search bots will read web components
- On styling, using scoped styles (a level 4 spec) works very well, as these will override at scope. However, using an object-oriented CSS approach makes this easier. It is, however, generally harder to make all of your CSS into OOCSS, which is more of a team/rigour problem.
- In the end, you're responsible for packaging and de-duplicating your resources. Web components will remove any duplicate files from the same origin, but it's still very easy to import two versions of jQuery. You are responsible for that.
Bruce Lawson has another very good write-up on this session, and web components in general on his blog.
Developer Tooling Panel
- Firefox's in-built developer tooling has come a long way from Firebug, with new features like deep scoping of function variables, great memory profiling and visual descriptions of CSS transforms
- Chrome's dev tools will have a better call stack flame graph
- Brackets does great in-line CSS editing and rendering
- Remote Debug aims to solve the complex workflow issue, where a developer knows how to use Chrome's dev tools, but not Firefox's, and it will let you use Chrome's tools with Firefox.
- JS Promises aren't in the dev tools sets yet. We need to experiment with the technology and then make tools. We can't make tools for something that doesn't exist yet
- Use debugger; rather than console.log - you may as well use alert()
- It would be great if we could inspect headless browsers with dev tools so that we can see what went on with a test
- It was also noted that contributing to dev tools is harder than it could be. Remy Sharp suggested creating some kind of Greasemonkey scripting capability for dev tools
Build Process panel
- Building software is not new. The Unix make command is 37 years old.
- We are trying to avoid a plugin apocalypse with grunt, gulp, bruch etc re-inventing the wheel
- A big mention in this session of the Extensible Web Manifesto as a basis for how we should be developing these tools
- There are things that belong in our build process, and things that belong in our browser. With Sass, variables should be in the browser, but minification and compilation should not.
- As a community, we need to be responsible with what we put into the standards
- Use git, not github. Use the npm protocol, not npm. The basics are great, but we can't rely on services like this
- More tools fuel innovation. A single task spec as a way to describe tasks between the task runners would be great, but this is on hold as they currently can't agree.
- On Grunt/Gulp - use the tool you're comfortable with, and make the most of it.
- We are still in the early days of build tools. Being locked in to a certain tools for 5 years probably won't hurt, because at least you're using a build tool!
Page Load Performance
- "Put it in the first 15KB for ultimate performance". Load content first, then enhancements, then the leftovers, and you'll get great performance
- Use sitespeed.io and webpagetest for your synthetic testing
- Looking at The Guardian - their branched loading model saves them 42% with their responsive site
- There will still be times when an adaptive site, with proper redirects, will give you better performance. I can vouch for this: Yell.com on the desktop is around 700KB, the homepage on mobile is around 60KB.
- HTTP2 will make spriting an anti-pattern, as it makes it easier to only download what you need. Remember, it's only the network that needs the data in one file
- If you own your site, instrument it. Target StartRenderTime, and use Window.performance for better timing. Look to improve the Page OnLoad Event.
- Resource Priorities and timing APIs will arrive soon. You're encouraged to use these in your Real User Monitoring (RUM) stats. Not many companies do this at the moment.
- Finding out is a user is getting page jank is a hard problem as you'd have to hook into RequestAnimationFrame
Pointers and Interactions
I was on this panel, so I've not got any notes! Thankfully, Google were there to video it for everyone
- For accessibility, complying to geo-specific regulations is important, but, complying with the law doesn't make your website accessible
- Are WCAG guidelines outdated? No, their values are still good, but there are more complex use cases since it was written. For example, gaming accessibility is about making visual cues auditory
- Mechanical audits of a website don't give you the full accessibility brief. It can give you ARIA roles, colour contrast, click regions and alt-text etc
- Try the Chrome Accessibility Tools extension, and the SEE extension, to see your page though different eyes
- If you want to know what it feels like to have a muscular problem, try using your mouse with you non-writing hand
- Accessibility needs to be considered at the design stage - if done at the QA stage, you've missed the point
Sadly, I wasn't at the event for this, but here's the video for you all to enjoy
I loved the conference, I've not learnt so much in a day in years! Here are some other write-ups from around the web: