Vendor Prefixes: Blink, and You’ll Miss Them

Google has announced that they will be forking WebKit to create a new rendering engine for Chrome called Blink.[1][2] Google will be able to remove millions of lines of code from Blink that isn’t needed for Chrome, and will be able to work more aggressively to develop Chrome-specific features into Blink.

This won’t unsettle things too much—at least not right away. The latest builds of Canary are already using Blink, and there is no perceptible difference. The CSS -webkit- prefixes even still work. So how long will it be until we have to switch over to using a -blink- or -chrome- prefix for our rounded corners? The answer to that question is: “wrong!” There will be no -chrome- prefix, or any other vendor prefixes at all, moving forward. Instead, you’ll have to change a setting in your browser to enable experimental CSS features.

Vendor prefixes seemed like a good idea. And maybe they would have worked, if developers had used them as intended. Eric Meyer wrote the seminal article advocating vendor prefixes, Prefix or Posthack. He explained that we could use vendor prefixes to escape the conflicting standards of the bad old days, when the same code was rendered differently by different browsers. Due to Internet Explorers broken box model, width meant one thing in Trident (IE’s layout engine) and something entirely different in Gecko (FireFox’s layout engine). Furthermore, because these differences were alive and living on the internet, there was no way for either browser to fix (or compromise) its definition to unify the web without breaking existing pages. Internet Explorer eventually fixed it’s box model, but it has been an arduous, difficult journey, one that has not yet reached it’s destination. Despite years of very clever attempts to fix things with the creation of quirks mode, doctype switching, and for IE, conditional comments and the problematic compatibility view (which still persists in IE10!).

With vendor prefixes, none of that would happen, because new CSS features would go through a vetting period, during which the feature could only be accessed by prefixing a code to the property name. After a time of testing, and when the standard was defined more definitively, the browser could begin supporting the property without the prefix, and nobody would have incompatible CSS anymore!

This quickly created a situation where the simple task of rounding some corners turned into this mess:

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

When it comes to background gradients, things are even worse:

background: -moz-linear-gradient(top,  rgba(255,255,255,1) 0%, rgba(255,255,255,0) 100%); /* FF3.6+ */
background: -webkit-gradient(linear, left top, left bottom, color-stop(0%,rgba(255,255,255,1)), color-stop(100%,rgba(255,255,255,0))); /* Chrome,Safari4+ */
background: -webkit-linear-gradient(top,  rgba(255,255,255,1) 0%,rgba(255,255,255,0) 100%); /* Chrome10+,Safari5.1+ */
background: -o-linear-gradient(top,  rgba(255,255,255,1) 0%,rgba(255,255,255,0) 100%); /* Opera 11.10+ */
background: -ms-linear-gradient(top,  rgba(255,255,255,1) 0%,rgba(255,255,255,0) 100%); /* IE10+ */
background: linear-gradient(to bottom,  rgba(255,255,255,1) 0%,rgba(255,255,255,0) 100%); /* W3C */
filter: progid:DXImageTransform.Microsoft.gradient( startColorstr='#ffffff', endColorstr='#00ffffff',GradientType=0 ); /* IE6-9 */

Not only does each browser get its own line, the WebKit browsers actually use two lines, due to a change in syntax. The messiness of this is annoying, but it isn’t actually that huge of a problem, in terms of web compatibility. In fact, it saved us from having to write crazy javascript to detect the browser and version number in order to give one syntax to < Webkit 4 browsers and another syntax to all others. The real problem is that this is how most developers began coding their sites:

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

or even just:

-webkit-border-radius: 10px;

You can see how this sort of abuse caused problems. Developers got used to Chrome being on the bleeding edge of CSS3 features, and began coding specifically for it without taking the time to check whether other browsers had supported the features yet. Effectively, vendor prefixes had become browser detection all over again, and developers were using this browser detection to block functionality from all browsers except WebKit. As you can imagine, this annoyed other browser vendors, like Opera, who were working hard to keep their browser up to date with CSS3, but which was blocked from applying the styles due to lazy coding. Developers had turned -webkit- into the very -beta- prefix which Eric Meyers dismissed in his article.

The -beta- prefix was proposed by Peter-Paul Koch as a compromise to his original proposition to abolish vendor prefixes. Koch’s vision was prophetic. In March 2010 he wrote:

Eventually Opera will discover that plenty of sites use -webkit-transition, but not -o-transition. Will Opera start to support -webkit-transition, too?

Despite a widely circulated call for action to web developers to change their ways and prevent “judgement day,” two years after Koch’s prediction, Opera announced it would support -webkit- prefixes. Mozilla had been planning to do the same. Microsoft’s massive attempt to change the bad behavior of developers had failed, and they too would follow suit.

It looked like things were going to get really messy. Then suddenly, Opera announced that it was abandoning its rendering engine Presto, and would instead use WebKit. While many saw this as a sad day for the web, the silver lining was that judgement day had been pushed back a little. Furthermore, Opera is adopting Blink with Google, and Mozilla is creating yet another rendering engine called Servo. So we aren’t looking at a rendering engine monoculture or a return to the “bad old days.”

So with all these new rendering engines, what will happen to vendor prefixes? They’re all going to go and live with the <blink> tag and IE’s broken box model. Mozilla is heading in the same direction as Chrome: “avoiding vendor prefixes by either turning things off before shipping or shipping them un-prefixed if they’re stable enough.” This is where all the vendors are headed, and the W3C working group on CSS has already put together a policy that rang the death knell of vendor prefixes.

This article originally appeared on the blog. It was cross posted by permission the owner of Shoutleaf (me) and the author of the article (me).