This is an unpublished draft preview that might include content that is not yet approved. The published website is at w3.org/WAI/.

Captions from ACT Rules for Manual Web Accessibility Evaluation Methodologies
Online Symposium, 14 March 2018

An EU Project

Meeting Record

This is the meeting record from the online symposium on ACT Rules for Manual Web Accessibility Evaluation Methodologies, held on 14 March 2018.

Introduction to ACT

Shadi Abou-Zahra: Welcome, everyone, to the online symposium on ACT rules for manual web accessibility evaluation methodologies. As a reminder, first of all, you should have a link to the slides in the latest e-mail so you can follow these on your computer. We are also showing them, sharing the screen. I hope this is working for most of you who are joining using the online web conferencing system. If you are dialling in, please have a look at the slides separately.

Before we dig in to the details just to put in context the long road until where we are today, and that’s important just to put some things in context, some of these are relevant today. So some of this work on conformance testing started really long ago. It is as old as WCAG 1, maybe even before that when the Trace guidelines emerged. Some people were thinking about tooling and evaluation schemes and ways to test for these guidelines, for these requirements. So there was a lot of work early on by Len Kadsay who unfortunately passed away, Wendy Chishom and many others who wrote the techniques for accessibility evaluation and repair tools. Flowed in to development of WCAG 2.0 to make it testable. And if people remember the techniques themselves and the failures, they have some small procedures in there. So at least intent and idea was there.

For quite awhile there was evaluation and repair tools working at W3C and that was focusing specifically on working with evaluation tool developers and looking at ways to improve automation and tool support, but also the consistency of evaluation and repair. And it was a lot of work there. It was also a lot of generations of different kinds of accessibility test tools and manual methodologies that emerged over the years and how the entire industry here of testing basically matured over the years as we went along.

An important point I think is the so called WAB Cluster. It stands for Web Accessibility Benchmarking. And this was actually three projects funded by the European Commission that were asked to put together the so called Unified Web Evaluation Methodology, UWEM. We are going to look at that in the next slide in a bit more detail, but it is actually one of the important attempts to try to have a coordinated and aligned methodology for testing web content. And this was actually within these three projects together, there were maybe 24 European organizations in there and there was a lot of work put in to try to make that happen.

And also one of the important things we are going to look at later is test samples for WCAG 2.0. One of the projects in this cluster, BenToWeb, and we have people on this project on the call today. So they may be clarifying some things. But those test samples are examples of good and bad content which were to test evaluation deals. So there was a lot of work put in to that and this is actually coming back in the ACT work through the idea of test cases. And we are going to look at that a little bit later on. A lot of the work that has happened in the past has helped inform our latest approach, our latest attempt and also some of the work invested in the past is stuff that we can take up on and build on for the new work that we are talking about here.

Let me talk just a little bit about the Unified Web Evaluation Methodology. And this was an important step to try to align different practices that were going on in different European countries and to try to harmonize the approaches between them. In my view one of the challenges with that UWEM was it was trying to do many things at the same time. It was trying to be large scale monitoring. So having here an approach for large scale monitoring and the issue with large scale monitoring you are looking more for indicators rather than quality. You are not really looking at the specific state of a website, but you are trying to kind of have an idea how accessible is it and trying to relate the degrees of accessibility among different websites or different regions and so on. So this you kind of sacrifice a lot of the quality for scalability when you are doing such large scale.

At the same time the UWEM was also trying to be a process for evaluating websites, qualitative to the evaluation for improvement. Somebody who wants to check their own website and see where they perform good, where they perform bad and what they need to fix. So the classical evaluation where people look at the site in detail to try to improve it.

And last but not least there was also another attempt here on certification. And in certification again you have a slightly different view on evaluation where you are primarily looking for problems, just looking for issues, just trying to confirm whether no issues are found, you are not really trying to exhaustively find all types of issues or all issues because, you know, typically in certification you don’t even suggest the changes or the improvements. You just look for failures. And if there are none, then you issue the certificate. And if there are issues then you ask people to revisit their website.

So and again all these three approaches are important for accessibility. But trying to combine them in one is really a challenge. I think each of these needs its own methodology and its own approach to accomplish, even if they are based on the same tests. And this is where this work comes in to try to have a basis tests that are common that can be used in different kinds of methodologies and scenarios.

Another issue with UWEM is that it was based on WCAG 1 and we all know that WCAG 1 had a lot of difficulties with testability. Things like saying provide sufficient contrast was a big issue because it is not a testable statement. You don’t know how sufficient it is. And so that created a problem. So these were some of the issues of why UWEM had difficulty, but at the same time I think also during that stage there was not a lot of interest from I think particularly some of the industry organizations but many others as well to basically share, have a common understanding of testing. Many people felt that how they test a website and these checks are their assets, their know how and that they need to keep this to themselves because it is really how they differentiate themselves from their competitors or from the others or they invested so much in work that they don’t want to give it up. So these were some of the issues at the time that made it very difficult to really get people together and agree on one common approach.

Moving on to slide No. 6 is really looking at the benefits of harmonization or the impact of fragmentation. What happens when people continue that whole process and have different evaluation results essentially, different understanding of accessibility, how they check something. And that really causes a lot of confusion among everyone involved in the development life cycle. And by developers here I mean the designers, front end, back end, the coders, the testers and so on. Really all the developers here but also suppliers to organizations. This happens very often. You are trying to plug in maybe something, some template or some tool, a player or something like that and that doesn’t meet your understanding of the requirements because the suppliers maybe have a different understanding of the accessibility requirements. And also third party evaluations when you are doing third party evaluations and there are differences. And that causes a lot of wasted effort. And we don’t have that many resources to spend in accessibility really to have duplicate efforts and these kinds of differences amongst us and in some cases actually it even causes loss of contracts.

So when you are purchasing a website and then you run a tool over it or you do your testing, and you say hmmm, that’s not accessible and the other person says no, it is, according to there. And so, you know, maybe even lose contracts, but in any case a lot of wasted efforts and resources that are very rare and that we need to be careful of. But more importantly it reduces the credibility of accessibility for the entire field. It hurts us to have these differences in interpretation. And this is in the last years one of the big changes that is happening which is many more people are coming to the understanding including tool developers who for years and years were trying to distinguish themselves based on their testing interpretations are coming around to the conclusion that by having a more common understanding and reliability of how tools and methodologies perform the more basically the stronger the market is for everyone and the more benefits we all have from that.

