Over the past six weeks or so I ran a poll about the use of old IE versions among web developers. It’s time to publish the results.
In total the poll got nearly 18,000 replies, and the worst-answered question about 1,150. So there will be some statistically valid signal among the noise, though you shouldn’t treat subtle differences as serious clues for old IE usage. Fortunately, there are hardly any subtle differences: web developers’ old IE preferences are very clear.
IE8 is still supported by nearly all web developers. IE7 by more than half, but IE6 support is at a paltry 10%. Not long now before it disappears from our radar, like IE5.5 has.
Web developers still occasionally test in IE7, and even in IE6. If the client demands IE6 or IE7 compatibility, web developers generally charge them extra. I’m surprised by the high number of people who even charge extra for IE7, and the relatively low number that charges for IE6 compatibility. (Though many of them will be web devs who don’t test in IE6 at all any more, period.)
|Are you required to support ... in your average project?||1%||10%||56%||92%|
|Have you tested in ... even once in the past year?||2%||39%||79%||n/a|
|Do you charge extra for projects that should work in ...||n/a||66%||42%||n/a|
There are three methods to detect old IEs (except for a browser detect, which you should not use). Conditional comments are clearly the most popular, followed at a distance by the IE versioning switch. Using Quirks Mode (i.e. no doctype) is not very popular any more, even though it switches all IEs back to 5.5 and thus gives you a stable, though crappy, base to work from.
|In the past year, did you ever use conditional comments?||79%|
|In the past year, did you ever use the IE versioning switch? (<meta http-equiv="X-UA-Compatible" content="IE=EmulateIEx" />)||28%|
|In the past year, did you ever use Quirks Mode; i.e. a page without a doctype that switches IE back to 5.5?||9%|
The reason I ran this poll is to find out what kind of old IE information web developers need. It’s clear that the market for IE6 information is collapsing, even though IE7 is still a going concern.
That admirably lines up with my current test situation, where I can test in IE7 but not in IE6 or 5.5. Thus I’ve decided not to test in IE6 any more, although old test results that are already in my tables will be preserved.
Upcoming speaking gigs: