This hilarious animated gif crafted by Tiago Franco (@Alarka) has probably been viewed by the entire web development community by now. It shows the responsive web designer’s instinctive grabbing of the browser handles whenever they open a new website. We all do it and there’s nothing wrong with that.
It illustrates a real problem in situations where web designers and developers try to design for this eventuality - to dazzle and amaze their industry colleagues while losing sight of the project’s audience. If you are building a website (responsive or otherwise) and your project personas become industry heroes rather than those you painstakingly identified at the beginning of a project then it’s time to worry. The project is in danger of losing sight of it’s original goals, the purpose it set out to serve and the value it was supposed to deliver. I would wager that this mostly happens around the concious incompetence and concious competence stages of the responsive dip.
Don’t get me wrong - I don’t have an issue with testing layouts in this manner for the right reasons, in fact I would strongly recommend stress testing your layout at random viewport widths. Try closing your eyes whilst dragging the browser handles to alter the window width, stop dragging at a random point and then open your eyes again. This removes any transitional1 hocus-pocus from the changing layout and gives you an idea of the state in which real users will load your product. Real users aren’t going to see changing layouts, the chances are they will see only one layout state2.
Don’t lose sight of who you are really building this for.
Per browsing session, it’s perfectly reasonable to assume a user may load your page on different devices at different times, in fact - almost a cert. Will they see the layout dynamically change between these sessions? No. Design for one state per session. ↩
I had an epiphany recently regarding one of the biggest concerns around adopting a responsive methodology for web projects — workflow.
People have been stating that responsive design projects require a fundamental shift in the web design workflow we have become accustomed to from years of building fixed layouts based on flat images that try to represent a website. I believe the shift is actually relatively small, in fact you might be working this way already.
After a critical evaluation of my previously established workflow, I came to the same conclusion a lot of pundits have already arrived at: flat compositions just don’t work. They are as useful as a photograph of a web page, they don’t represent interactions, the infinite canvas, how typography behaves, and so on. Compositions set an unrealistic goal, a rainbow for web developers to chase to the end of a project with scattered showers of compromise and disappointment along the way.
I realised an additional enlightening thought: the design phase is ethereal throughout the project, an entity that permeates through the different stages of a project lifecycle. It moves, it’s supple, it adapts. Sound familiar? Allow me to walk you through my refined process for responsive projects1.
Wireframes
When I wireframe I usually iterate from very low fidelity sketches with initial ideas up to a medium fidelity sketch with general layout decisions, content considerations and components to document in a functional specification2.
This is where the small changes to my process begin. After annotating my wireframes to correspond with the functional specification I use an additional copy of the wireframes to label typography and modular components with reference numbers. Similar pieces can be labeled with the same colour or other categorisation techniques to help with the next step of the process.
The wireframe begins to resemble a set of instructions for an Airfix kit. Identical pieces of the wireframe that are intended to be reused are given the same reference number. The reference numbers correspond to the next part of the process which may be the biggest change depending on how you work.
Style Guide
The wireframe numbers correspond to components in the live HTML style guide. This is the heart of the design phase. After accounting for all pieces in the wireframe, you will be in a position to construct a complete style guide covering typography and modular components ready for use throughout the website. It serves a similar purpose to the pieces of an Airfix kit.
We are now in the ideal position to design in the browser — there will almost certainly be assets you still need a graphics package to create, I’m not dismissing their effectiveness for quickly testing colour and type combinations altogether either. Photoshop/Fireworks still serves a definitive purpose in my workflow albeit a much less significant role than before — mainly asset generation.
Functional Prototype
This step of the process remains unchanged. Here I build the functionality (not design, not layout) documented in the functional specification. This stage is only complete when all functionality is tested and complete.
Customised Prototype
The wireframes and style guide play a pivotal role in this methodology. They are schematics — instructions for assembling model using pieces from the style guide kit. The typographical components and modules in the HTML style guide are the pieces that you assemble according to the wireframes to build the model. This is where the design in the browser happens.
