Does that 20% come from your own stats or a third party? Most of the third parties look at worldwide market share. IE8 is disproportionately represented in third-world countries which you may or may not care about in your business.
Best I can tell IE8 is somewhere around 6% in the US (these numbers are rough to find because most of the places that collect statistically significant numbers charge for the info). On my site (which centers around a first-world disposable-income-heavy hobby) IE8 is basically non-existent. Your mileage may vary.
On my largest site with a couple million visitors a month, IE makes up about 16% of site visitors. About 25% of those are using IE8. So, about 4% overall. I specifically ensure that the site still works in IE8 and up and was still quite functional on IE7 when I last looked.
Anyone not ensuring their site works on IE is leaving money and customers on the table. Note that it doesn't have to look exactly the same, but it should still work and allow customers to learn, buy, sign in, subscribe, etc.
It depends on your market. I've worked on apps that have never even seen a visitor using internet explorer at all. This happens all the time in b2c software. So that analogy of leaving money on the table is not very realistic.
Of course, it varies. But there are far more users of desktop IE than desktop Safari in the US, for instance. So, once your product spreads beyond early adopters, iPhone users and Mac users (all of which are the minority of the market by large margins in each case) it's important that your site already works with the new visitors as word spreads.
It's not just early adopters vs late adopters. And nobody is saying to disregard all of IE, only very old versions.
We're talking about dropping support for a browser that is mostly used on a thirteen year old operating system that even Microsoft has dropped support for.
My site is unlikely to ever have a significant number of IE8 users. It revolves around a nerdy hobby that people easily spend $500/year on, often in big chunks. They aren't letting their tech get that out of date.
I fully admit that this isn't every market, but you absolutely CAN make decisions about your specific market and decide that IE8 just simply is not and never will be worth it.
On the other hand, there are certain markets that will probably have to worry about IE8 for ten more years.
Gmail is part of Google Apps. That's pretty mainstream. Personally, where I work, anyone using any version of IE is screwed. We don't test for IE. I haven't checked to make things worked in IE as a regular part of development since 2009 and I have worked at a pretty big company that is often talked about on HN.
A few years back I was working on an app with some moderately fancy CSS and JS. We desperately wanted to drop IE6 support but after we pulled the revenue numbers per browser, we worked out our IE6 users could pay for a junior-mid range developer to just sit there and fix IE bugs all day.
I wonder what that's doing to the maintainability of your code base. I'm sure the overall cost is higher than just the pay of a single developer ruining your architecture. Supporting IE6 is more than just about fixing bugs.
Also, something to possibly consider: reach out to customers and ask them how much they need IE6 support. Unless it's healthcare or government, I'm not sure why anyone would use IE6 by choice. Would deprecating support for IE6 lose every single one of those customers?
A big chunk of the old IE users were military people deployed in the Middle East. I'm of the understanding that they have/had some sort of Internet Cafe setup on base, and I guess they were running older browsers.
And what happens when the amount of your users drops slightly or that hired dev is now worth more money? And what about the other costs on complexity in your codebase? I think the revenue needs to far eclipse one dev's salary to justify keeping IE6 support.
I believe (but cannot prove) that the 20% is mostly due to China. Because it's such an incredibly populous nation, it can dominate international statistics like these.
Unfortunately my home nation (though very small) is also on something like 20% WinXP (thus limited to IE8). It's ridiculous, but as long as clients are willing to pay us to support IE8, then that's what we'll do, much as we hate it. In my country it hopefully won't be long before it becomes prohobitively expensive to support IE8, so there's that... :)
How long should the mountain wait for Muhammad, though?
In the UK, analogue terrestrial television was eventually turned off, and replaced by digital, following a lengthy publicity campaign.
I would guess there are very few people staring at blank TV screens and wondering what to do now. My point is, you need to give people a push sometimes.
Anti-virus companies use an IE8 user agent string sometimes for finding dangerous sites.
Other bots and crawlers still use the IE8 string, although it seems like many are switching to Chrome.
I've never been able to find a resource about this, but is there any data available for the breakdown of bots/crawlers and their effect on browser usage statistics?
One reason I ask is that using the web with IE8 is pretty miserable and a lot of sites are flat out broken. I know there is an aspect of corporate networks on it, but 20% just seems really high if you realize that 20% or more of the web is broken for IE8 users.
Second, IE8 is still very popular browser with 20% market share. You don't want to lock out 1 in every 5 of your users.
Third, a major point of using frameworks is not having to worry about browsers. This is why making frameworks are hard.
Fourth, people who are stuck with IE8 or below are typically not by their choice.