And there is still a lot of ways to differentiate also between methodologies. I’ll get in to that a little bit later. But the idea here is there should be different types of methodologies. Maybe they are going to be more educational. This is very often in say the educational sector or in the public sector where you are trying to get people to do evaluations who are not regularly accessibility experts but need to be able to check some of the content.

And so these people need maybe very educational or basic methodologies that have a lot of explanation and hand holding and guidance. Now in other areas, let’s say in methodology for a certification authority where the primary users are accessibility experts who need to know what they need to check. And for them maybe such an educational methodology would be too verbose and would be too distracting and they want to have more, you know, exact information about what they need to test and that’s it. And they will just go through that. So there are different ways or different kinds of methodology that each have their place and the distinguishing factors between methodologies and tools should not be based on the interpretation of the results but on other factors. For tools there are things like the performance of tools and how well they fit in to your development process and many other things where tool makers can actually differentiate themselves. Also the coverage, how many tests they cover and so on. But it shouldn’t really be on interpretation. And so I think this is a major point here that as I said a lot of in the community is actually coming to that conclusion and so this has been a major step that led to this latest work on ACT, Accessibility Conformance Testing.

Now meanwhile, on slide 7, while this was going on so to say in the evaluation area we also had the release of WCAG 2, the Web Content Accessibility Guidelines, 2.0 in 2008 which as most of you will know has more testable requirements than WCAG 1 had. Also this notion of techniques and failures I think is very important. The idea is that you can extend, provide more detail as you go along as you find out things, as new technologies emerge. We saw the changes of HTML4 to HTML5, introduction of WAI ARIA and some of this WAI ARIA being integrated in to HTML5. Technology changes very quickly. And by having this idea that you can have techniques and failures that adapt to the current situation, the current technological situation but that the requirements themselves are more stable. This has proven to be really useful and something also where we as testers can come in and adhere to the testing aspects.

So most of the time here we are looking at failures. A lot of the checks that we are talking about actually map to failures that the content does not meet. Also one of the important things about WCAG 2 is that since its release there has been much more international harmonization. Back then with WCAG 1 or several years ago there were many more derivatives of WCAG 1.0. The U.S. was different than Europe or the European Commission was saying. The individual European Member States have had different standards. Also around the world there were yet more different interpretations of WCAG 1 mixtures between WCAG 1 and WCAG 2. And so the situation was much more dire. It is not perfect yet. There are still issues and differences between countries. But there is certainly much more harmonization meanwhile which I think is a good situation to be in to know that these testing work we are doing would cover a much broader area, much broader market. So this strengthens our case and helps the work that we are doing.

Also meanwhile there is Website Accessibility Conformance Evaluation Methodology, WCAG EM. It is actually yet another different type of methodology. It doesn’t go in to the details of how to check individual requirements, but through the process of how to evaluate a website because WCAG focuses really on pages, on individual pages more. And this is a non-normative specification. So this is not a W3C standard, it is a so called Working Group Note that explains how to check a website, how to select individual pages from a website. So they are representative and so on. So that you can get to a useful answer at the end when you have done that of how accessible a website is.

And meanwhile also there is a so called Auto-WCAG Community Group and I need to explain that a little bit. At W3C we have Working Groups which are established by the W3C membership to create a standard or a specification. Whereas a Community Group anybody can create, and anybody can get work started there. So this is kind of pre-standardization, what we call incubation. And when enough community interest in there or something really proves to be coming along then the membership may decide to make that actually in to a Working Group.

This is actually how a lot of this latest work on Accessibility Conformance Testing started up. There was yet again another European funded project called EIII project. This project is one of the descendants of the WAB Cluster. Their focus here was on more the large scale monitoring of websites in regions. So what they were trying to do is basically create yet again a set of tests which they can run across sites, automated tests so they can run across sites to get an idea of how accessible they are. Instead of just doing that work within the project itself, they started the community group and actually also on the call today is Wilco Fiers who is the co-facilitator of that work today at W3C. They basically started to work in the open in this community group. And it started to gather momentum. People started to join that group. Started to discuss, started to exchange and again it was focused really on automated testing. Maybe in hindsight that name is a bit misleading because it has the automation kind of connotation whereas meanwhile we broaden the work much more. But anyway, that was really a kickstart in terms of having here something from the community where we could see that there is much more interest around this, around harmonizing than we had in the past, in the years before where people were really not as keen on having a common interpretation.

So that takes us to slide 8, the creation of the Accessibility Conformance Testing work within W3C. As I said, this Auto-WCAG group and incubation work started to gather sufficient critical mass and there was enough membership within W3C to get that work started, a task force was created. This is W3C terminology, but basically a task force is a subgroup of a working group. So directly within the Accessibility Guidelines Working Group which is developing the accessibility guidelines. So developing WCAG. Directly embedded in there is this task force that’s tasked with Accessibility Conformance Testing. As I mentioned earlier, the co-facilitator here is not only Wilco Fiers but also Mary Jo Mueller from IBM who is also on the call today.

These are the facilitators of the group but we have had a lot of involvement from other organizations, from Siteimprove, from the Daisy Consortium where they are looking at digital publishing, from Level Access. University of Illinois was participating and several others. I’m sorry that I’m forgetting some. And the mission of this task force is to develop three distinct things.

First is the so called ACT Rules Format. The idea being here that if we want people to share their test rules and contribute them we need a common way of writing down these rules so that we can actually share them. So we can actually have people contributing these test rules. Another piece here to develop by the group is the ACT Review Process. So once somebody has contributed a rule, what should be the process from that contribution to make it in to a formal, a vetted and approved rule? And last but not least is actually those rules that get vetted and reviewed and approved to collect them and provide an open repository of such rules that people can take and use in their methodologies or in their test tools.

