Marcy Sutton: @marcysutton | marcysutton.com | Deque Systems

Show Notes:

• 01:07 - Deque Systems
• 01:54 - Accessibility Tool Integration and Testing
• 05:26 - Configuration and Success Criteria
• 07:04 - What is accessibility? WCAG
• 09:22 - Spurring Adoption of Accessibility
• 12:09 - The Accessibility Matrix
• 16:56 - Accessibility-First Development
• 18:12 - WCAG and ARIA Roles
• 24:57 - Test Automation vs Human Interaction
• 28:56 - Empathy Building
• 30:45 - Porting to the Web
• 35:57 - Accessibility in Single-page Apps and Focus Management

Resources:

• axe-core
• aXe
• aXe Developer Tools
• WCAG (Web Content Accessibility Guidelines)
• Web Accessibility for Designers
• WAI-ARIA Authoring Practices 1.1
• First rule of ARIA use
• Access Works: Usability and Accessibility Training
• Marcy Sutton: Notes On Client-Rendered Accessibility
• a11y on Slack

Transcript:

CHARLES: Hello everybody. Welcome to The Frontside Podcast Episode 61. My name is Charles Lowell. I'm a developer here at The Frontside. With me also is Mr Robert De Luca, a developer at The Frontside and today we have with us, Marcy Sutton who is going to be talking with us a little bit about accessibility, both in the large and the small. Welcome, Marcy.

MARCY: Good morning, everyone. Happy to be here.

CHARLES: I know, I understand you're actually calling us from the parking lot of a ski area.

MARCY: I am at the legendary Mount Baker ski area outside of Bellingham, Washington where we have the winter that is just going on and on and on and getting after it on the last few days of my birthday vacation.

ROBERT: Oh, wait. Happy birthday.

CHARLES: Yeah, happy birthday.

ROBERT: Happy belated or happy birthday.

MARCY: Yeah, it was Sunday so still on that shiny birthday week.

CHARLES: Well, thank you for getting with us on your vacation and on your birthday but doing a little bit of work, you actually work at Deque Labs. What is it that you guys do over there and what's your particular area of interest and work there?

MARCY: Deque is an accessibility company. We have people who work on products and services for accessibility. We help people avoid lawsuits and make their websites and mobile apps more accessible to people with disabilities. My slice of that work is on the product team, where I work on browser extensions, APIs for developers. Basically to make it so you don't have to write every single accessibility tool or test yourself.

You can pull in these APIs and get some of that experience that Deque has built up for years and years and years, which was part of the reason I went to work there was to learn from them. We make tools that make it easy for you to make use of that knowledge in your applications.

ROBERT: That's awesome. It’s like a base JavaScript library that can be ported anywhere, like to browser extensions. I know we use it in Ember accessibility testing. That’s really cool. That’s where I've gone for the way I write JavaScript. It's in a base library so everybody can use it and it's even more awesome that it's testing and like wrapping tooling around accessibility because I know a lot of developer-minded people want to see like a failed built.

CHARLES: Yes, what does that experience look like? I mean, coming from someone who's never even heard of these tools, how would I integrate them into my project and what would change about my workflow? What information would it surface?

MARCY: The best place and the reason I work on these products is that I saw projects go out the door broken a lot of times, when working in agencies or maybe testing isn't part of your methodology. Personally in my career, I just knew there had to be a better way. I got into software testing and the more I learned about it, the more I thought that it was sustainable, you could pull in other APIs to help you write better tests.

I went to work on axe-core, which is the JavaScript library that we've talked about a second ago. That really is bottling up all of these accessibility tests that you can automate some of the accessibility checks for things like if your HTML markup is in a good state and you're using attributes correctly. Basically, saving you from having to write all of those little microtests, some of which can be sort of complicated. It's all about getting test coverage for the automated things that we can actually test for.

CHARLES: You described a pretty wide-ranging coverage. How do you go about actually implementing that into your CI process? Do you just install the axe-core? Do you have to load up your browser and then pointed it out? What does that look like?

MARCY: Ideally, you would already have a test suite and you could just pull in the test harness. There's all different versions of aXe. There's versions in JavaScript and in Node. The core thing that you need to test is get your app running in a browser, whether it's a headless browser or could be a mounted browser but we need those actual DOM browser APIs to check things like color contrast. We need to be sort of coupled to the DOMs so that we can run our full set of tests, which is a distinction from, say some shallow rendering that you might be doing in React testing or something like that.

