wheresrhys

I read somewhere the other week that jQuery is deprecating its jQuery.browser method, which means that in the future (though not yet, as the method still works but will not continue to be developed/supported, and will eventually be dropped) you will not be able to directly ask jQuery which browser it is being run in.

The rationale behind doing this is on the face of it sound: The reason you want to check what the browser is is that you want to check if a feature of javascript is available in that browser… so why not just ask if that feature is available, then the question of which browser it is becomes irrelevant. This will enable your script to continue to function (most probably) even in the unlikely event that eg Microsoft bring out a patch for ie6 that makes it fully support javascript.

But there is a flaw in this reasoning. It assumes that the only reason you would want to serve some different script to a browser is because of flaws in the browser’s javascript implementation, but in fact there are other reasons for doing this too, principally the browser’s failure to render CSS correctly.

While the new jQuery.support method does include tests for boxModel (ie6 and 7′s incorrect rendering of the CSS box model being the principle reason behind serving different CSS to different browsers) and one or two other CSS bugs, this simply isn’t up to the task. The box model is broken in ie6 and ie7, but broken in different ways, so normally to fix a layout bug I will want to serve different CSS to both.

Using various combinations of jQuery.support, eg

!jQuery.support.boxModel && jQuery.support.objectAll

you can still detect ie6 and ie7… but is this really an improvement on jQuery.browser?

jQuery.browser.msie && jQuery.browser.version.substr(0,1)=="6"

With jQuery.browser it is wholly transparent what you’re trying to detect, and if the code within the conditional’s braces contains a hefty amount of CSS rules your average developer should be able to work out the reason for the browser detection is to fix CSS bugs.

One other reason for explicitly detecting a browser (which I need at the moment as I’m developing a jQuery plugin) is that different browsers render form elements differently (the most obvious black sheep is safari’s making most elements look blue and glossy, but there are subtle differences in other browsers too). If I, say, want to fake a <select> element  by using a <ul>, and want it to look convincing in all browsers I need to supply each browser with differentiated CSS. jQuery.browser would be the obvious way to do this. No matter how many CSS bugs get fixed in browsers, how to render form elements by default isn’t specified by CSS/html, so there will always be this reason to want to detect browsers.

jQuery.browser is a very useful feature and it’ll be a shame when it goes.

Marking up web pages semantically is normally more or less a doddle. Most tags are named (or have abbreviated names) that are straightforward. If you’re writing a  paragraph, wrap it in a <p> tag. If you’re writing an ordered list put it in an <ol>… and so on and so forth. And it’s actually less effort than typing eg <div class=”list”>.

But there are nevertheless lots of examples where the tags offered by (x)html seem hopelessly limited in their semantic meaning. Recently I worked on this site. On the home page is a list of holiday offers, and I ummed and aahed over how to mark it up for a while. Previously I would always follow the same practice – lists of complex items would be marked up as a list, and whatever more complicated markup needed would be placed inside the <li> items. But I’ve read somewhere that putting too much information within a list item can actually be a hindrance to accessibility,  so this time things would be different.

Design of list of holiday packages

The design superficially lent itself to a simple nested list:

<ul>
  <li><h3>Blue Reef Resort</h3>
     <ul>
        <li>All inclusive</li>
        <li>8, 15 dagen</li>
        <li>Marsa Alam</li>
        ...
     </ul>
  </li>
</ul>

But it’s hardly satisfactory. The items in the sub-list are in fact properties of the particular holiday offer. In fact, the whole thing is more like a table, which becomes clear if we just move some bits around.

Modified design to show true semantic structure

So to achieve the required design all we need is to hide the table header and move some of the cells around. The first task is easily completed with a display:none, but the second is alas not very feasible. While support for floating and positioning table, tr and td elements is supported in ie8 and other good browsers, in ie6 and ie7 there are big problems as they more or less ignore that the <tr> element is there. You can’t float a tr, you can’t give it position:relative (in order to absolutely position td’s within it) but, bizarrely, you can absolutely position a tr, though this is of limited use. Racking my brains for a use the only one I’ve come up with is in order to give you the ability to absolutely position the tr’s td’s absolutely in relation to the tr, but then you would need a different style for each tr to avoid them layering on top of each other. Which admiteddly could be generated dynamically as an inline style, so perhaps not all that hopeless a situation.

ie6 and ie7 also don’t support floating on table cells either, in case you were thinking that was an option.

As it turns out, I may have discovered a new bug in Internet Explorer 8. While writing this I set up a test page to make sure what I thought was the case was correct (turns out I was, in my main thesis, incorrect, as I thought HTML/CSS had no provision for positioning table elements). Here are 3 screenshots:

Correctly positioned table elements (firefox)

Incorrect table element positioning (ie8)

Incorrect table positioning (ie6/7)

Once again, ie has problems dealing with relative positioning on tr’s. On the face of it it’s not as bad in ie8 as in ie6/7, but what’s happening is that instead of letting the tr’s confuse the positioning completely, it just ignores them, and goes ahead and positions relative to the next positioned ancestor, which is een worse if you look at where the cells get put relative to where they shoudl be – ie6 and ie7 nearly get it right (but this might be a fluke that a larger table would unmask), but ie8 is completely wrong.

The long and short of this article is that there are lots of instances where a table should be the natural choice of markup (given a little thought about what the content’s structure is), but once again thanks to Internet explorer, it’s going to have to be a nice idea that will have to be shelved for the time being.

So, unless you fancy individually positioning each tr with its own style rule, nested lists it is for now.

Below is a list of links to the browser download pages for each of the major web browsers:

• Firefox
• Safari
• Internet Explorer (ie)
• Chrome
• Opera

Each design is essentially the same – a little bit of text, some images (normally of the browser) and a massive link to download the latest version. All, that is, except for ie. (It also bears mentioning that the firefox page commits the usability cardinal sin of not actually mentioning it’s a web browser).

Here is a screenshot of ie’s page. It’s difficult to see on this small image (but click it for bigger), but I thin kit must be one of the most ill-thought out, though supremely important, pages I’ve ever seen:

• Despite ie7 still being the officially stable and sanctioned browser, the most prominent text and download link is for ie8.
• You will already have clicked “get it now” once from the ie home page, but the top part of the home page and the download page are exactly the same, so it looks like “get it now” is just a download link that doesn’t work.
• Once you figure out it’s not the download link you have to scroll down the page to actually download the software. Scrolling down to get to the most important bit of the page?! In the words of Kyle from South Park, “Really??!!”
• You have to choose your operating system in order to download an installer which then downloads components appropriate for your OS. Given that it ends up downloading additional files anyway, why not just make it a generic download that detects your OS and then downloads all the relevant components. Or even detect the OS with javascript, with the select box as a fallback.

It beggars belief that usability mistakes like this can be allowed through.

Having said that, I bet most users never see the page as ie updates automatically… but still!

On two not unrelated notes:

• I wish open source projects, in particular maybe the core design of sourceforge, would make use of prominent download buttons for the latest stable version, instead of the maze of  download links for every version under the sun. It woudl surely increase the use of what can be very useful tools.
• I downloaded ie7 finally as I needed it to install IETester , a great tool for installing ie5.5, ie6. ie7 and ie8 simultaneously on either windows vista or xp.