All these pieces are in development. All these things are still in draft. So there is a lot of opportunity here to comment and provide input. A lot of moving pieces but it is firming up. And next on slide 9 maybe that gives hopefully a bit of an overview of how these things, how these pieces fit together.

In the very bottom of this diagram is the ACT Rules Format. Again this is the common format of how you write down such rules and so that you can share them in to the repository and there can be different sources here. We have commitments from several different organizations. The commitment that they will be sharing rules. We also, since November 2017, have an EC funded project called the WAI-Tools Project supporting this work. Specifically this presentation today, this symposium is organized by this WAI-Tools Project. And I’ll mention more at the end about that.

So we have several contributors here in order to accelerate the work. These different sources will be contributing ACT rules. And then on the left hand side is this vertical bar and that represents the ACT Review Process. That really takes it from the source to the review and vetting process in to the rules repository which already has implementations. So part of this review process is that you need to demonstrate implementation of these rules in order for them to be adopted in to the rules repository. And then we can have further implementations.

The word implementation here applies to not only test tools but also to manual methodologies. That’s an important aspect here. That ACT Rules themselves are not intended to be used directly by evaluators. These are not what you actually go and test a website with, but these get interpreted in to a test tool or in to a methodology. Because you have the different flavors of methodologies and different flavors of tools. So the ACT Rules are really addressed more to people developing tools and developing methodologies rather than directly at the people carrying out evaluations. These implementations, these tools and methodologies are used by the evaluators. The hope here is that we have more consistent results between these different evaluations. Regardless which methodology you use or which tool you use you have more consistent results at the end. That’s the overall goal.

Moving on to slide No. 10, I should mention, going in to the individual pieces here of this diagram is the ACT Rules Format, 1.0. This is a working draft which we have just published another version of on the 8th of March. Just a few days ago. The review period goes until 5th of April. So this is your opportunity to comment. I will remind you of this later on in the discussion. Really looking for feedback here, your input. It is starting to firm up. We hope that our next draft will be the last working draft and then we will be going in to the next level of maturity at W3C. And so we are making progress here. But we do need your input and actually specifically the purpose of this symposium today is to introduce most of you who, from seeing your responses, are involved in manual evaluation, manual testing. So we would like to know from people developing manual methodologies how does this really fit with, what you are trying to do with your development of methodologies and your development of test rules.

So let’s have a look at the outline. These are the pieces on slide No. 11 that this rules format specifies, needs to be in an ACT Rule. First of all, a unique identifier. Every rule needs to have a unique identifier, descriptive title, and a description. The accessibility requirements that it maps to, things like which success criteria of WCAG does this address. Some limitations, assumptions or exceptions, accessibility support. Aspects under test. This is relevant for tools as well. So you are looking at the DOM tree. Are you looking at the direct HTTP response that you get from the server? Are you looking at the code or the rendered code? What is it that you are actually testing on the content and providing that information is really important.

Also a very important concept and we are going to be talking about that in a little bit more detail in just a bit is the applicability and expectations. When we first started this work a lot of this was based on so called steps. Testing steps. Do this first; do this second; do this third. The issue with that is that it doesn’t map well to things like let’s say tools or approaches that are doing this, implementing this differently. Let’s say Artificial Intelligence or other types of testing that do not really have this procedural approach but looking more at what is it that I need to test.

The concept of a rule group, of combining different rules together to get to a common test, to test particular requirements and finally test cases. And test cases that I have said much earlier those test samples. Test cases are examples of content that meet this rule and content that doesn’t. And that’s really, really important. This is how you check your implementation. Whether your implementation is a test tool or a methodology. This is how you check if you have the correct interpretation of the rule or not.

Let’s look at a few examples just to make this a little bit more concrete. On slide 12. I’ll switch over to a browser. The first one we will be looking at is the ACT R2, which is a rule that we have published along with the rules format just to demonstrate how such rules would look like. This rule is called proper use of ARIA described by. What you are trying to check here is whether ARIA described by has been used properly or not. And what you have is the accessibility requirement, is which success criteria this really maps to and in this cases is success criteria 1.1.1 which is non-text content. You then have a description and some background, and some of the WCAG techniques and failures that it maps to. Some of the assumptions that you are making here in order to test. There are certain assumptions. What I talked about the aspects of the test, so in this case you are testing the DOM tree and the CSS styling. These things are defined in the rules format. Then the really important piece that I wanted to show is the applicability and the expectation.

The applicability part you describe as precisely as possible the specific aspects that this rule applies to. So in this case it is an element with the semantic role of image that has a non-empty ARIA described by attribute. And it goes on. This is really what is applicable to. It makes it very specific what it is applicable to. And we moved away from what we previously had as things like CSS selectors or things to try to codify this. Because it has too many limitations and things that restrict the implementations, especially when we get in to things that are more manual or not fully automatible.

Then you have the bunch of expectations which all need to be applied to this applicability in order for this test procedure to be checked. So, for instance, the first one is that each target element of this ARIA described by has a direct identifier. This is the first expectation. The second expectation, first of all, you have to fulfill the first expectation but also this described by ARIA attribute. So you write down the expectations of what you need to achieve. Now what you don’t do is you don’t say which tool you should use to do this or how to do it in a browser or so again this is not directly geared to the evaluators themselves but people who are going to take that and implement it in either a testing methodology. One could be more verbose or the other more educational, the other more technical. There could be different flavors of these methodologies, but also for tools. Also by writing it down like this you could have more tool supports. Even if this rule may not be fully automatible maybe tools can do different parts of these checks. For instance, to check that certain attributes exist, a tool can do that. But you may need to point out to the tool or support the tool to understand what the references are.

By writing down the exact testing, first of all, we achieve a more common understanding and more consistency in the results but also allow for more tool support to take place, to have more scalability basically.

Moving down in the rule, again these validation tests, these are the test cases that’s really important. Here you have some test cases and some examples of code that should pass. Some that are expected to fail. And some that are expected to be not applicable. That’s what’s your result is. When you run your methodology or you run your tool on these test cases, that’s the results that you should be expecting.

I want to also show just two other rules that are in development because they are more related to manual testing. This is just an early draft. You can see from GitHub it is still only an issue. It is not even a so called pull request yet. It is the earlier stage of development. I am looking now at the keyboard trap R1 which is in the slides, the URL with issue 67 at the end. This one actually we have been working on that through the WAI-Tools Project. This is actually derived from the methodology that is used by the Norwegian Government agency called Difi who are also on the call today. They have a check in checking the keyboard trap. In their methodology they lead the evaluator from one step to the other what they need to do. And we extracted from that the actual essence of the check. The applicability and expectation. While doing that we saw that we need to actually separate it in to two subrules, R1 and R2. This R1 only checks that you can actually move away from the element. Many people don’t remember the success criterion of that part. The success criterion for keyboard trap, something along the lines that you can move away using standard keyboard controls. That you can move away from an element that you have tabbed in to or moved in to using the standard keys. By the way, standard keys is not defined in the guidelines. Or the user is informed about how to move away. So this or there, this needed a second rule to map that.

This first rule R1 checks the first part of the requirement, which is if you can move in to an element you can actually move out of the element again. We have put an initial definition of what these standard keys are and these are the escape key, tab key, shift tab and enter key. This will be for discussion but the idea here we started to define this more specifically. What does it mean to move out using standard keys which is not really specified in WCAG at this moment.

And then R2, as I said earlier R2 covers that or part of that success criterion of that requirement which is that help information is available. There are certain expectations again mapped in to how help information can be made available if you cannot use these standard keys. This is what we call a rule group. You have to combine in order to check the success criteria. There are two different kinds of subchecks in there or two atomic pieces.

These were some examples of these ACT rules to give you an idea of how these ACT Rules look like. The important part here is that this is really geared towards developers of methodologies, not directly at evaluators. Then, what I have mentioned earlier, is the ACT Review Process. We saw those two rules, actually three rules but, you know, two sets and one looked finished. We just wanted to show the first one with the ACT R2, we wanted to show how it would look like when it is done, when it has completed this entire review process. The keyboard ones we have looked at most recently which is step 1 which is the creation stage. Step 2 is the validation stage which you make sure you have the test cases and in the right format. So, for instance, these keyboard tests right now they don’t have all the test cases that they should have. That’s why they are still considered in step 1. When they have all test cases and all pieces that they need, then they would then pass the validation stage. Then comes the verification stage where people debate and discuss. For instance, this definition that we have of what standard keys are. Somebody may challenge that or maybe somebody will request that another key combination get added. These are different kinds of things that need to get agreed on. Step 4 is the implementation of these rules. Every rule needs to have at least two independent implementations before it can even be considered by the Working Group for acceptance. This is how we try to do this work in the community process, outside the Working Group, so that we are more scalable. We do that work through community development and demonstrate that these rules are broadly accepted and agreed on, and there are implementations for them.

Step 5 is put in brackets because it is not really a step. This is actually an ongoing thing, the maintenance. At any time there could be new technologies, any reason why you would need to change a rule or maybe deprecate it or maybe there was a bug in it or things like this. There is this maintenance aspect in this review process.

Last but not least, when there is acceptance and sufficient consensus, it may get accepted by the Working Group in to the ACT Rules repository as a completed ACT Rule that is ready for broader use.

Moving on to slide No. 14, we have the same diagram again that we had earlier. Hopefully this will actually make a little bit more sense now, this diagram. Again, at the very bottom is ACT Rules Format which is the basis, defines what belongs in rules and how these rules should look like and this is out for draft. We are asking for your input on this and also review process, also this is for discussion. This also something we would like to have input on. On the side it is written ACT Task Force. This ACT Rules Format is what is standardized and developed by the W3C Working Group. The rules contribution and the rules review and processing really happens in a community group outside the W3C. Right now it is the Auto-WCAG community group. That’s the community group that we have. But there is no reason not to have others. For instance, there could be a community group focusing only on rules for digital publishing let’s say or we were for a while debating whether we needed to have a manual-WCAG group or have both in the same group side by side which is kind of what we are leaning in to.

The layer above that is implementations. We start to have implementation already during the review process. We hope when you release rules in to repository that will get more implementations of this. Last but not least at the very top are the evaluators who use the implementations to come to a conclusion.

Let’s just continue to slide No. 15. I want to emphasize the auto-WCAG group, since it is a community group, is open to anyone who wants to participate. We do ask you do get involved if you want to create rules, if you want to contribute. For instance, if you have ideas on the keyboard rule that’s being developed, well, chime in. Put in your comments. Put in your issues. Let us know what your thoughts are, or if you want to implement these rules and we will have a look at these repositories in a bit. The URL is there but I’ll also show you later on.

Questions from the Survey

Shadi Abou-Zahra: I’ll start by looking at some questions that we got from the survey. Unfortunately there were not as many as we were hoping. Many of you actually it seemed like wanted to know more about the ACT work and what it is which is why we have kind of switched more to introductory mode and to have more kind of background explanation. There were some interesting questions and I wanted to cover some of these.

So some of the questions revolved around the idea of, you know, accessibility evaluation cannot be completely formalized or you cannot have, you know, to what degree can you make it so mechanical. One comment was even saying accessibility will get boring if everything was completely formalized and mechanical. Absolutely. There are key limitations. There are many things that cannot be tested automatically or even formalized, but what we are trying to do here is to document what we already know. To get agreement on some of the things like keyboard trap or things like headings hierarchy. There are several parts I think in WCAG where we can just document having a common understanding of and move on. There will be actually quite a couple of things still open.

The next question actually also relates to that very much is kind of, you know, we are looking here at ACT Rules on an element level or very atomic level of issues. But very often issues are kind of in combination or in context. There is a big picture, what if you are dealing with an entire component rather than bits and pieces. That’s another limitation of this ACT work. This is another thing where evaluators will probably need to bring things together. Aggregate some thoughts like, you know, this keyboard thing is happening on navigation or versus it is happening somewhere else. What is the failure here. That’s the issue. There aren’t many things here that ACT does not fully address, but at least again we would have common understanding for some of these atomic level things.

Some other questions where well, there is limited set here in repository. Not a lot to see. That is yes, just, you know, chicken or egg. We are just starting to develop some of these things. We do hope that we will see the commitments from some organizations. We have this European funded project that will support the development of these rules. We hope that in the next while once these pieces start firming up, that we will see the development of more rules here and this repository will get fuller as we go along.

