[humungous auction site] has a policy of no user left behind: No matter what antiquated hardware you have, no matter that you might be clinging to Navigator 4.7 (remember Gary?), no matter that you bought a WebTV box in 1998 and refuse to give it up...no matter what, you shall be able to use [humungous auction site]. It costs them tens of $millions every year to maintain a staff of codemonkeys to do nothing but test and come up with workarounds, and we often joked, rather ruefully, that it would be more cost-effective for [humungous auction site] to simply buy new computers for all those troglodytes (given the existence of $100 commodity boxes intended to seed the Third World with computers, that's not really an exaggeration). You can do that when you're [humungous auction site] with revenues in the tens of $billions every year. But what's a freelancer to do?

A freelancer has to practice triage by the numbers:

IE7 (and alas, IE6)
FireFox (and other family members like Nutscrape Nagivator 7, by which Netscape clings to a sort of half-life)
Apple Safari 3

and if time permits and the client has deep pockets, you might tackle minor niche browsers like Opera.

But there's another problem here that's more fundamental than economics: The newer browsers (defined as those that attempt to comply with the international standards) have features not supported in older browsers. There's no point setting up a lab full of oddball computer and browser combinations if the browser just can't do the job.

Much of the technology side of "Web 2.0" is based on the availability of a component in modern browsers since IE5 called "XMLHTTPRequest". It's the little guy that lets script on a "page" strike up a conversation with the server without requiring the browser to destroy the Web page and load a new one. It's present in IE6/7, FireFox, and Safari 3, and with a little work ironing out the minor inconsistencies, it's possible to make all those browsers perform cool Web2.0 stuff. Without it, nothing. Dead in the water.

The "big three" that provide that level of support account for about 97% of the browsers in use, and that just has to be good enough. It's not really a matter of asymptotic economic investment, but more of a steep sharp cutoff: The others are not in the running.