CSS3

Since the idea of CSS2.2 was raised, there’s been some discussion as to what it should encompass, who should be responsible for the spec, and what it should be called; here’s what I think:

First, it doesn’t matter what it’s called. Whether it’s referred to as CSS2.2, CSS2.1+, CSS3 Interim, or whatever, makes no difference. It doesn’t need to have a name at all; the important thing is that we have it.

Second, it doesn’t need to be an official recommendation from the W3C; in fact, it may be easier if it’s not. The optimal solution would be communication between developers and browser manufacturers, and – crucially – between the browser manufacturers themselves. What’s needed is an agreement as to which features are implemented, and to make sure those features are implemented in the same way; a de facto standard.

Finally, what feature should it include? For me, it has to be the elements which have already been implemented and tested in at least one browser for an amount of time sufficient for developers to have used them.

The most-requested feature is multiple background images; if you’re going to have that, background clip, origin and size would be wanted too. Border images would also be useful, as would an agreement on implementation of border radius.

Opacity, and with it RGBA and HSLA, box shadow, and text shadow would round off the decorative declarations.

Even if those few could be agreed on, a lot of workarounds could be avoided.

I would have said that multi-column layouts were less urgent, but as they are already part of the Gecko engine and about to be introduced in Safari 3, it seems that that should be part of the standard. Media queries would be pretty necessary as we move into the mobile era, too.

Nothing I’ve mentioned above would be unrealistic; most have already been implemented in at least two current or imminent browsers. As they are available, why are we being kept waiting before we can use them? Think of all the extraneous markup we could be freeing ourselves from!

Come on, browser makers: open up lines of communication and get talking to one another; float the ideas on your company blogs, see what your readers have to say. There’s a whole big community of developers who love to download nightly builds and test new features, and are hungry to improve their pages.

While preparing an example of using the colours specified in the CSS3 Colour module, I come across something after Minefield (the nightly builds of Firefox 3) updated itself. It now supports the resize property from the Basic user interface module.

Unfortunately, Minefield loads, then freezes and eventually doesn’t respond when trying to use this. This is unsurprising of a nightly release, and was likely due to the other CSS3 properties I was using on the example. The strange thing however is that the examples on the css3.info resize preview page do not work at all.

I decided to strip out a lot of the extra styles and content that I was using, and have an example where resize works both in WebKit and the latest Gecko builds. As this is part of another example that isn’t quite finished yet, there are a number of bugs, which may or may not be my fault. There are also a number of styles which probably don’t need to be there, but are part of a bigger example, that will be coming soon. I just wanted something quick that would show this in action. Feel free to suggest any improvements.

You can check out the resize example here.

So what are the issues? In the example I wanted to make a faux window, much like a desktop application window. The whole window should be resizable as one block. For this I placed the resize on the app-window div. The main contents of the window had to be scrollable when there is too much contents for the size of the window, but the rest of the contents shouldn’t scroll, such as the title bar (a h2 element). Therefore I put the scroll on the contents div.

In Safari there is a problem where both the h2 title bar and the main contents resize independently of each other. There doesn’t seem to be a way to resize the whole containing div where the resize is set. In Firefox this works correctly (or at least how I expected it to work). In Safari the content also can’t be resized smaller that the initial value of the div. This wasn’t an issue in an earlier build I was using while developing the example.

Minefield can be resized down to the minimum size and width requested, but it only resizes correctly horizontally. The div resizes vertically but the containing content div doesn’t resize at all. This may be a bug of mine as the contents seems to break out of the container slightly at the bottom. I haven’t had time to look into this.

One strange thing about Minefield’s support for resize is that it adds a cross-hair icon that allows one to move the contents around the page. This works correctly, unlike the resize, but I can’t find anywhere in the spec which specifies this should be allowed when using resize and it isn’t expected behaviour, or at least I wasn’t expecting it, and I didn’t specify that the div should be moveable. It would be nice to be able to specify this with a different property however.

Also, as I use a Mac, I expect the content to be only resizable from the resize widget on the bottom right hand corner (for better or worse). Minefield ignores this and has its own controls, which can’t be styled as far as I know. I also can’t remove the resize controls after clicking out of the resizable content. They stay until the page is reloaded. As resize has probably just been implemented, it is expected that their will be a number of bugs like this however, so things should improve as more people test out this and other CSS3 properties. Resize with HSLA/RGBA and overflow: scroll; is particularly buggy for instance.

I’m working on the original example, that I’ll post here at a later date, along with an updated version of the overview of the CSS3 Color Module that I posted on my personal blog recently. I guess I’ll have to report the bugs I’ve found to the appropriate places, unless it turns out I have bugs in my code instead.

I mentioned earlier this year a tutorial which shows how to use CSS 3 selectors to enhance hyperlinks. Of course, you don’t need to limit yourself to just the href attribute; with CSS 3 selectors, you can use the same technique for any tag which has an attribute. I’m going to give a couple of quick examples, which will output the following result (of course, you’ll need a browser better than IE6 to see them!):

• Home
• Download a PDF
• En Français

All icons are from the fantastic collection at famfamfam.com.

The first list item uses the general attribute selector to look for any tags with an accesskey attribute and display an icon to alert users to its presence:

#test_selectors a[accesskey] {
background: url('icon_key.gif') no-repeat 0 50%;
text-indent: 20px;
}

The second item looks for links with a type attribute value of ‘application/pdf’ (you do mark up links to documents using the type attribute, right?) using the exact attribute selector, then inserts an icon advising you of the destination document:

#test_selectors a[type='application/pdf'] {
background: url('file_acrobat.gif') no-repeat 0 50%;
text-indent: 20px;
}

Finally, if you have a multi-language site you can link to another version using the lang attribute with the value of that language, then use the language attribute selector to apply that country flag:

#test_selectors a[lang|='fr'] {
background: url('fr.gif') no-repeat 0 50%;
text-indent: 20px;
}

As always there is much more you can do with the selectors; if you use or develop this idea on your site, post links here for us to see.

Now that got your attention, didn’t it? You’ve probably never seen a site where Safari has a market share of 84%… Well neither had I. Until I looked at the browser stats for this site… I just had to share it with you all:

The Short Answer:

None of it.

The Slightly Longer Answer:

I’m in the process of updating the Preview area at the moment (sneak preview), and what’s immediately apparent is the low level of implementation of the new CSS 3 features across the major browsers. As IE6 is still the most widely-used browser, roughly 50% (and, slowly, falling) of the market has next to no CSS 3 support at all. A sobering thought.

With IE7 introducing support for attribute selectors, roughly 50% of the market can use those. You will still have to provide fall-back support for IE6, however, either with conditional comments or through graceful degradation.

Next most-widely implemented property is opacity; with support in all the key browsers other than IE, perhaps 25% of site visitors will see this effect if you use it. Again, make sure that your designs degrade gracefully if tempted to add this to your code.

After that, you can more or less forget it. The properties in the Backgrounds and Borders module have patchy implementation in browsers, and almost all use browser-specific prefixes, which you probably want to steer away from in a production environment as they are subject to change (see the border-radius conflict as a good example of why they are tricky to implement).

text-shadow should gain support from Safari 3 and Opera 9.5, but even being generous that’s only around 5-10% of the market. Most of the other properties have little or no cross-browser support.

What You Can Do About It:

Get behind Andy Budd’s ‘CSS 2.2’ idea. Think about it. If you have a blog, discuss it there. Write to browser manufacturers and the W3C. We’re putting together a campaign website to promote the idea, so get in touch with us and offer support.

We want – no, need – these new properties, to do away with many of the non-standard or non-semantic solutions we have to use today to provide complex solutions for simple problems. CSS 3 provides many of those solutions, but they won’t be implemented cross-browser until they become standard; that can be via the W3C, or a de facto standard agreed by browser manufacturers. But however that standard is made, it won’t happen unless there’s concerted pressure from the development community.

I’ve written a number of posts about CSS3 on my personal blog, so when I was asked to write on CSS3.info I jumped at the chance. To get the quick disclaimer out of the way, my day job is working for Opera Software as their Chief Web Opener. Any thoughts are my own, and any use of CSS3 properties doesn’t imply that they will or will not be supported by Opera in an upcoming release, unless otherwise explicitly stated.

With that out of the way, I was discussing the difference between opacity and RGBA in the office, and thought that it would be useful to write an example showing what the later can be useful for. In the main part, using RGBA is different from opacity, as the transparency only applied to the property you are setting the colour of, while opacity sets the transparency of the entire element. For example, setting opacity on a p element will set text as well as the background colour to the setting chosen, while setting RGBA will only make the background colour transparent, and not the text, if background-color property is used. I decided that overlaying a text block over an image would show this quite well, so I’ve knocked up an example using this to style a figure contain within an article. This example can be found here.

To mark-up the figure, I wrapped the img and the p containing the caption in a div element. While this wasn’t strictly needed, I feel like the figure is a division of the page, and so give some semantic meaning and convenient grouping of the figure’s components. I gave it the class name of figure as suggested in Dan Cederholm’s blog post on marking up figures and Chris Messina’s post on a similar topic. I then added a class of caption on the caption, to state explicitly what the element contains.

The first important point to note when using CSS3 is that only the very latest browser versions support any given feature, if any do at all. Therefore as Peter points out in the previous post, it is important to use graceful degradation. this is especially important when using RGBA colour values, as browsers are instructed to ignore the CSS statement if it doesn’t understand the Alpha channel part. It doesn’t fallback to just using the Red Green and Blue channels. In the example used here, the figure caption has been set to red with a alpha channel value of 0.6, using background-color: rgba(255, 0, 0, 0.6);. A value of 1 would mean the colour would be fully opaque, while 0 would be fully transparent. For browsers that don’t support RGBA, there would be no background colour at all, and the text would be illegible. To solve this, a regular background colour is given, in this case background-color: red;. This has to be included before the statement using RGBA, so that it will be overridden by browsers that do support RGB with Alpha channel.

To get the caption to overlay the image, the paragraph in question was set to use position: relative;, and given left and top values that position the element in the correct position; in this case the bottom right corner of the image. A positive z-index value was given to ensure that it was displayed above the img element. I was tempted to use em values for the sizing and placement of the caption and the padding/margin of the figure, but as the image uses pixels for it’s height and width, it made the cross browser placement of the caption inconsistent across the browsers tested (Opera, WebKit, Minefield). Even using pixels, Minefield (current nightly of Firefox 3) is 1 pixel out on the vertical placement compared to Opera and WebKit. I’m sure there is probably a better way to place the element than what I’ve used in this example.

As this blog is about CSS3 I added a couple more features. I gave the caption a slight shadow (okay, not strictly CSS3) using text-shadow: 2px 2px 2px black; (This works in Opera Kestrel and Safari) and gave the whole figure a slight 3D effect by specifying a shadow around it. This was done using box-shadow. In truth no browser supports this without prefixes but Safari 3 supports it using -webkit-box-shadow. I also initially added a CSS3 selector to style the first line of text in the article as bold, using .content p:first-of-type:first-line. Unfortunately Safari 3 currently has a bug where it will apply this to every p and not just the first one, while Firefox totally ignores it as it doesn’t support the first-of-type selector. Safari will probably fix this bug before version 3 is released as I noticed they’ve just fixed a similar bug with using p:first-of-type:first-letter, but as I can predict when the first paragraph will come in this example (after the figure) and it works across each of the three browsers I tested, I just settled on using .figure+p:first-line. If anyone has any questions comments or has any better way of doing things, then feel free to leave a comment.

Graceful degradation means that your Web site continues to operate even when viewed with less-than-optimal software in which advanced effects don’t work.

Fluid Thinking, by Peter-Paul Koch

With CSS 3 so tantalisingly close (and yet so far away!), it’s fun to play around with some of the new cosmetic features. In fact, we can even start to implement them on websites – as long as provision is made for users with older browsers.

I’ve made a quick example of a horizontal, tabbed navigation using CSS 3, which degrades gracefully for older browsers.

Here’s the markup:

<ul>
<li><a href="">One</a></li>
<li><a href="">Two</a></li>
<li><a href="">Tre</a></li>
<li><a href="">Qua</a></li>
</ul>

The CSS is quite lengthy, so view the source of the example page and you’ll get some idea of what I’ve done.

If you’re using IE6, you’ll see just a row of plain, grey boxes which don’t animate when you mouse over them. IE7 and Opera users get a slightly better experience, with the text going white when hovered:

li:hover a { color: #fff; }

Firefox and Safari 2 users get it better still, with border-radius and multiple backgrounds respectively supplying curves to the tabs:

li {
url('corner_fff_topleft.png') no-repeat top left,
url('corner_fff_topright.png') no-repeat top right;
-moz-border-radius-topleft: 25px;
-moz-border-radius-topright: 25px;
}

It’s when you get to Konqueror (and, presumably, Opera 9.5), however, the new styles really kick in; use of the nth-child declaration allows for alternating colours, and text-shadow give a slight illusion of depth when hovered:

li:nth-child(1n) { background-color: #9f9; }

li:nth-child(1n):hover { background-color: #0c0; }

li:nth-child(1n):hover a { text-shadow: 2px 2px 2px #000; }

li:nth-child(2n) { background-color: #6f6; }

li:nth-child(2n):hover { background-color: #090; }

This was a very quick and easy example, but I’m sure you can imagine much more creative uses for graceful degradation. It’s a useful technique if you have the time to implement it, and comes in very useful when trying to persuade people to upgrade to a better browser: “If you think that website looks good now, take a look at it on my computer!”