Some of the other questions that I liked about the usability of these ACT Rules. I would like to filter them. I would like to have them integrated with other resources. I would like to be able to use them more directly as a developer. This is also something that we are looking in to. I mentioned earlier we have the WCAG EM, evaluation methodology. We have the How To Meet WCAG 2 Quick Reference Guide. So there are several bits and pieces where we can tie together these things and have these rules appear in these tools and resources that the evaluators use more directly. Have them appear there, but first we need to have this repository kind of more rules in order to start connecting some of these thoughts.

Let’s see, what else? What is out of scope of ACT? That’s a really great question. There are a lot of things that are out of scope. We talked about the limitations, just a little bit earlier. ACT really tests on an atomic level, on a very detailed level. It misses out a lot on the kind of grand scheme of things. More importantly ACT, one of the things that we really tried to or one of our important limitations is that we will not try in any way to redefine the requirements of. So that’s really important. We are not trying to extend or change or modify the requirements of a Working Group. That’s completely out of scope to try to define new requirements here.

There were a few others but I think I will go to the queue and I will actually start opening up and have some discussion. If people don’t get on the speaker queue, then I will actually maybe call on some people. Wilco and Mary Jo, can I ask you both to unmute and if you wanted to add anything or to respond to some of the questions.

Questions from the Floor

Wilco Fiers: So to anybody who is not familiar, I’m Wilco. I work for Deque Systems. Thank you, Shadi, for putting this all together and for the presentation. It has been very helpful. The thing I want to emphasize most again is what I’m hoping to accomplish with this work is that we can start to harmonize our efforts. There is a lot of work going on in a lot of different organizations. And we are duplicate on a lot of stuff. So and all of that just means not just extra work. It means that differences are going to arise. People are going to see these differences. People are going to be confused about why does organization A say one thing and organization B say a different thing. And that costs us things. We lose people if they start getting confused about accessibility and really getting a clear message helps all of us. So that I think that’s really the essence of what I’m trying to get at, what I am hoping to get at with this kind of work.

Mary Jo Mueller: I’m Mary Jo Mueller and I’m with IBM. I am the accessibility programs manager. I think that pretty much everything has been said about what our real ultimate goals are. It is important that there is a very consistent resulting message with tools doing testing. That they kind of agree upon their results. And then there is no questions that arise as far as who is right or what is exactly right with the test for accessibility barriers, what are the real accessibility barriers. And we get a nice consistent way to test for conformance which we need to report for our products.

Shadi Abou-Zahra: Okay. Thank you both. So let’s see, some questions from the floor. So, first of all, Mark asks what does failure of a rule mean. Does it mean that a certain particular technique does not meet the success criteria or that the success criteria is violated. So the design of ACT Rules goes directly to success criteria. We don’t test techniques. But we actually test directly for the success criteria. So a rule maps to a failure. So basically you could think of it as a more elaborate failure. There are existing WCAG failures but we need more. Failures are sometimes also more general, like failure to provide text alternative or something general. We are breaking down these failure conditions in to more atomic and more specific bits and pieces, but if there is a failure of a rule, then actually that would map directly to a failure of at least a success criteria. Sometimes it may map to more than one success criteria.

Wendy asks have we looked at the results format at the Sirus oasis. Yes. We are planning to continue using EARL which Wendy, you have been involved in creating. I would be interested to hear more of your thoughts, whether you think that there is a particular benefit of Sirus. We start to use JSON realization of EARL.

Wendy Chisholm: My name is Wendy Chishom. Nice to see the reference there. Thank you. Some of our tools are not accessibility tools. And so we are looking for some formats so we can use between policy areas and looking for something that is more generic and so no decisions have been made. We are looking at some formats right now, Sirus, that is something that has been talked about recently. I am curious if others feel that also or if EARL is broad enough. One of the issues that we are having and it is funny, I haven’t looked at EARL in a while and I need to go back and look at it again, Sirus doesn’t have the breadth of fields that we are capturing, screen shots. Something we are investigating now.

Shadi Abou-Zahra: Okay. Would love to hear more of your thoughts. I think GitHub, there is already this JSON realization of the role, trying to use I think we will take back that idea of revisiting Sirus and seeing if there are any particular things that it provides or benefits over EARL, but if you have any thoughts, let’s each look at what we haven’t looked at in awhile and see if we can come up with anything.

Wendy Chisholm: Yes, if you have done a gap analysis or planning to, that would be interesting.

Shadi Abou-Zahra: I will take a look. Next question is from Brian. Brian, do you want to introduce your question yourself?

Brian Bors: I’m Brian. I work for the Accessibility Foundation. I’m part of the European projects. And we are dedicated to making these rules. And I was interested in how such a manual implementation of a rule would look like and automated implementation would be a tool that actually tests this rule, that have bound and manual implementation would look like. For us as an organization stating yes, we use this rule. Would that be enough or would you have to provide a few steps that our evaluators are following?

Shadi Abou-Zahra: That’s a very important question. So with implementation in order to verify that there are implementations it would need to be publicly documented. And there are a few, there is the

Brian Bors: So like an instruction manual for a tester?

Shadi Abou-Zahra: Yeah, yeah. Again this instructional manual can be more educational if you are dealing with public sector bodies who, you know, are not the expert. It could be more technical if you are using that in your organization. And you are addressing more the QA testers. I think this manuals would look very different from one to the other.

Brian Bors: Yes. That answers my question. Thank you.

Shadi Abou-Zahra: All right. And Kathy writes on IRC just related to that the trusted tester is being updated now to include WCAG 2.0 Level A and AA and still in progress.

Wilco Fiers: Right. Just real brief. I wanted to add that just to avoid confusion, the rules format is not about an applet. However you express your data be it in some sort of serializible format like EARL or Sirus or whatever you are using, a plain Word document, I think that many still do, just you write your report. It doesn’t matter. The ACT Rules Format is really about making sure that results are consistent. About having a procedure that you follow to ensure consistency across different tests and across different organizations. So then how you express those results, that’s the second question, that’s a separate question. There are some references to EARL in the ACT Rules Format simply because you can’t talk about accessibility testing without saying okay, this is a pass and this is a fail. And we need to say that in one way or another. But the serializible format that you express those results in are not part of the rules format.