For accessibility tests, we need an actual DOM so you could get axe-core on npm and then pull it into your project and then you basically either require or import it, depending on what your stack looks like in JavaScript. Then you have access to all of these tests. It's pretty useful since our ecosystem has evolved to cover things like npm. I've found that it works pretty well.

ROBERT: That is pretty neat. You require it into your test and then you visit a page that's fully rendered and then you do aXe check, like you call a method that runs all these checks?

MARCY: Exactly. You would call axe.run and then you configure it to run, either specific tests or just one test. One of the tricks that has been helpful to know is that if you disable the color contrast rule, you don't need quite as many of the DOM APIs so it will run faster in things like JSDOM, which doesn't implement the entire browser APIs. But you could call axe.run, either in your unit tests or more likely it would be in your integration tests because you'd already have a browser instance, either through Selenium-WebDriver or karma-chrome-launcher or something like that. Then you basically call axe.run, passing a callback function and then it will return to you at set of JSON results and then you can do things with those.

ROBERT: When you call run, can you pass options of what you want to check? Can you filter out things that you know might -- because I imagine like if you put this into an existing app that's been going for a while, I imagine you're going to get a bunch of fails and it might be overwhelming. Is there a way to peel a back like an onion and start working at it that way?

MARCY: Yes. You can get pretty specific with our API. The GitHub for axe-core has our entire API configuration. You can get pretty specific. You can filter by tags. I imagined we're going to talk a little bit more about what WCAG is but there's a set of standards that you can break accessibility down into things that you can actually assert that they are either accessible or not. There's all different kinds of what we call success criteria.

All of our rules are mapped to these actual guidelines and standards because that means that our tests are helping you solve things that are actually helpful so you could filter by the different levels. Maybe you want to configure it with custom rules. We have some additional products for that. You can get pretty specific with what you want it to run.

ROBERT: It's extensible too so you can add your own stuff.

MARCY: You can and we do a lot of work with some of our clients to actually help them write custom roles so that's a service that we offer. But the API is pretty configurable on the JavaScript side so you can do quite a bit of configuring on your own as well, which is cool.

ROBERT: That is pretty awesome. You alluded to WCAG, I guess now we know how you can integrate a testing library into your JavaScript apps, let's take a step back a little bit and what exactly is accessibility and then you can start explaining WCAG because WCAG is a very big document that tells you how to go and be accessible.

CHARLES: I assume WCAG is some acronym?

MARCY: It is. Peeling that back a little bit to what is accessibility. I'm more on the digital side. There is physical accessibility as well for spaces. But when we're talking about digital accessibility, we're talking about making apps and websites that work for people with a broad range of abilities. Say, you had color blindness or a low vision or you're fully blind, you would need to be able to zoom in, you need high-contrast colors, you might use a screen reader if you're blind.

But then there's other categories. People might actually fall into more than one category including motor disabilities, where maybe you can't use a mouse and you have to use a keyboard only or a keyboard with one button, which is how we think about a switch control --that's another device. You might be deaf or hard of hearing and need transcripts or close captions so any audio or video content needs an alternative of some kind.

Then there's cognitive disabilities where people have learning disabilities. Maybe the language used on a website is too vague or too marketing copy speak and we need to simplify, people with traumatic brain injury like Stephen Hawking has ALS. I discovered at some point in my career that I could actually make the web a better place by supporting all different kinds of people. That's really what it's about for me is doing good craftsmanship and making sure that you're actually making things as accessible as you can.

The WCAG thing that we mentioned, it stands for Web Content Accessibility Guidelines. It's just that. It's a set of guidelines, sort of a map to help you get there. You have to actually interpret those guidelines and put in the work to do it. The guideline is just a guideline but it gives us a really good roadmap of how to implement all of these different areas of accessibility.

CHARLES: I actually had a question and this is a little bit harkening back to the discussion about the axe-core but also kind of straddling. How do you spur adoption, both the technology and the value inside of your development team? You know, we definitely make our web apps as accessible as we can because we have Rob on the team. But for teams that don't have Rob, how do you spurred option? How do you pitch it to your team and to your management structure? Like testing. Testing used to be controversial. I think in some pockets, it still is but it was something that you had to pitch or agile methodologies was something that you had to pitch. Now it's kind of accepted. It's a core-value of development, I think. I hope.

MARCY: Definitely more so. I agree.

CHARLES: Do you see a future where making applications accessible is just a tenet of development in the modern era and how do we get to that point? How do we pitch our teams to adopt that value?

MARCY: Part of what I'm trying to do is meet developers where they're at and make tools that make it really easy and free to integrate things so it doesn't cost you anything to npm install a library and pull it in your project or to use a free browser extension. What we're trying to do is really help developers get there by lowering the barriers, just kind of a funny way to put it because that's what we're doing with accessibility is removing barriers for people that get access to things.

I'm pretty optimistic about it. We talked a lot in the accessibility world about education is really needed because often, it's just that people don't know about it. I've made it my mission to spread the word as much as possible by doing talks and blog posts and just trying to get as many people on board as possible, instead of making them feel bad about it like, "Oh, you don't know about this? You’re terrible."

ROBERT: Oh man, you're speaking to me.

MARCY: "-- You can do this." I try to bring people along and make them feel welcome because it's not really a fun experience to be like, "Oh you're bad because you didn't do this. You don't think about this thing." That's what I try to do.

ROBERT: One of my first experiences in accessibility was like somebody giving me that moral argument like, "You're ruining people's lives. They can't do things on their computer." I just remember the response I had and it wasn't that, "Oh, you're right. I should go make this accessible. It was more of like I had a flight or fight response. I start to justify the reasons I didn't do and that wasn't a good experience so the way you put it, like meet the developers where they're at, I love that because that's how I've been operating too.

I think accessibility is just another engineering problem and it can be an engineering problem that would be fun to solve. The accessibility matrix gets really hard and hairy as you get into it like --

CHARLES: Oops! Jargon alert! What is the accessibility matrix? Does the accessibility matrix has Neo?

ROBERT: The different AT combos and since my experience stems from screen readers --

MARCY: Assistive technologies?

ROBERT: Yeah, assistive technologies -- I'm doing a poor job here -- Basically, you have three levels that you work with here. It's the operating system, the type of assistive technology and if we're talking about the web, it's the browser. You could have like the matrix, the beaten path is MacOS, VoiceOver and Safari. That's going to be your matrix. Then on Windows, it could be Windows JAWS and Internet Explorer or Windows NVDA, which is another screen reader on Windows. JAWS is also a screen reader. The browser for NVDA would be Firefox.

Then it can just fork in any of those different combinations that you could possibly imagine that makes it hard to debug for. But that's why I think this is a cool programming problem is because we can build awesome tools to help us do this and test for it like aXe.

MARCY: Yeah. I would also argue that it's almost even more of a design problem. It's part of the additional challenges that we have to get our design friends and colleagues on board as well because the more that they are thinking about it before they handed off to us, the less we're going to be caught in these situations where we have to make it work in one browser and assistive technology but then it's broken somewhere else because we're trying to use really experimental APIs or we're just trying to do things differently for the mouse versus the keyboard. I can tell you that could be really difficult.

The more we're thinking about making things straightforward and intuitive from the design side, not to say the easier job is going to be but the more successful, I think we can be as a team because it's more than just developments responsibility. There's good resources for designers as well, like a web accessibility for designers. If you just Google that, there's a great checklist from WebAIM. I think it's helpful to make it inclusive to people that we work with, not just in the development side because we really want them to set us up for success or else were really just fixing problems that not at their core. You know what I mean?

ROBERT: Yeah, as they come down the pipe, we're kind of dealing with them instead of getting ahead of it.

CHARLES: That reminds me actually of an experience that I had, a pair programming with Rob, probably about a year ago as we were making an interaction model for a select box. This was for a custom client. We actually stripped it away and we're like, "Let's just focus on what is the state machine behind this thing," so we drew it out on the board and it turned out that we were really just capturing the interaction apart from any rendering so we had a very strong model.

With each state's transition, we were able to basically radiate that information with a screen reader in this case. But it was actually very trivial to do because we've actually forgotten about the DOM, forgotten about the fact that we were actually chasing a visual interaction and like I said, what is the actual user interaction? What is the information coming in and coming out? It turned out once we kind of flush that out and have developed that, hanging the interface on that skeleton was very easy and we could do it in multiple media. It feels like a similar concept where if your designers are very upfront about really exploring the information architecture of an application then being able to represent that information architecture in multiple forms becomes much easier because the joints and beams are very, very clear and they aren't bound to a particular form of representation.

