The support issue is, in my mind, more or less covered. There are now several major IT vendors supporting Linux, so that enterprises don't have to feel they'll be stranded out in the cold.

The cost advantage is trickier. In many ways, Linux distros emulate Windows so well that you can hardly tell the difference. And neither can the big CEO types. They know about the security benefits, but if they've already got a Windows team handling that and it's not straining their bottom line, then inertia is the route of least risk. A migration is complex and could be problematic, might require software rewrites or Windows emulation, and might require them to shake up their IT team to boot. And even then, there's no guarantee they'll be better off.

So I think the best bet is to find businesses that have been bitten hard by Windows problems, or who are being financially strained by licensing/support issues, and get them to switch then let their case study do the talking. But I think companies being affected by those issues wouldn't be the enterprise so much as small/medium-sized business. And personally, I think this is the real target to get change started.

Enterprises move slow, and a lot rides on a highly successful conversion because all the other big corporations are watching to see if this Linux thing is worth it. And IMHO massive and ambitious conversions are likely to at least have some unexpected roadbumps at this point in time. Linux on the desktop is still "new" and the existing market is mostly limited to techies and a few in-process govt/enterprise conversions. So rather than aim straight for the biggest fishes in the sea, why not target small/medium-sized businesses, moving them over a few at a time, and improving Linux as you go along by filling in gaps that appear?

As someone in that target market (small to medium sized business), I would say that the biggest hurdle for Linux is learning curve. Linux is really about "piecing together" an OS from a bunch of different parts, but most Desktop users just want to hit install and have everything they may need installed for them. (i.e. like with OS X and Windows) There's also the issue that the filesystem, for example, is very hard to understand, which becomes important beause in smaller businesses users do some degree of maintenance on their machines.

Developers, similarly, want a single, clear target where they know what they need to bundle and what will be provided by the OS. This is currently handled by complicated dependency scripts, but writing scripts that handle all software dependencies is just one more thing devs have to learn and maintain that they don't need to deal with on Mac and Windows.

IMHO, the simpler you make it, the more people will start to move over, which leads to more mindshare and even more 'switchers'. Linux has been tweaked quite a bit on the 'outside', but the insides are pretty much still a system built 20-30 years ago and optimized for client/server terminal usage. Mac OS X made the decision to regard this as a legacy filesystem, and IMHO they've benefited greatly from it.