Shadi Abou-Zahra: Great. Thanks, Wilco. Let’s move on to the next question. Antonio says on IRC in chat, I’m very interested in understandable principle not only from a formal point of view. He goes on to ask accessibility content is less covered by WCAG 2, even in WCAG 2.1. Would like to contribute at this level in writing ACT rules. And the answer is as earlier said one of the clear limitations here we are not trying to define here any additional requirements. That’s a whole different beast to actually come to agreement on what is actually an accessibility requirement. We leave that exercise to the Working Group itself, which is having a lot of fun with that. If you want to get involved in that we do need all the hands that we can get on deck. There is a cognitive accessibility task force. There is a lot of work going on there. There is a lot of complexity of why this is so difficult to put some of these requirements in to testable statements.

From an ACT perspective what we have agreed on is that there will be so called the conformance tests, these are what absolutely meet WCAG and then best practices. We have agreed to have best practices. So we know that a lot of things kind of pass conformance but are really ugly and really bad practice. So yeah, we could have rules that say hey, you know, this technically passes the requirements but it is still a warning or a bad practice or something. So we could have these kinds of rules. We are trying to focus right now obviously more on the more low hanging fruit just to demonstrate and get this snowball rolling first, that’s our first priority. We can do some of this work as so called best practices. Later on I think a lot of the cognitive aspects are also addressed by techniques and how can create things in good practice. So we can have rules for that. But we do not intend here with ACT to define any additional requirements. I think that also answers Jennifer’s support for Antonio’s comment here of the relationship between ACT rules and additional WCAG requirements.

Charles says from the discussion possibility of a percentage outcome or other model in addition to simple pass/fail outcome. Here this is a recurring issue as well of so called metrics. Measuring accessibility and it would be so great to have, you know, to say this website is X percent accessible. But we couldn’t really come up with that means. We meaning W3C, we looked at this before, I will put an IRC and link for those who are not on IRC. It is W3.org/wai/2011/metrics. And so years and years ago in 2011 we held an online symposium like this one, only it was longer and we had people actually submit position papers on how they actually do some metrics calculations, how they come up with measuring the state of a website. We came to the conclusion that until today that’s valid, we haven’t found a single widely accepted process for doing that. There are a number of different approaches be it in Germany. There is an approach that is being used by the BITV Test and the idea of Barrier Walkthrough developed by Georgio Brajnik and others. There is several attempts but they have become so complex to calculate or a measurement that doesn’t reflect the accessibility requirements anymore. We have come up with four, I believe four requirements. I can’t remember them. Accuracy, sensitivity, complexity and there was another four dimension maybe to measure how qualitative a metric is and we haven’t found any that meet all these requirements. Thank you, Glenda. Thank you. Validity, reliability, complexity and sensitivity. And if you are trying to come up with how to measure the state of a website, you should try to address these like how valid is really the metric. How reliable it is and so on. This is also another piece that is out of scope of the ACT work. I know this comes very often in relation to testing and evaluation is having some percentage or some outcome. So far the WCAG model is pass/fail. And this is what we are using in ACT as well. The whole idea of aggregation and coming up with a final number, that’s out of scope of ACT work. And something that I think that continues to be very interesting research work here. And maybe Charles, you are involved in some of that, maybe Silver is the exploration for the new generation guidelines, 3.0 maybe that was, you know, maybe that will be a different model there.

Daniel Montalvo: I’m Daniel Montalvo, working in a group of ONCE in Spain as an accessibility coordinator. My question is, you know, whether or not you are thinking in this scope of the ACT project to incorporate some of the assistive technology behavior once we can conclude the rules have passed. And I gave you the example of the ARIA described by attributes. I think something could be done to document, you know, the screen behaviors and the fact that they are described, value is or going to be held after the accessible main because it computes as accessible description. The same could be said, for example, if you are going to create rules about the title attribute and so on and so forth. I don’t know if you are thinking about that or something out of scope or what are your thoughts about it.

Wilco Fiers: I knew it was coming my way. So the answer is yes, but only to a minimal extent. So what we couldn’t really do as part of ACT rules is specific documentation of things that may not work. Right? So we can’t explicitly say screen reader X doesn’t support Y. Because they may contest that. There may be different things that are disabled by default. There may be bugs. There is all sorts of gaps and difficulties that come up with your work as accessibility supports and that’s kind of what this is about. This is an accessibility support question. How much of this is supported in different assistive technologies. What is part of the ACT Rule is a section on accessibility support as part of every rule where it may be necessary that you document hey, we know there are assistive technologies that have some limitations. So you could in a rule where there are known gaps in assistive technologies you could add that information in. Say hey, as of the writing you know there are some limitations of certain assistive technologies without going in to high level of detail what those are and make that notes. Where it then continues and the reason why we put that section in we hope this will encourage organizations that actually do the limitation. So to developers, so accessibility agencies to take that information and run with it, to do some of their own testing to make sure that that information is then exposed to their clients to their users. So we did pick this up as one of the things that we are trying to encourage, but we can’t explicitly add that in to a rule because then you are essentially hard coding bugs in to documents which hopefully should live much longer than the bugs. So I’m hoping that kind of answers that question.

Daniel Montavlo: Yes. It does. I mean the accessibility and I was going to ask but you already mentioned. So thank you. I was going to ask about what your expectations, not just from your host, but from the way in which the other organizations are going to use that and you have pretty much answered me. So thank you very much.

Wilco Fiers: And I know that different organizations are looking at this. Some of them more explicitly than others. And it probably is a thing that the accessibility community as a whole has been kind of ignoring and thinking not being too sure about what to do with that. But I’m hoping by putting it in to the rules format that we do encourage people to start thinking about it more often because I think we do have a rather long way to go before we can do accessibility support properly.

Shadi Abou-Zahra: Thanks. We also have Kathy. Kathy, do you want to speak up or maybe introduce yourself and some of the work that you are doing if you want to answer your question? Okay. I’m not hearing anything. I am not sure if your audio works. Kathy writes I am curious how automated would be one set of ACT Rules.

