Internet Explorer 6 (abbr. IE6) is the biggest thorne in a web developer's behind at current times. This legacy browser, released almost 7 years ago, is known for its multitude of offenses on security and standards compliance and still has a sizeable user base to this day. Its market share makes it impossible for us developers to ignore it still, despite how much we would want to do just that.

So if we can't ignore the IE6 audience now, can we at least have a timetable for when we could? I was just thinking just that recently, as we were disucssing half-jokingly to drop IE6 support from our start-up Octabox. Octabox is not a web site in the content delivery sense, it is fully a web application and requires much of the browser. As we are approaching our private beta release (two weeks, exciting times! ), IE6 debugging remains as usual one of the last things on the to-do list.

I will now try to predict the future: Using several statistical sources, I have plotted the usage patterns of the three major browsers (IE6, IE7 and Firefox) and extrapolated the timespan in which IE6 will become redundant.

I used three sources for my data:

• w3Counter, a hosted web-analytics service (no relation to the w3c)
• w3 Schools, a veteran repository of web development tutorials
• The Counter, a website analytics tool

As could be expected, the statistics provided by each site differ quite substantially, however since I am looking only for the trends of browser usage in theory it shouldn't have a big impact on the results.

I plotted all three browsers usage percentage against their timeline, calculated the rate in which IE6 usage is decreasing and used some basic algebra to derive the timespan for IE6 theoretical oblivion.

The graphs:

That R-squared box you see in the graphs is the correlation coefficient for the linear approximation on the usage pattern of IE6. It stands for how well does the data represent a linear pattern, with values closer to 1 mean a better fit. Luckily, for all three graphs I got pretty good matches, with data from "The Counter" bottoms out at a 0.96 match.

As a side note, I found the data from "The Counter" to be the least consitent and reliable - all of its 2008 data is smoothed out over a period of 150 days, providing much less detail in the numbers, its precision level is low (no decimal point) and if you look at the IE7 data you will see a strange bump in the beginning of 2007 not seen in other sources. Pretty strange.

Continuing with the analysis - since I got pretty good fit for my linear approximations (pretty good for this blog post anyway. I won't be submitting any of this to Scientific American), I will use a linear equation to extraploate future usage values.

Formulating:

- Usage percentage is a function a of time (t)

- We approximate usage to a linear function

- We want to know the time at which IE6 usage will drop to 1%

- Some minor formula rearrangements give us what we need

The time we would want to count from would be the present day, so we will assign the last value for each graph as usage at t=0. The slope coeffcient 'a' is the slope extracted from the graphs. This gives us for each data-set the amount of months from now that IE6 usage will drop to 1 percent:

As I've mentioned previously, the data from "The Counter" looked suspicious so it's not surprising that the results it produced are skewed from the others.

How relevant are those numbers - The approximations I performed here are very basic and contain no error treatment. It also does not account for outside factors, such as the release of new browsers and operating systems. We have however an indicator for the life expectancy of IE6 - it has somewhere between 2 to 2.5 years of major shelf life.

Don't throw away your IE6 browser just yet

(The excel file I used is provided for your discretion here.)