My favourite pieces to commission when I jobbed as a packer at a chainsaw factory were the health and safety instruction videos. "Never check the level of petrol by holding a lighter to the opening" and "Never keep the blade between your legs when trying to start the chainsaw" were just two of the highlights.

Health and safety measures are important – they ensure that our work environment is enjoyable and will not make us sick. Just because our job is handling user agents and typing funky words into an editor does not mean we shouldn’t follow some of our own.

Right now, DOM scripting with JavaScript is hot again (the over 3000 unique hits on my post about outdated JavaScript techniques proved that to me) and we run into the danger of overusing it or using it in the same obtrusive ways we used JavaScript when DHTML was the flavour of the month.

When your tool is a hammer, everything looks like a nail, or – in some cases – a thumb.

Let’s not bash anything in sight or mince our own thumbs. DOM and JavaScript are there to enhance the structure and interact with the presentation of a web site – and not replace them.

Ideas that prevented me from repeatedly hitting my thumb:

1. Don’t create HTML that should be there without JavaScript. Re-use what is already in the page instead. An accessible site starts with a semantic, well-structured HTML document. If that is not given, there is no chance we can be accessible. Not creating HTML via DOM makes the product a lot easier to maintain. Non JS-savvy colleagues won’t have to butcher your code to make a change on the page.
2. Generated HTML follows the same rules as written HTML: Don’t create invalid HTML. We stopped using CSS to make elements look like headers instead of using real header elements. The same applies to the DOM - redundant HTML elements to fix a design or add a design feat are – well, redundant, no matter what technology was used to add them. By the way, this also applies to server side scripting.
3. Be aware of visitor and user agent restrictions. An example are my fabulous clickable headers of the Unobtrusive JavaScript course. They look handy, work cross-browser and were developed with a clear separation of structure, presentation and behaviour. Where they fail is when you try to use them with a keyboard. The power of the DOM seduced me to turn a header into a hover-able and clickable element, but there is no way to use it without a mouse. That is why I will add a real interactive element to the headline – a link – to make them work for keyboard users (and possibly screen reader users) as well (thanks must go to Stéphane for flagging that up to me).
4. Understatement is class. Before adding all kind of cool new JavaScript features and make everything hover, click and move it is a good idea to lean back and think: "Is this really necessary? Do I add it to help the user or do I add it because it is cool?"
5. Don’t break too many conventions. As posted earlier, I see the web as a secondary media, and assuming that the visitor pays detailed attention to our web sites is more narcissism than reality. If we need to explain functionality to the visitor in a piece of text, there is a big chance that we will confuse rather than help. Example? Users hitting the back button on AJAX apps and pure Flash sites.
6. Leave a clean desk. Especially in distributed developments it is of utmost importance that everybody speaks the same "code language" and that handovers are painless and quick as the code is already properly documented. Define an in-house coding standard, comment your code where applicable and there will be a lot less stressed faces and moaning when people get assigned to projects.

Notice: The following is a “best practice document”. You can follow its advice and live happily ever after, but there might be situations where you cannot apply the ideas mentioned within. This is especially the case when you have to maintain an old product or complex web application. You cannot replace everything in those in one go – as you are very lucky indeed if you get the time and budget – but you can tackle those step by step.

According to many web designers at @media2005, JavaScript is sexy again, and in demand. Ajax is the new CSS was a quote in one of the presentations, that – taken out of context – makes my stomach go wonky.

Working largely with b2b sites and restricted environments, I never knew it was out of fashion.

It is a tool, much like a shovel, a screwdriver or a towel. While a towel is handy in every situation, as every HHGTTG reader knows, JavaScript has its place and its merits, but also its limitations and places it shouldn’t be applied to.

You can use a screwdriver to screw in screws or to clean your ears, however, the latter needs real skill, determination and a lack of fear of injuring yourself. It is much the same with JavaScript, you can use it for great things, but you can also use it to make a product inaccessible, confusing and not usable.

Here are some danger signs you encounter in many a bad script example and tutorials written in the spur of the moment rather than following proper research.

Things that should make you go Noooooooooooooooooo!