Now so we actually started off with the idea of having rules that are categorized, let’s say automated and then other, semi automated and others that are manual. And then maybe with time as we have more Artificial Intelligence, and, you know, more advanced tooling and so on that some of these will wander off from being automated to semi automated or semi automated to automated. I came to the conclusion that it doesn’t make any sense at all in my head anymore. I think that’s a common for the rest of the Working Group is also at this stage. But this is where I am. And we actually move away from this categorization of automated, semi automated and manual because the intersection is just so strong. So let’s say this keyboard one, this rule that we looked at, well, you know, maybe rules will start showing you some of the components. Maybe they will be able to test the let’s say native components, native form fields but they can’t do that in things that are developed through JavaScript or something. So maybe you will have certain limitations. Maybe the tools can do five of the eight tests and then three others the developer has to do manually or evaluators.

We are not planning to differentiate anymore or to say this has to be done manually or this has to be done automatically but maybe provide here the applicability and expectations. This was one of the major changes between the previous draft and this draft is to make that shift to applicability and expectations that describe what the outcomes will be, not really how to do it and let whatever can be tool support be tool support. And whatever needs to be done manually and that also may be the difference between one methodology and the other. The question that we had earlier about how do these different implementations look like. Maybe one methodology, you know, involves the use of tools that will do more semi automated checking while the other methodology that is more manual of that’s a matter of taste and flavor. But this is where the methodology developers, this is where you people, where you take these rules and interpret them in to your methodologies or tools. They are not intended directly to the evaluator. I am seeing some agreement. Glenda, you have been so patient. Why don’t you go ahead.

Glenda Sims: I wanted to voice my support for the actuals and as a person working at Deque leading the accessibility experts there, we have more than 50 without our internal description of what is a failure and what’s not, I’ll get inconsistencies. That’s how I have got to consistency at Deque. I have put a lot of work in that. But if we want accessibility to move forward and to be attainable and easy and sustainable by people that haven’t done accessibility for ten plus years this is exactly what you need. So I’m thrilled about this. I can’t wait to contribute to it. Because it is just going to make those areas of WCAG that are more complex and nuanced to understand more attainables for everyone and with that being on that merry go round of is this a failure or not. I’m delighted and very supportive.

Shadi Abou-Zahra: Great. Thank you. And you had some comments on the categorization as well.

Glenda Sims: Absolutely. Absolutely. So when I heard you say that originally you had listed whether this rule was better suited, actually better suit for automated or was it something that you would expect semi automation or something you are like you know, the only way we can think of it today is manual, I do see value in that. But I get where it may not be in scope in this first round. Sounded to me as you were talking about that first round is let’s agree on defining how you test the accessibility test rule. And then you swing that through perhaps an update from an efficiency standpoint because knowing there are some people that I’ll walk up to and they will test something and they will be doing it manually where there is a tool that could have done it for them in under a second. That’s wasted time. So I see that that categorization while tricky to maintain its accuracy for is filled just manual, will be excellent for research for efficiency purposes.

Wilco Fiers: Just to add on a discussion. It was a painful thing to drop the categorization to me personally but we didn’t entirely drop. So what is the proposal is that instead of making it part of a rule and saying this rule is automated, this rule is not, what we are going to do we are going to track to implement the rule because we need to know that anyway because when you are when you want a certain number of implementations before a rule gets published and as part of that process tracking who implemented was it automatically, was it manual and that way we still get that data. But it might be different for different organizations.

Shadi Abou-Zahra: Yeah, thanks for this clarification. Good point actually. I didn’t talk much about that because we are not there yet but we do actually hope to have implementers declare not only which check they implement but also which individual test cases and how, whether they do that automatically or manually or whatever. And this will first give people an overview of which methodologies they can use for which purposes. But it would also I think also give us an understanding of where in researchers where there is more room for automation and work here.

Charu Phandi: Hi. This is Charu. I also work for IBM accessibility research. And I wanted to say this is an excellent effort. Thank you, Shadi, Wilco, Mary Jo for leading this. I think it is much needed because the interpretation of the WCAG is so subjective and, you know, it leads to inconsistency in the validation from organization to organization. I think this effort will really help harmonizing all that and also save all the duplication. Like, Wilco, you mentioned each organization is doing their own thing and they have so much not only consistency but duplication. I’m so glad. And hope to contribute more.

Shadi Abou-Zahra: Yeah, thank you. And all the participants in this work. This is a group effort of an entire task force and the invitation here for everyone to get involved. You can get involved either if you want to develop more of the specific area of the work, the rules format and the review process or if you want to get in to writing rules or commenting on rules I will provide again the links toward the end. We are actually starting to close. So if you want to say any final words now is the time to get on queue because we are kind of going towards the end here.

Daniel Montavlo: Yep. As I said in the IRC just I saw those references to the different abilities that user profile should in order to be able to perform these kind of semi automated testings or manual testings. In fact, there is the confirmation for the evaluators to actually confirm that the role has passed or not. Have you dropped this already as kind of dropped the automated and semi automated thing or not?

Shadi Abou-Zahra: Wilco, do you want to take that? In auto-WCAG there is information on user profiles. I think the answer is that this is outdated because the auto-WCAG is following on from what has been defined from ACT, but can you confirm that?

Wilco Fiers: Yeah. The auto-WCAG rules while like Shadi said latest drop got updated less than a week ago. We are dragging a bit and it will take time to update the rules. User profiles, I’m not sure that we are going to keep this. I think that falls in line with the question of what do you actually automate or not. That may be more of a question for implementers than is something that should be part of the rule. But that is a discussion we haven’t had yet.

I had a question. And the question was kind of for the rest of the group, especially people not involved in the work, and it is what, how could we help to let you contribute to the support. The more voices we have the better the quality and the more we use and are going to get out of this. Just a thought. Please jump on queue if you have anything to say.