MARCY: Yeah, I think it a way that's definitely true. One challenge I would issue for this part of prototyping would be to consider all of the user inputs. Make sure that you're considering a keyboard user hitting an escape key to close that select or maybe they're using a screen reader on a touch device and like the single finger swipe, it's already allocated when that screen reader is running so if you have an interface that was only swipe left or right and there were no other affordances like buttons that you could actually activate, that would be an unusable interface to a mobile screen reader user.

What really helps to make that information architecture stand up or hold out when you're developing it, like stay true to your vision through the process is making sure that you're considering all of those user inputs. Sometimes, developers aren't thinking about keyboard user so they're not thinking about focus styles, really trying to activate it another way. I do think that's a helpful exercise.

ROBERT: Yeah, and to be fair at Frontend developers, we already have a lot to think about. It's just a lot to juggle so I can understand that's why we have tools like aXe. But what Charles is talking about, I think is actually kind of neat is we were experimenting with accessibility-first development so the people do TDD -- test driven development -- and I was trying to see if we could build something. I wanted to see if what we're writing would yield better software if we did it with an accessibility in mind from the outset.

I think that's true. It was a more accessible typeahead. It was better, more well-defined user experience around the typeahead and it was because we thought about accessibility and all of the different edge cases. We really boil it down to the core problem.

CHARLES: Right. We were driving it first with keys and nonstandard interaction methods. It meant that we actually got more clear interaction model lying underneath. It was decoupled from the actions that drove it completely because we had to support too from the get go.

ROBERT: I thought that was neat.

CHARLES: Yeah that was a fun exercise. You know, we should have blog about that because I think that actually results in better software.

ROBERT: Yeah, I had a conference talk brewing in there somewhere. Just never got around to it. Talking about the web accessibility guidelines. There's different levels to it. Now, you have an A, AA, and AAA. What do those mean and where does that play into ARIA roles and stuff?

MARCY: There's WCAG 2.0 and actually 2.1 is an update that they're working on right now but WCAG 2.0 is --

ROBERT: Oh, yeah. I saw that.

MARCY: Yeah, there's some new stuff coming out. It's mainly for low-vision users and mobile touch things. But the WCAG 2.0 is the blessed standard that we're working with right now and the levels are different conformance levels. There's different things that you can achieve with A, AA, or AAA. Most people go for AA. AAA is pretty restrictive in what you can do and if you make it support WCAG 2.0 AA, it doesn't necessarily mean it's going to be intuitive to use. You could make it technically conformant but it won't necessarily be that beautiful or accessible. There's a bit of a dance that we have to do around that to meet these guidelines but do them in an intentional way so that we're actually making something usable.

I think that goes back to that idea of craftsmanship and caring about your user to know if this actually going to work for them. There's a number of success criteria in WCAG that are broken up into different categories. There's perceivable, operable, understandable and robust. Within each of those, there's all kinds of different checkpoints that you can look at to inform how do I make this keyboard accessible. There's all kinds of really helpful documentation. That's the WCAG guidelines and within each of those, there are a number of different ways that you can code something.

As I'm sure you know, there are infinite ways to code the same thing, pretty much and part of what that cover is techniques for making things accessible. They'll tell you all about Native HTML and what tools you can use within that standard. Then there's this other standard called WAI-ARIA and that's the Web Accessibility Initiative – Accessible Rich Internet Applications. That was originally created back in the day when we didn't have as many browser APIs and we didn't have great ways to expose accessibility information to screen readers. They made this API in browsers that implemented that you can actually bolt on some of the same information that you get from HTML.

It’s helpful if you're writing as VG or XML, where you just don't have those built in semantics so we have things ARIA role states and properties. You may have seen things like 'role="button"' or 'role="main"' or 'role="search"'. You might see that somewhere and that is just exposing programmatically bolting on a role to any element. You could put on 'div role="button"' and there's a little more that goes into that to make it an accessible button. Anytime we mentioned --

ROBERT: The tab index.

MARCY: Yeah, the tab index. You have to make sure you have a keyboard event but that would be a programmatic way to create a button element. You should always start with the native button element because you get all that stuff for free but ARIA gives us an API to actually implement accessibility information. You'll see those techniques come up a lot in WCAG of how you can accomplish the same thing multiple ways. Those are some of the things that we test for in our animated tests in aXe. We check to make sure that you've only use roles that are actual roles because there is a set standard of them.

