The Hidden Internet Explorer Tax

It's easy to tally the costs of upgrading Internet Explorer, but the costs of not upgrading are harder to see.
Many companies justify their reluctance to upgrade from Windows XP and/or Internet Explorer 6 by pointing to the high costs of buying new systems or upgrading old ones. These are the costs that are easy to see, since they end up as a line item in the IT budget. When management asks the question, "Is this really necessary?" it can be hard to make the case for spending that money when faced with other priorities.

Yet there are hard costs associated with not upgrading, and they can be significant. This past weekend I was at the jQuery Conference in Mountain View, Calif. Yahoo's Nicholas Zakas presented data showing that although IE6 and IE7 only comprise about 20% of browser market share, they require as much as 40% of the development time for large Web projects. Several Web development companies at the conference confirmed that basic estimate. One developer put it this way: "We're very up-front with our clients about supporting older versions of IE. It usually takes about 30% more work to support them. Our company bills by the hour so it's the client's choice, but even then we'd prefer not to support them. It's tedious work and never seems to come out to anyone's satisfaction."

The differences can be enormous. On Internet Explorer before version 9, the CSS-related features like rounded corners, drop shadows, gradients, and transitions don't come for free just by mentioning them in a style sheet; transparency for PNG graphics doesn't work in IE6 either. Scripting features like geolocation, cross-domain AJAX, client-side storage and caching, and canvas are missing as well. Instead, the developer must include special code or images to make any of it work in a minimally-crippled fashion when the user has an older version of IE. It is nearly a completely different code path and has to be tested extensively to make sure it works within the limitations of these workarounds, which cannot do exactly what the native functionality can do.

Zakas made a short but elegant argument about whether Web sites need to look exactly the same in every browser. In essence, your site should look different in every browser, because it provides the incentive to upgrade. The trap we have fallen into is thinking that just because it is sometimes possible to make IE6 work using an expensive Herculean effort, that it is a good idea to do so. When the CEO of the company is using IE6 on their own computer, the demonstration of the latest internal Web app will look like something from 2001. When it's running in IE9 or Firefox 4, it will be good looking and responsive. Is that a bad thing? Perhaps not, as long as it works to some extent, doesn't throw errors, and reminds the CEO that their corporate priorities are what made the app look that way in the first place.

So there you have the hidden tax of staying with the older versions put into some hard numbers with some concrete justifications. It is somewhere between 30% and 40% more expensive (and time consuming) to develop for the older IE browsers that hang around on XP. This isn't a one-time tax, either. Every project a company develops will have to pay the price, and every third-party service will fall prey to it as well.

One reason why developing for older IE is so expensive comes from something that Zakas calls a faulty development model of Web pages. He says that although we often try to compare a Web site to a book, a Web page is not a printed page. Instead, a better model is that a Web site is like a television channel, and the Web browser is like a TV set. Clearly, the channel will only look as good as the TV it is being shown on.

To Zakas' analogy, IE6 is like an old black-and-white tube set whereas the newer browsers are like 3-D flat-screen sets. We are unfairly expecting developers to deliver a flat-screen 3-D experience on a black-and-white TV set. It is possible to use the "shims and polyfills" technique to coerce older browsers into running some new features, but inevitably those Web apps will end up being slower, harder to debug, and longer to develop than when the browser supports them natively.

There are some encouraging signs that users and companies are upgrading their PCs. Just this month there was a report that Windows 7 has surpassed XP in the United States. This milestone has already been reached in several other countries, particularly in Europe. Internet Explorer 6 and 7 usage in particular continues to drop along with the Windows upgrades, further strengthening the argument that older browsers should not be a focus of--or impediment to--Web-based development.

Microsoft has given XP users a difficult choice when it comes to browsers. Since IE9 doesn't work on XP, companies that want to stay with XP for a few more years can't go any further than IE8 if they want to stay with Internet Explorer. They can, however, move to Firefox or Chrome, which do support Web standards well and run on XP, so the lack of IE9 on XP is not a reason in itself to avoid the old-IE development tax. Transitioning to a new browser may not be completely painless in cost or time, but staying with an ancient browser will hold back your Web development efforts and cost money as well.