Shadi Abou-Zahra: Very good point. What can we do to help you get more involved. Brian actually asked a very good related question. He is trying to get some of his colleagues up to speed what does he need to give them basically. What do they need to know. What should they read. I will answer that question in the closing and give some pointers but as Wilco says this is by no means the answer. This is what we currently have. We know there is still a lot more help that we need to do. We know that a lot of these terminologies and groups and it is sometimes confusing. So your feedback will be very helpful. Also after the call feel free to send us your thoughts and comments and we will consider this as we go along. And your input is very valuable and very much needed.

Katie Haritos-Shea: My name is Katie. I have been with the WCAG Working Group and at the W3C on various WAI groups as well as a representative for accessibility and other W3C Working Groups. But when Wilco first started this community group it was more than three years ago, wasn’t it? Like four? And everyone, this is obviously such a great idea. And, you know, it was a few people on Skype and besides watching the technology change, you know, at the W3C and somebody gets an idea and then it turns in to this real thing, some of them don’t make it and some of them do. But this is something that couldn’t make me happier being so much of my life having been involved with the development of WCAG 2 and working on WCAG 2.1. This is just something that is so needed. It has been needed. And for some reason had a hard time getting off the ground. But everyone who can please, please join this group. Either the Working Group or the community group. This is going to help all of us in our evaluations, in our work and are providing consistency really for the world and giving credibility to what accessibility is and how important it is and how it can be not something as contentious as somebody’s opinion. It doesn’t cover everything. We know that. But this is a very direct, organized way for competing companies to still be able to do what they can do but make sure that the outcome for users which is where, why this whole thing happens is going to be consistent. I don’t have any more time but I am thrilled.

Shadi Abou-Zahra: Thank you for these words and we will contact you, but actually this question I wanted to pick up. I think, you know, several questions now are how can we get involved or get other people involved either as good, either get involved yourself or push others to getting involved and that’s fine as well. There is a lot of work that needs to be done and there is a lot of ways in which you can contribute. So let’s move on to wrapping up and closing and giving you some of the pointers, some of the information on how you can get involved. I agree with Louis, I know a lot of the W3C stuff is confusing and overwhelming in these different groups.

Wrap-up and Closing

So slide No. 19, just to wrap up again, a reminder the ACT rules format is out for public review. The link is in the slides but I’ll read it out. It is w3.org/tr/act-rules-format. And that’s the technical specification really defining how these rules look like. It also points to some of the exampled rules that we saw today. So this is really the basis for all this work. We hope that this will be our last as we are maturing on this. Please send your comments by the 5th of April on this. This is one way you can contribute is by sending your comments on this. Anybody can do that.

The second is the auto-WCAG community group. This is as I said earlier a community group which is open to anyone. This is the place where rules are being developed. And in a community, so outside the Working Group. But if they follow this review process, demonstrate consensus and implementation, then these rules will essentially wander over to the W3C repository and become “formal” rules. The link is also in the slides. I will read it out. Auto-WCAG.GitHub.io/auto-WCAG.

The auto WCAG group is open to anybody who wants to participate if you want to send comments on rules that are in development. If you want to develop rules, if you have ideas for rules that you’d like to develop, if you want to be involved in the development. So, you know, this keyboard rule that is being developed if you want to be involved in that development, if you want to implement such rule that’s really the place to interact. This is where the rules development will be happening.

And then next is the ACT task force itself, which is part of the W3C Working Group and that’s where the rules format and the review process is being defined. So this is open to the W3 member organizations. In some cases we can have invited experts under very specific conditions. So but you can always comment all the work is being done in GitHub. You can always send comments and issues and the URL for this work is w3c.GitHub.io/WCAG-Act-rules. If you go there, Brian, that’s your question earlier, where do you send colleagues. We are trying to make that link, the last one that I just read out, that link really the place to go to. If you go to that link you will find some initial information and overview where the different bits and pieces are. We are going to be updating that page and providing more information, like the diagram that we showed today, maybe that helps also explain, put things in to perspective. So we will put a bit more information there. But that will be the place where you can send people. If anybody has further ideas of how we can provide more support, how we can help you or your colleagues get involved, do let us know.

So these are the places where you can get involved. And if you have any questions, write me or us. I see Wilco already taking the names of people. And we’ll be reaching out to you as well in order to help you get involved.

Good. I think we are coming towards the end of the call. Before we do that just, Wilco and Mary Jo, again any wrap up comments?

Wilco Fiers: No. Just to say thank you very much to Shadi. I wouldn’t have had the time for doing this. Thank you very much.

Mary Jo Mueller: Yeah, totally. Thank you for organizing this, Shadi. And I really do encourage folks to participate and contribute to our work. And we feel it is very important and it sounds like from you very important to you. So it all doesn’t come together unless we all work together to get it done.

Shadi Abou-Zahra: Well, speaking about staying in the loop, so this online symposium here today was organized by the WAI-Tools Project, the EC funded project. A lot of thanks for that contribution. Without this, a lot of the work would not be possible. Through this project we are planning to do more such webinars, seminars, maybe also face to face meetings. We are looking at maybe in the fall in Germany a face to face meeting. There is also the ICT testing symposium in the fall where we may be doing or, you know, hoping to do a pre-conference workshop maybe. So we are going to be trying to do more kind of events in the coming period with the support of this so you can get more involved or get your colleagues more involved. If you do have more questions or ideas, also if you go to that Web page, there is a list that you can subscribe to a newsletter to get updates from that project. So you get informed, when we organize such an event through the project. So if you are interested in getting such newsletter update go to that Web page and subscribe. It is only when we have events or certain results from projects, things like that we send around. So we will be doing more of this stuff. Today was specifically focused on people who are doing manual evaluation and developing manual methodologies because we need to do more work here.

We are really looking forward to hearing from you, hearing your thoughts how we can improve this work to better support you in developing your methodologies and making those consistent with the standards and interpretation worldwide. So let us know how we can help you in that mission.

I want to thank everyone for all this work, in particular the working group participants, task force participants and people in the auto-WCAG. And all this work is done by the people in the groups. And you have heard how many people it takes to contribute to get this international consensus. So thank you, everyone. This concludes today’s symposium. And we’ll stay in touch and all the best for everyone. Thank you.

Back to Top

This is an unpublished draft preview that might include content that is not yet approved. The published website is at w3.org/WAI/.