We check to make sure that all of the ARIA values that you might use are actually allowed for that. Sometimes, if you're using 'role="list"' for whatever reason, you can't use a real list. It is possible to create a list with ARIA but if you had the wrong child role or something, that's a pretty easy thing that we can flag with aXe so we're sort of saving you from yourself. It helps me sometimes when I get a role wrong because we're human and we do make mistakes. There's a lot of things to remember so that's pretty key technique that aXe will help you with. That's making sure that your ARIAs used correctly because it is pretty easy --

ROBERT: That's really nice.

MARCY: -- to get it wrong, to be honest.

ROBERT: Yes. I've definitely done that. Being through the spec document is not the most fun. Trying to read the standards language is a little bit complicated so having a tool like aXe is really helpful for me to pick my way through it like, "aXe will tell me that this is wrong," so it narrows the problem set down for me where I can go and look at the standard and kind of tunnel vision in on that one, rather than get overwhelmed looking at that whole standard documents like there's so much here.

MARCY: Yes, there is. One thing that might help with the is the initiative that people are working on called the ARIA Practices Guide, the ARIA Authoring Practices and it sort of breaks down these techniques into what is the keyboard navigation model for that component or it will break it into known patterns. This is really helpful also for designers to know what are some known patterns and how can I implement accessibly. They can really help you jumpstart to using those patterns with this more easily digestible information to tell you how to do it correctly. That has come up in the last few years that I found really useful.

ROBERT: That's awesome. I think I've seen this. Is it where they tell you like, "If you're going to reimplement a checkbox, here's how you would do with ARIA?"

MARCY: Exactly. I've dropped a link in the chat so we'll expose that in the show notes, I'm sure. There's more resources out there now that are really helpful. There's another one called ARIA in HTML and that one is also from the W3C and it's a note on using ARIA and HTML. That one I found to be very useful as well because they tell you this first, second, third, fourth, and fifth rules of ARIA use.

The first rule of ARIA use is if you can use a Native HTML element or attribute, you should absolutely use the built-in one first. That's a big --

ROBERT: Yeah, let's stop reinventing.

MARCY: Yeah, you know it's tempting because you can create these custom elements and try to bolt on ARIA but the reality is that if you're trying to make it really backwards compatible, it's just so much easier to support the native things. There is an assisted technology called Dragon NaturallySpeaking, that's a dictation method and they didn't support ARIA until 2014 so you can easily imagine some of your user base with an older assistive technology. That might be completely broken for them so that's why we really push using the native things first just because of the better support on every platform.

CHARLES: I have a question about the test automation. We've been talking a lot about aXe in the way that you can do this. Did I get it right? Are my roles correct? And all these things. What's an example of something that you just can't test for in an automated fashion? It just requires human interaction just to perceive it. I mean, this would be right now, kind of in the visual sphere, the state of automation for testing like did I break the layout still requires a human. What are examples of that in terms of accessible interface where you just do the things that you have to be on the lookout for that you can't cover with automation right now?

MARCY: I think context and content are some of the most difficult like writing good all text. That can be really challenging just because what makes a good alt for an image and that supposed to be a text alternative to say, "This is something useful," and Facebook has solved that by using artificial intelligence to dynamically guess what's in an image.

A blind colleague of mine that works there has written about and he said he always felt left out when he would read his news feed and someone would be talking about their first love or some kind of vague status update. With this new feature, it could say, "Oh, this image that they're talking about their love is a pepperoni pizza," or something where --

[Laughter]

MARCY: It's really missing the context so they've started to do automatic all text. For us doing accessibility checks, we try to keep our solution as light weight as possible and without false positives. We can check whether you have an all attribute missing like you don't even have the alt attribute at all which means that the file name would be read in the screen reader which is often terrible, depending on what your filenames are so we can check if that's missing but we can't really tell you what would make a better alt attribute, if you already have one. That's one is a bit difficult.

There’s another one that we're working on right now with color contrast where we can't really tell if you have a background image that's behind some text. If it has multiple pixel color values in it, even if we could read those colors, it gets really hard for us to say whether text meets color contrast when it's over an image for multiple reasons. That one's a bit tricky. I think there are some other examples throughout WCAG that we can only automate. Depending on which rule set you're using, we estimate between 30% to 40% of issues, we can actually catch with automated tests so there is quite a bit that we still need humans for.

But however, I think some of these really basic ones that we can check to help you do those easy wins so that you're not getting messed up by using the attribute Aria-role when it's just role. Those kind of things. It's like we're helping you so you can save that time for those more complex task that might require a human. There's definitely no substitute for trying to use the keyboard to make sure that your app is usable from the keyboard. Test it with a screen reader, you can find people in the web accessibility Slack that might be willing to help you test it, if you're extra nice or maybe you can give them a gift card or something.