Using the wireframes as a guide for shaping the layouts in terms of proportions we can then quickly piece together detailed design layouts on top of a fully functional prototype, we can make adjustments to things that don’t work so well and quickly iterate. It’s almost like painting by numbers — it’s that easy.
I wasn’t fully sold on designing in the browser until I tried this approach. By stepping into the browser sooner, you can make more informed design decisions without shooting yourself in the foot down the line. The style guide is key to the success of this approach, the more flexible the pieces are, the better they will fit into the overall model.
In summary, the biggest change is dropping flat compositions as a deliverable — they are a heavy piece of the process that set false expectations. Perhaps they may still be of use internally for quick visualisation purposes, but they are damaging to responsive projects if treated as a deliverable. Wireframes should be given more prominence as a set of instructions that inform both the creation of the style guide as well as informing the shape of the layout when it comes to customising a functional prototype using the style guide. The wireframes should be a conceptual model of the final product, not the definitive model. This process involves greater communication with your clients which is a good thing. The style guide is presented much like Samantha Warren’sStyle Tiles with the message that it contains the building blocks of the complete model. Like any process, there is always room for improvement — I would love to hear your thoughts if you have any further ideas or insights from your own process. This works for me, I hope it helps you too.
This isn’t the complete process, I have left out the discovery and planning phase as it will be unaffected by the changes above. ↩
A functional specification is a document detailing how components work along with corresponding references to the wireframes. It’s a useful deliverable that doubles up as a test plan. ↩
A stub1 is the part of a document torn off and kept as a record, it usually resides at the end of documents like cheques, receipts and tickets.
I can’t think of a more apt description of this handy little technique I am about to share with you to show you how I would approach print styles in responsive projects.
Print layouts are limiting. There isn’t much you can do in the sense of advanced layouts, even floats can cause problems. For print it’s a matter of reducing the visual noise, extracting the content and determining useful supplementary information.
Lucky for us if we are developing from a mobile first standpoint we can involve print styles at this level too so that most of our layout resets for the print media type are already taken care for us before horizontal layout occurs.
<link rel="stylesheet" href="css/global.css" media="all">
<link rel="stylesheet" href="css/layout.css" media="all and (min-width: 33.236em)">
<!-- Floats and horizontal layout techniques live within layout.css where there is room, global.css contains all styles apart from horizontal layout and any adjustments for large contexts -->
I’m going to generalize for a moment: mobile first layouts are mostly full-width blocks that contain the full set of styles (minus horizontal layout specific styles like floats and widths) with a harmonious vertical rhythm. When making a print stylesheet, we usually cancel floats, reset widths and hide the unnecessary interactive elements like navigation and video. The floats and widths (generally) aren’t set in our global stylesheet so it makes perfect sense for a print media query to live at this level, so at the bottom of your single column stylesheet, you could have:
Then you would change how you reference your stylesheets in HTML by changing the media type referencing your CSS file for layouts greater than a single column to this:
<link rel="stylesheet" href="css/global.css" media="all">
<link rel="stylesheet" href="css/layout.css" media="screen and (min-width: 33.236em)">
<!-- changed media="all and.." to media="screen and..." so when printing, the device won't look in layout.css for print styles -->
For more fine grain control over the quality over printed material from your project, you can make use of print media query values to map styles to printers with certain levels of detail:
@media print and (min-resolution: 300dpi) {
/* Styles for printers with a print resolution greater than 300 dots per inch */
}
Putting your print styles at the bottom of your single column stylesheet means you are only pruning the styles set at a this level before horizontal layouts occur meaning they are will be quicker to produce, easier to maintain and lightweight. Responsive web design has a knack for traversing mediums, lets take advantage of that.
Ironically if this short article resided on Wikipedia, it would be classed as a stub ↩
We are accustomed to the phrase “design patterns” which encompasses an established solution to a design problem consisting of purpose, function and appearance. There are lots of libraries cataloging these solutions to help designers and developers make sensible decisions when designing and building a product.
Two years ago Ethan Marcotte introduced Responsive Web Design, which has become the biggest movement on the web since The Web Standards movement. With Responsive Web Design we changed the way we make websites, our tools remain largely the same but our approach is different.