1. document.write What this does is write out content to the HTML document, inside the body of the document and merrily mixed with the markup. Bad web developer, bad! Go to your corner! A better solution: reach the element you want via getElementById or getElementsByTagName and then insert your newly created content (via createElement and createTextNode) there.
2. This you encounter a lot in “Accessibility Tutorials” and it should “make sure that every user can use your site”. Actually, it is a sign that the script you use (or did) is bad and should have never been used in the first place. It is like using white-out on the wallpaper after you used your crayons there. Comment: I was asked to clarify this, as web application developers who have to rely on JavaScipt (which is a flaw in the application design IMHO) use noscript to tell the users they need JavaScript enabled. Normally this is added as a warning message at the start of the non-working page. The more logical option in this case would be to have the "no JavaScript" message in the document and replace it with a link to the application when Javascript is available. Check this example page on how to avoid noscript. Turn JavaScript on and off to see the difference.
3. href="javascript:", onclick="javascript:" There is no such thing as a JavaScript protocol on the web. Links use protocols to connect documents. Create real links or don’t use any link at all – do or do not, there is no try!
4. onclick="void(0)" Why would one put effort into creating something that by definition is not doing anything? Glossing over bugs and bad scripting, that’s why. A link that only points to a JavaScript function should be added via JavaScript.
5. document.all , document.layers, navigator.userAgent Unless your defined project environment is MSIE 5.x and Netscape Communicator 4.x (my condolences if it is), there is no need for that any longer. Object detection is so much better than trying to guess what the browser in use is, and the W3C DOM is also more likely to be used in future UAs than the bespoke Netscape or Microsoft ones. (If you say “huh, what?” now, just trust me that everything with document.getElementById is much more stable than the other two mentioned earlier)
6. onmouseover="myCall('I','pity','the','foo',1233,'I aint going on no plane')" Anything crucial to the user experience that you generate via JavaScript needs to be in the document anyways – for users without JavaScript. Reusing this markup is a lot easier, cleaner and more maintainable than sending a lot of parameters. If there is any reason to send parameters, a simple ‘this’ does the trick in most of the cases, as you can navigate through the DOM from there.

On request

And what about innerHTML?

As some pointed out in the comments, innerHTML is another way to create content on web sites. In some cases, it is even the fastest method, as Peter-Paul Koch found out when comparing methods to generate content. However innerHTML is neither standard, nor applicable to any DOM scripting outside the browser environment. There is an excellent discussion about innerHTML on developer-x. My personal view? innerHTML makes it a lot easier, but advertises HTML documents as strings rather than node trees, and that makes it hard for junior developers to go further in their scripting ventures.

Partly inspired by Robert Nyman’s recent post about JavaScript

The reactions of the audience at @media, and many a chat afterwards made it obvious to me that the concept of JavaScript adding the behaviour layer above the HTML markup (or structure) and the CSS (or presentation) has become commonly known.

This is nothing new, after all, Zeldman’s book explained that in 1999 with a nice movie metaphor:

If your web page were a movie, XHTML would be the screenwriter, CSS would be the art director, scripting languages would be the special effects, and the DOM would be the director who oversees the entire production.

However, when it comes to accessibility and ease of maintenance, I’ll dare to make things more confusing now, by telling you there are five levels of web development, each of them with its own technologies and methodologies. Only when all of them work properly and are aware of standards and accessibility / usability constraints we can work without breaking a sweat.

The levels are:

• The behaviour layer – Executed on the client, defines how different elements behave (Javascript, Action Script)
• The presentation layer – Displayed on the client, the look and feel of the site (CSS, images)
• The structure layer – Displayed on the client, the mark-up defining what a certain text or media is (xHTML)
• The content layer – On the server, the text, image and multimedia content that are in the site. (XML, DB)
• The business logic layer – On the server, defines what is done with incoming information and determines what is returned (Backend, various environments)

The Business Logic Layer needs to

• provide the different states in a styleable and changeable format

The Content Layer needs to have

• Text that is easy to understand
• Text that makes sense without a visual representation (wrong example: “click on the links below”)
• Alternative text for every image, sound piece or video that is content
• Text that is fit for the web (KISS, structured into headers, paragraphs and lists)
• Explanations of Acronyms and Abbreviations
• Content images need to be unambiguous for the colourblind and text in images needs to have a sufficient size and contrast.
• Information to the users of changes necessary to her environment (example: “opens in a new window” or “PDF document, 182kb”)

The Structure Layer needs to

• provide a consistent document structure with the content available as early as possible
• provide the “invisible accessibility enhancements” like accesskeys, link relationships and table serialisation aids.

The Presentation Layer needs to

• ensure that text can be zoomed without making the site unusable
• ensure that the interactive elements of the site are easy to find
• ensure that images and foreground and background have enough contrast and are unambiguous to the colourblind
• give the site a consistent navigation
• Aid the user through business processes
• Separate content into easily understandable units

The Behaviour Layer needs to

• ensure that all the functionality is available to the user regardless of input device
• make the user experience as easy as possible by cutting down on options until they are necessary

Raising awareness of this might help in the future, and accessibility / usability will not be considered a “plug-in” added in the final week of the project lifecycle any longer.

There has been quite an interesting thread on css-discuss about solutions how to make a blog post construct like

headline
date
text

link. In compliant HTML you cannot nest block elements in a link, therefore you would need to repeat the link three times – one inside the header element and two inside the paragraphs for date and the text. This poses an accessibility issue of three links to tab through / listen to / appear in the document’s links list.

Ingo Chao came up with a pure CSS solution.

The only problem with that one is that the link text will be just far too long, and you have no structure in the text any longer.

I decided to harness JavaScript to do the dirty work.

I just finished EasyNav, a PHP script that automatically highlights the current page in a site navigation.

All you need to use it is generate your navigation as a nested list of links and include EasyNav in your pages. The current link gets automatically replaced with a “strong” element and gets the ID “current”. The parent list item gets the ID “active”.

This allows you to use the CSS examples of Listamatic to style your navigation.

EasyNav also allows for different navigations to be generated, dependent on the ID of the root UL.