There is an organization called Knowbility and they have this thing called Access Works where if you need to find a user with a disability to deduce a user testing for you because that's a great thing to do. It's very important. They can help you, as a business think up with someone who can test your app. I would definitely check out Access Works. That's really what's the missing piece. As a developer, I'm okay using a screen reader after doing accessibility for a few years but it's not my primary way of navigating so it's really helpful to have real users to test your app and that's a good way to find someone to actually test it. That sort of makes up the rest so you can get that really valuable feedback.

ROBERT: I'm a firm believer in testing but also, I really do think a lot of accessibility work is just kind of empathy building and the way you do that is just sit down and actually use this assistive tech that these people will be using. In that way, you can understand as you're building it, how somebody might move their screen or cursor over the top of this and you can start to think about what the screen will read off and stuff like that. I think using a screen reader as a developer is powerful. But I agree, it will never reach the level like my mom that has been using a screen reader for seven years now. I'll never be able to use it as well as she does. It actually putting in the hands of people that do this day-to-day and live this. A far better idea and that goes beyond accessibility too. You want to user test all your apps anyway.

MARCY: Yeah, exactly. I think that should be a big thing that we demand just from our organizations like how you were saying it was kind of controversial. I feel like user testing is another flavor of that where we have a bit of emotional tide of these things that we create and we want them to be perfect in the way that we have envisioned but not everyone interacts with things the same and it's really humbling to watch someone use something that you made and have it completely not get it at all. I think that's a really valuable experience. I've watched my mom or my dad or people try to use something that we assume is really intuitive and it's just not. We look at the web all day -- day-in and day-out being professionals and it's really helpful to show it to people who maybe aren't as fluent, aren't digital natives like that.

CHARLES: We talked about actual user testing. We talked about the checking where you render your application and you run a set of checks. Do you have any experience with actually -- this is kind of an idea that just occurred to me, although we did a little bit of it when we were doing native applications -- using the accessible interfaces to actually drive your acceptance tests? Is that anything that you have experience with? Because it seems like on the face of it, you've got this assistive technology that surfaces the key levers of your application so is it a good idea to grab those levers from within your test case? Within your acceptance test to manipulate your application and thereby kind of front load your accessibility because in order to verify it, you must have those levers in place.

MARCY: Yeah, from understanding your question correctly, you're wanting to just run your tests using accessibility features?

CHARLES: Yes. For example, when we write our acceptance tests in our application, what we do is as part of setting them up, say we want to click here and I want to enter this text into this text box and I want to move this over here and that implies actually dispatching mouse events, keyboard events and then also being able to find the elements in the DOM that I want to dispatch those events on so we're kind of doing it in, I think we use CSS selectors to find them and then we use the jQuery event interface to actually create the events and send them to those elements. But it seems that part of ARIA roles or something else is like identifying the role that this element has in your application and basically saying, "For my test cases, I'm going to use these roles and I'm going to use these things and I'm going to use different access methods, keyboard mouse or whatever to manipulate my interface." Does that makes sense?

MARCY: Yeah.

ROBERT: I think this makes sense in the native world where in order to get the label, I think you have to use the accessibility label.

CHARLES: They do that when you're functionally testing iOS apps so why not --

ROBERT: Does it port to the web, basically.

CHARLES: Yeah, does that port to the web?

MARCY: It does --

CHARLES: It's really long, way of saying that, I guess. Sorry you all.

[Laughter]

MARCY: No, and I wanted to clarify because I was wondering if you're talking about driving it with actual assistive technology, which we can't quite yet. We don't have any tools for that. But yes, you should --

ROBERT: We should explore that in Ember.

MARCY: Yeah, we just don't have the hooks for that. Maybe Python and NVDAs, since it's open source, maybe AppleScripts.

CHARLES: What would that look like to drive it with assistive technologies?

ROBERT: We talked to some people at Apple with Ember accessibility team and if I remember correctly, we could only drive VoiceOver on MacOS with AppleScripts but there was no way to do it in any other way so you only could do it with VoiceOver on MacOS and that was still kind of murky.