At a component level I believe we are still learning the ropes. A design pattern established for one device may not translate so well for another. It’s easy when designing and building responsively to ignore design pattern functionality problems lost in translation and concentrate solely on layout issues. For example a large multicolumn dropdown built for navigating multiple sections within a site may provide a good user experience on a desktop machine - it may translate poorly on smaller screens and harm the user experience.
Another example would be the lightbox design pattern. The purpose of a lightbox is to display a larger image corresponding to the selected thumbnail version while keeping the user on the same page instead of linking directly to a page showing the full image. Again, there are an abundance of solutions for solving this problem but the vast majority of existing patterns translate poorly to smaller displays. In fact you may argue that a lightbox shouldn’t even exist on small displays, the image may already be displayed at full width, and where do you go from there on a narrow viewport? These are the questions we need to ask ourselves when translating design patterns across devices and viewports.
We should challenge our current arsenal of patterns, ask difficult questions of them - how will this work on a narrow screen? On a huge screen? A TV? A touch screen? For a user with low bandwidth? I try to ask these questions at the wireframing stage so I can quickly test what works and what doesn’t, then refine and iterate. Even with a Mobile First mentality I found that you still need foresight to cover all eventualities - planning, documenting1 and getting sign-off on your pattern behaviours pays off so you have a reference for testing and controlling project scope.
We are dealing with an entirely different beast in that we are building experiences that users will be touching with their bare hands - there is nothing between them and the interface. We need to ensure our tried and tested design patterns translate appropriately for all scenarios so that we can build a new library of responsive patterns that will assist us in making great things for the web. Don’t be afraid to hack or invent, re-invent and re-invent again.
I find it helps to allow this documentation to evolve with the project. Locking it down at a wireframing stage can lock you into building functionality you later realise doesn’t work during prototyping or with the content. ↩
There are lots of solutions to address the problem of responsive images. The solutions share a similar set of requirements, some meet the majority of the requirements, some don’t and specialise in doing one thing well.
I recently read a framework for discussing responsive images solutions by Jason Grigsby and his first point really hit home with me. Scaling a large image for mobile devices may not produce the desired results every time, content authors should be allowed to specify different images for different contexts.
Most solutions offer this approach including the picture tag proposal. Where this falls short, in my opinion, is that there is still a disconnect between the picture tag breakpoint values and the media query values specified in your CSS. Each time a picture tag is written you need to specify your breakpoints again. In a templating environment you would use variables to take care of this but you will still be setting your breakpoints in different places.
Then I recently read Jeremy Keith’s fantastic Conditional CSS article which helps connect your CSS media queries to JavaScript to perform actions based on breakpoints. The possibilities are endless with this little gem. One of those possibilities is using it for specifying breakpoints for responsive images that are set in our HTML.
We can store image URLs in data attributes to ensure they aren’t loaded initially, only the original src is loaded. This approach borrows from Josh Emerson’s Responsive-Enhance technique where you load the fastest image you can for mobile and replace on the fly where it’s suitable to do so for the end device. By combining this philosophy with Jeremy Keith’s Conditional CSS, then we have responsive images based on our media queries.
CSS:
@media all and (min-width: 45em) {
body:after {
content: 'desktop';
display: none;
}
}
As outlined by Jeremy’s Conditional CSS technique, this sets a string of text we can detect using JavaScript to determine when the media query is active. You would then repeat the above code for your other contexts eg “tablet”, you could extend it for a 2x retina display with the -webkit-device-pixel-ratio media query.
JavaScript:
var size = window.getComputedStyle(document.body,':after').getPropertyValue('content');
if (size == 'desktop') {
$('img').each(function(index) {
var large = $(this).data('large');
$(this).attr('src',large);
}
}
if (size == tablet') {...
So to recap on the code above, it’s checking that when the desktop media query is active by checking the existence of the content in the “body:after” element, then it swaps the src with the value of the data attribute “data-large”. And of course you would repeat this for the other media queries too. Note I’m using jQuery above, you (the smarter one in this dialogue) could do this without a framework, I’m just proposing an approach. Users without JavaScript enabled get the low resolution, compressed mobile image specified in the img src, I’m fine with this as users without JS are getting a lo-fi version of the web anyway.
