The prevailing response to the request for an integrated compatibility layer in Steam for Linux to allow playing Windows-only games has been "if they add a compatibility layer, there'll be no reason for publishers to make Linux versions of their games!". Someone linked a couple of tweets by John Carmack, where he was told first that a game ran well in Wine, then responded to a request for a Linux port with "no plans for a Linux port, it runs okay in Wine". The person who did that was trying to imply an unwritten "because", whereas I saw only a "but".
What makes anyone think that a game not running satisfactorily in Wine will make the developer any more likely to devote resources to a Linux port, any more than it would motivate them to contribute to the Wine project so as to get their game working properly in it? The only games which have had Linux ports announced are ones where a Linux port adds minimal effort to the developer's existing workflow - games using Unity3D, for example.
My counterpoint is that a substantial proportion of the Steam library does not have a Linux port and will never have a Linux port. To underline my point, I'm going to list the games I personally have bought or had bought for me since I signed up for Steam which do not currently have a functioning Linux port, and then furnish you with some statistics. (Bear in mind: by the standards of Steam users, this list is obscenely short.)
How many of these games already have Linux ports which haven't been added to Steam? One - Jamestown. (I have no idea why the Linux version isn't on Steam - I e-mailled the developer about it some time ago when I was first accepted to the Steam for Linux beta program. Their lack of any response whatsoever is poor customer service, and I am not in much of a mood to purchase from them again.)
How many of these games have a non-Windows port at all, excluding Jamestown? Three - GTA III, GTA: San Andreas and GTA: Vice City. (For reasons unknown, the Mac versions of these games appeared separately to the versions I already installed.)
Now, of the remaining 23 games, how many do you think are likely to grow a Linux port at some point in the future? My guess: four, possibly five. Valve are one of the few companies committed to Linux releases, so it's likely that Half-Life 2, Half-Life 3: Episodes One and Two and Portal, being Source games, will all make the crossing; Garry's Mod might come along for the ride if whoever is behind it cares to make it happen.
So that leaves 18-19 games with no hope of ever becoming Linux-native. Running Steam in Wine is cumbersome - every time I want to play a Windows game I'd have to switch one client to offline mode, switch the other one online, play the game, then switch it back again afterwards. Have you ever switched Steam to offline mode? It takes a while - it has to completely exit and then relaunch itself. Given how long it takes my computer to boot (it's no Usain Bolt, since I don't have an SSD, but it's pretty fast) I feel it'd be faster to dualboot.
The compatibility layer absolutely shouldn't be provided for any new games - that would be terrible for precisely the reason given by the detractors! My point is that 90% or more of Steam's library prior to the announcement of Steam for Linux is never going to make the shift, so a compatibility layer for those games is appropriate so that customers can play the games they've paid for already with a minimum of hassle.
* I know these are "supposed" to be called Half-Life 2: Episode Foo; I'll talk about Half-Life 3 some other time in the blue section.