MARCY: Yeah, exactly. The idea would be, rather than just testing the browser, we would actually be able to run a simulator programmatically, to know is the screen reader actually exposing this information. Because a lot of it is there are things that get lost in translation, sometimes where we're following best practices and standards because we have this agreement that people who implement browsers and screen readers are going to follow those standards. It's definitely is not always smooth sailing with that. But there's sort of this disconnect between the browser testing and then actually firing it up in the screen reader and make sure it worked. We take that on faith a lot of time, which is getting back to your original question, why it's so valuable to have tests that use these interaction methods.

Absolutely, either in your unit tests or even in your integration test, they can live in either place to have tests that assert and closes with the escape key or it operates with the enter key or whatever the user interaction should be, that we have tests that assert that because that way, if you leave your team or heaven forbid, you got hit by a bus or something, you have a test coverage that makes a contract of how this component should work and you have accessibility support, actually built into your test infrastructure. That is super valuable.

At least we know that that part of it is there. We know we can drive it from the keyboard, which is how a lot of screen readers work. They operate on top of the keyboard so we can get really far just by having basic keyboard support. Then, if you pull in an API like axe-core, you can have it tell you if you were using ARIA wrong or something. It's sort of a combination of both where those feature tests in your actual project where you're writing something that it works with the escape key, those are custom tests for your application. I find that they're really valuable just to have in there, especially if you work on a component library or something reusable so that everybody who is contributing knows how this thing is supposed to work. I think that is really valuable.

ROBERT: Absolutely. I want to talk about accessibility in single-page apps. The problem with accessibility in single-page apps is while using a screen reader, you click a link and to the screen reader user, all it says is the link was pressed. They don't actually know that the content has changed. But in Ember, we kind of solve this by focusing the outlet that has changed but in other frameworks, in your experience everywhere else, how do you combat this? What are the best ways of attacking this?

CHARLES: Yeah, what are the problems that you encounter in single-page applications?

MARCY: I've done quite a bit of research and blogging and conference talks on this. I'm working on the Angular team for a while. The issue with the single-page app is the page isn't being refreshed when you make a raving change or something happens dynamically. The user's focus is never refresh to the top of the page so they don't hear a title change or things like that.

There’s different techniques that you can employ to make that experience more accessible. The first and foremost tool to have in your toolbox is focus management so that you're programmatically sending the user's focus to this new content. Say, I have a sidebar with links in it and I click one of them, I can send focus to content wherever it loaded on the page. That way, they are both alerted to the new content because depending on where you send it. There's different techniques for this but often, we will send focus to the wrapping element so that everything will be read aloud and you can accomplish that by using tab index of -1 in your HTML. That will make this wrapper catch the focus, essentially but it won't add it to the tab order of the entire page. That's a technique that we used to shuffle focus around.

I've also seen people use what's called an ARIA Live Region where you have this element somewhere on your page that's not visible. It has to be rendered so you can't use 'display: none' but you can basically pipe messages to these live regions to announce what's happening on the screen. I've just saw a React example where they put an ARIA Live attribute just on that wrapping element, instead of the focus management so anytime new content went into that element, it would just be announced.

The challenge with that is that you can't always control everything on the page. That works if you control everything and you know that only this one element is getting updated at the time. But often, we work in this big ecosystem where there's a bunch of things happening. Depending on how complex your app is, you might need some sort of a focus manager, some sort of a utility that will keep track of what's focused and routed around at a correct place. That's the biggest tool for creating accessible single-page apps, that's focused management.

I mean, not only for the reading content purpose but also to have their focus in the more accurate place so if they hit tab or they try to start interacting with something that they're in the right part of the page. A good example, if you think about like a modal window -- a modal window may open as a new layer over something -- that requires focus management on open so that your focus is sent into it, either to the first focusable element or to the wrapper. Then when you hit escape or close the modal, it just send your focus back.

ROBERT: To the previously focused element, right?

MARCY: Exactly, so that if you are using a keyboard and you can't actually use a trackpad or a mouse to get back then you're in the right place or if you're screen reader user and you can't even see the screen, then you're always in the right spot. That's actually, I think really cool. Something that's become more common place with dynamic JavaScript apps is that we can do these really cool focus management techniques. I think they're really cool, they can be challenging but that is something that we definitely need to think about as developers of single-page apps.

ROBERT: Absolutely, especially since none of the single-page app frameworks out there were libraries. Actually maybe with the exception of your work on Angular, they don't come with a router focused-library built in so this is something that you have to actually think about and then pull in and do yourself. Does Angular have it, by default?