So now we have an approach that ticks a few important boxes:
Breakpoints defined in CSS and only in the CSS
Images can be whatever the content author specifies (no automatic resizing)
No bandwidth hit for mobile devices
No code repetition
Fallback for users without JS
I would love to hear your thoughts on the approach above and any improvements you can think of.
Responsive Web Design allows us to change layouts and element appearance based upon the width and/or height of the destination device. It’s the ideal scenario for users, coders and content coordinators alike - same site, same code, same content, just optimised for your experience depending on the display properties of the device you are using.
There can be stumbling blocks along the way to a beautiful unified site that runs perfectly on all platforms, which I’ll not go into as they are well documented (think, responsive images, page weight and so-on). The one that has been teasing me for a while is Content Choreography. Trent Walton wrote a great article on the matter. My key takeaway from his article was that when it comes to the narrow width single column view, content stacking is unavoidable and the approach the majority of us take is to simply stack the page elements in the order they appear in our document order.
What happens when this order isn’t ideal for a narrow single column viewport but it works as we want it to for others? We want to make the most of the confined space and let key content and components flourish, but sometimes our hierarchy of elements can prevent that from happening. Say for example we want to present an article in the narrow single column view, but before the article appears in the stacking order we have: a header (containing site name etc), navigation, maybe even a banner advertisement, then the article. The heart of the page is buried beneath items that may be less important in this context. Rather than brutally hiding these items with a display:none property we can reorder them using another CSS property - flexbox.
The flexible box layout module allows us to do more interesting things than our current box model allows for producing layouts, including the ability to reorder elements. Sounds exactly like what we wanted to do with the article example mentioned earlier, right? The flexbox property box-ordinal-group let’s us reorder elements based on a group number assigned to elements in our CSS. They appear in the order you specify starting from 1. So in our example we would priortise the elements on the page:
Header (1) - Contains the site name, most important
Navigation (4) - While a key element, it’s not crucial in our example - we just want to present an article
Banner advertisement (3) - Important for our advertisers, we are compromising for our example and dropping them below the article
Article (2) - The second most important thing on the page apart from the site’s name.
When we apply those values to box-ordinal-group for each of the elements, the browser would render them in this order:
Header (1)
Article (2)
Banner advertisement (3)
Navigation (4)
All this happens with minimal fuss - in your single column media query you apply display:box to the parent container which houses the elements you wish to re-order, assign the box-ordinal-group values according to the order in which you want them to appear, save, refresh and - ah-ha! They appear as if they have been floated horizontally, what happened? Flexbox arranges elements horizontally by default, because we only want to apply this at the single column view (I’ll explain why later), we need to add one more property to the container alongside display:box, and that property is - box-orient:vertical. Now we are working with our familiar stacked block-level elements only this time we have the delicious power to re-order them as we please.
I mentioned briefly that we would only want to do this when our layout is viewed in a single column, allow me to explore this a little further - there are advantages to applying flexbox at a single column level. One is that it is easier to get your head around moving objects up and down rather than horizontally moving and stretching and filling and stacking. Another is that by aiming to do this at a single column media query (I hate associating devices with viewport widths but for the sake of illustrating this point this would be the mobile view) we can enjoy the incredible support for flexbox on mobile devices. In addition it is (arguably) mobile devices that feel the pain of the need to re-arrange elements for optimal presentation in that context. Also if a mobile browser can’t recognise flexbox (IE9 Mobile is the only major one missing from the list so I’m assuming it can’t), then it will simply fallback to the default document order which isn’t the worst outcome.
So unless anybody points out some fundamental flaws in this technique, and please feel free to do so if there are any, then I plan to use this a lot when building responsive sites. I can see it causing some extra foresight when building for mobile first although I don’t think that’s an entirely bad thing, you would essentially build for mobile fallbacks and the larger context’s document order first, then re-order to the ideal choreographed scenario when finished. To be honest, I’m still figuring that part out, the best way to approach this, as with everything else in responsive web design is to build and refine and repeat.
