I think everybody is explaining the same technique, but looking at it from different ends. Some people are taking great care to get exactly the amount of exposure that will get the brightest highlights without clipping, which automatically gets you the most shadow detail (without clipping the highlights.) Other people are arguing that you should expose for shadow detail, but would also presumably not wish to clip highlights at the other end. Same thing, basically, badly phrased, perhaps. This is also assuming that nobody would want to completely blow the highlights in either film or digital, which can be done in either medium.
It's also possible to ETTL, and let the highlights go. This works in dark churches, where the blown highlights may give you a "spiritual-type" glow, while still letting you look at some of the details in the statuary, frescos, et. But you should know when you're going to let the highlights be blown...
I'm interested in the comment about the linearity of digital sensors, because it brings up the question of why somebody doesn't design a firmware tone curve for *exposure* that simulates a film curve (bring back my Kodachrome!) Shoulders are pretty damn useful sometimes.
Somebody's going to tell me it's been done, aren't they?
JC