MARCY: No, we never added a focus manager utility. There were some things to try and clean up that HTML, which ended up being, honestly worse than the original problem. But I've written a blog post about focus management techniques. I just dropped that in the chat. There's a smashing magazine article I wrote and it really is framework-agnostic so it sort of covers all of the things that you need to think about if you're writing a client-rendered application using Ember, React or Angular.

It is something that we have to think about as developers because from the framework level, it's impossible to know what the right situation would be in your app in a given moment so we can only get so far with magic at the framework level. It's something I would like to see more of. Maybe if there is some sort of a layer manager, I think that is a tool that someone could write that would be super useful -- to make sort of an intelligent layer managing system for focus management.

I've heard the Facebook team talked about how they do it internally but it's not open source so I have yet to see an open source solution for this. We have to tackle it in our own apps but once you know that that's the thing, you can really make sure that you're covering it. If you have someone with a visual disability or impairment that try and use your app, they'll probably uncover that problem pretty quickly. That's the value of user testing in case you forget. Maybe there's a few views --

ROBERT: Need to sell it.

MARCY: Yeah, or maybe with your application, if you don't have visible focus styles turned on, you might not see that the focus isn't being sent. That is one trick, I will tell you in development. If you're working with focus management, turn the focus outlines on and then if you were trying to send focus before it got fully rendered or something because it has to actually be rendered to catch the focus. That is good debug flag, if you can all agree on the focus styles, for all users. I found that to be really useful in our app. You just to have those turned on so you can debug it.

ROBERT: And make it really loud like this is a giant red outline.

MARCY: Yeah, then you'll know, if you forgot to add tab index of -1, to make it catch the focus or like I said, maybe there's a rendering thing where you need to wait a tick by using a set time out or something. That is a good technique that I've used recently.

ROBERT: Awesome. Basically, what it boils down to in single-page apps is manage your focus and enhance your focus, some might say.

MARCY: Yeah, let's think about keyboard ergonomics, like if you are doing things dynamically on the screen and then you want to start typing, I think the most common example I see is autofocus. The developers, even if they aren't thinking about accessibility, they'll ask for autofocus. That in a way is focus management. The difference with autofocus is that you can only use it once and it will send your focus there automatically. But in a similar way, that's the idea of what we want is to get the user's focus point into the right spot so that they can do the right activity on the screen and they know what content they're looking at.

ROBERT: Right. Sometimes, it's like navigating around a website with your keyboard, that's like power users who have Vim or Emacs or anybody that's a power user of computer that doesn't like to leave the home row, you can make your application awesome for you to use and also lay the groundwork for accessibility, if you can navigate your website with just a keyboard.

MARCY: Exactly.

ROBERT: Let's try to pitch it to people in that way. It's still a developer problem.

CHARLES: I like that because it really highlights the fact that there is this kind of deep interaction model. The user actually is focused on one thing at a time in the application and if you track that, then it's going to be a benefit for all of your users. If you are deliberate about thinking like this is the subject of interest at this moment. You're just going to reap a lot of benefit for everybody.

ROBERT: Keep coming back to it, building accessible applications yields a better application for everybody.

MARCY: Absolutely. It might enable you to support some futuristic device that you haven't even thought of yet. If you have your actions decoupled from the actual input and you can do everything declaratively, that really makes it easier to try and support of use cases you haven't thought of like we need to borrow up that other keyboard combination or some touch device. It just really helps to not have everything buried in a jQuery event.

ROBERT: Yes.

[Laughter]

MARCY: Like, "Oh, man I need to call that same functionality for multiple events. Crap." You need to decouple that real quick.

ROBERT: "Let's obstruct this."

CHARLES: Right. I think we're about the time. I know you've got a hard stop. You got some skiing to do.

MARCY: I do.

CHARLES: So we will let you get up on the mountain but thank you so much for coming by. This is been a great conversation.

ROBERT: Yes, thank you for dropping all the knowledge.

CHARLES: Yeah, I'm feeling lots of knowledge right on top of my head --

MARCY: Awesome.

CHARLES: -- That I got to go and process. But for everybody else out there, I would say go experiment with aXe. The idea is going to be easy for developers. I know I'm going to experiment with it and then you said, there was a browser extension as well to help you out and probably call out every website that you ever use, right?

MARCY: I'm dropping some links for you, just now.

CHARLES: There's some links to go along with the knowledge so go check them out and you are @MarcySutton on Twitter?

MARCY: That is correct.

CHARLES: All right. Fantastic. Thank you so much for coming by.

MARCY: Yeah, no problem. Thanks so much for having me.