As liquidat points out quite correctly, it is NOT Opera that is the first to support all the tests in our CSS Selectors test. Those bragging rights go to Konqueror, as we pointed out in January already. Release 3.5.6 of KHTML was the first to support all the tests, not any other browser, sorry David :) .

We, the CSS3.info team, are very very proud to see our Selectors Test being used as the de facto test of selectors compatibility though, and we hope that other browser projects will be using it too!

While I’m at it, please remember that our preview section is only a showcase of css3 properties that are currently implemented in any browser. If any browser were to support all the things in that section, this does not constitute “full css3 support”, as there is no such thing yet. This also means that it could be that we are missing features in that section that are currently implemented in a browser, if that’s true, please drop us a line!

Opera have just announced version 9.5 of their desktop browser, currently codenamed ‘Kestrel’ – and as you can see from the screenshot they’ve chosen, it passes the CSS 3 selectors test with flying colours!

They mention that CSS 3 support will be improved (text-shadow is provided as an example), as well as superior SVG support, and a new Javascript engine will allow better access to websites with wonky coding.

This sounds like a really exciting release, and should give Opera the accolade of being the most CSS 3 compliant cross-platform browser. Weekly builds will be available shortly.

With the release of Safari 3, there are now two browsers with (browser-specific) implementations of border-radius; unfortunately, the two implementations are different. The problem is that there is an unresolved ambiguity in the CSS 3 working draft.

The draft proposes four declarations, which describe the four corners of a block:

border-top-left-radius
border-top-right-radius
border-bottom-right-radius
border-bottom-left-radius

Each of them should accept two values, which define the radii of a quarter ellipse that defines the shape of the corner; this allows for irregular curves (take a look at the diagram in the draft if you need clarification, or see this example of a box with border-radius: 5px 20px, horribly rendered in Safari for Windows).

Safari, with the prefix -webkit-, accepts these. Mozilla, with the prefix -moz- (and differing declarations), accepts only a single value and, therefore, only regular curves.

At first glance, it would appear that Mozilla are in the wrong; however, their implementation is due to the ambiguity I mentioned earlier.

This ambiguity comes about in the border-radius shorthand property; if you enter a double value in this you’d expect to apply the irregular curves to all four corners:

border-radius: 5px 10px;

If you wanted to have four different irregular curves on the box, you’d have to provide eight values to the declaration:

border-radius: 5px 20px 10px 5px 10px 20px 20px 5px;

But what if you wanted to have two corners with one value, and two corners with a different value?

border-radius: 5px 10px 10px 20px;

The problem is that this could be confused for four corners with regular curves. In order to get around this, you’d still have to provide eight values:

border-radius: 5px 5px 10px 10px 10px 10px 20px 20px;

In fact, from the brief testing I’ve done (and I can’t find any documentation), it seems you can’t do any of that; unless I’m missing something, the shorthand declaration in Safari accepts only 1 or 2 values, to provide either regular or irregular curves which are applied to all four corners. If you want different irregular corners, you have to supply values to all four declarations:

border-top-left-radius: 5px 20px;
border-top-right-radius: 10px 5px;
border-bottom-right-radius:  10px 20px;
border-bottom-left-radius:  20px 5px;

Mozilla avoid this by going against the spec and allowing only regular curves; so you can provide 1, 2, 3 or 4 values and it’s all perfectly clear.

This problem is down to interpretation of the draft. I personally think Mozilla’s non-standard solution is better – it’s less flexible, but easier to understand – but can’t blame the Safari team for following the standard in their implementation.

It will be interesting to see which comes out on top; in the meantime, if you want to use border-radius in your code the only way to get them to appear the same on both browsers is with a single value for four regular corners:

-moz-border-radius: 10px;
-webkit-border-radius: 10px;