There's been some talk about whether this or that lens is "fast" or not since the DP1 came out sporting an ƒ/4 fixed lens. How "fast" is the Olympus E-P1 lens, really?
The concept of "lens speed" (referring to maximum aperture) has been variable over the years. Without launching into a history lesson, it's all been relative—Tessar-type designs are limited to fairly small apertures, for example, and first reaching ƒ/2.8 with a Tessar-type was an achievement. Generally speaking, though, lens speed was a prestige camera feature in the middle third of the 20th century, more or less, and cameramakers competed to see which could out-do the next. The market rewarded this tendency until ƒ/1.4 became common in normal lenses (the further from "normal" a focal-length is, generally, the harder it becomes to make the lens a fast one). Several cameramakers reached further than that, most notably with Walter Mandler's famous Leitz Noctilux, and several makers, Canon in particular, reached ƒ/1.4 with some fairly wide lenses for 35mm. The other famous fast lens early on was the 85mm ƒ/1.4 Zeiss Planar, which has long and interesting history that continues right down to the present day, in a fine lens released very recently. But the market as a whole didn't much care—ƒ/1.4 at the normal focal length turned out to be the point of sufficiency for most photographers where lens speed was concerned. Options that do better than that have sporadically been available, but have never achieved much in the way of popularity.
Of course, film speed has to be factored into that. During the only two stretches of time that I used Leica rangefinders—the first time for a period of about two years in '91–'93, the second time for something closer to a single year around '96 or thereabouts—I used Summicrons. (Leica, unique among cameramakers, names its lenses according to the lenses' maximum aperture; all the "Summicrons" are ƒ/2 lenses.) I was shooting Tri-X, rating it at E.I. 200. Also—somewhat oddly in retrospect—I never used either of those Summicrons at ƒ/2! The first main lens I used was a fourth-version (now called the "pre-ASPH") 35mm Summicron (which I once rather infamously dubbed "the King of Bokeh"), and the second time I used a custom-coated collapsible 7-element 50mm Summicron from the early 1950s. Neither lens was much good wide open, so I avoided the open aperture (and, for the most part, one stop down as well, if I could).
One might well ask, why use a Summicron if you're not going to use the lens wide open? It's a good question, especially since, on a Leica, you don't need to use the lens to see in the viewfinder. That, of course, was the other driver of lens speed: SLRs caught fire in the market starting in the 1960s, and, when you look through an SLR, you're looking through the lens you have mounted on the camera, generally at its wide-open aperture. The slower the lens, the dimmer the view. And why is that important? For focusing, of course. Photographers who have come of age since 1985 or so are probably aware that everybody used to focus manually, but only if you experienced being dependent on it can you really appreciate how important it was to have a good, snappy groundglass and a crisply fast lens to see through. On SLRs, you might never shoot at ƒ/1.4 with an ƒ/1.4 lens, but that maximum aperture was still very important nonetheless.
Zander at age 1, a much-loved family snap taken with a very fine fast lens, the OM Zuiko 100mm ƒ/2, wide open (can you see the telltale of slight spherical aberration?). This was a true candid; my uncle had just plopped his own hat on Z, and Zander responded by looking up at him and smiling. I got off one shot before my uncle took his hat back.
The increasing popularity of zoom lenses in the 1980s and '90s changed the equation again. Very broadly speaking, zooms are two stops slower than equivalent primes; if a premium normal prime lens is ƒ/1.4, a premium normal zoom is ƒ/2.8; cheaper, smaller lenses in each case might be one or two stops slower than that. This was really only practicable because of the parallel advent of autofocus. Slow zooms, especially mated to cheap finders, would have been very difficult to focus manually.
Nowadays, of course, another factor has once again changed the meaning of lens speed is the high-ISO capability of digital cameras. I distinctly remember the brief juncture at the beginning of the 2000s when digital image quality became good enough to tempt serious photographers, but the "speed" (we call it ISO, even though it strictly isn't) of sensors didn't yet match that of film. That changed quickly. Image quality still declines with increasing ISOs, but the general consensus is that every sensor/firmware package has a fairly well defined limit of "usable" image quality—and, with larger sensors, the number is usually amazingly high when you compare it to similar limits applied to film. My personal standard is that if a DSLR has good quality at ISO 800 and usable quality at ISO 1600, I'm happy. My current DSLR does considerably better even than that.
Image stabilization yields yet another twist to the story. IS is a "tripod replacement," not a "film-speed replacement," and that does have implications for how it's useful and how it's used. Even so, one or two stops' worth of further flexibility in low light or with long teles is often quite valuable.
The long and the short of it is this: in 1992 I was shooting with an M6 and Tri-X rated at E.I. 200, using an ƒ/2 lens that had poor image quality wide open, meaning that I avoided shooting wide open. The new Olympus E-P1 that we've been discussing for several days here has an ƒ/2.8 normal lens. But it probably also has "good as Tri-X" image quality at ISOs up to 800 or so, and its built-in image stabilization probably yields one or two stops' worth of shutter speed into "tripod territory." Let's say for the sake of argument that the newer lens design is also acceptable in terms of image quality wide open (even though that still remains to be seen). What that means is that, even according to these fairly conservative calculations, that I gain two stops in exposure index ("film speed") (ISO 200 to 800) and up to two stops in shutter speed (thanks to IS) using the two lenses at the same effective widest aperture (ƒ/2.8). And you're not dependent on the lens for the brightness of a TTL (through-the-lens) optical viewfinding image.
Or, put it this way: given the presence of all these conditions, ƒ/2.8 is the new ƒ/2.
You can say "more speed is never bad," but in practice, the market will eventually decide on a point of sufficiency, just like it does with all technical parameters in which various makers compete to satisfy real demand. The "point of sufficiency" is where most people decline to pay for further improvement (or buy it only for status). We might already have reached that point with megapixels and sensor speed, with 24 and usable 6400 respectively in the most advanced devices. (Only time will tell. The manufacturers will overshoot the point of sufficiency, but as they do, fewer people will pay for the increased performance.) However, for a small personal snapshot camera designed to be handy and portable—given that it has a large sensor with good high-ISO performance and IS—an ƒ/2.8 lens seems fast to me.
P.S. Comment: "But I'm not so concerned about exposure, I'm concerned about achieving limited depth of field to isolate subjects."
Not picking on the person who wrote these particular words (or any individual who might have written similar words) personally...but this is total internet hooey. It's something the "thousand monkeys in a room" on the internet have convinced themselves is an issue. It isn't. Or it shouldn't be, to any pictorial photographer shooting pictorial subjects.
In this particular case: a) you don't use a 35mm-e on anything to "isolate subjects"; and b) the difference between ƒ/2 and ƒ/2.8 for "isolating subjects" is very slight, marginal at best; and c) isolating subjects has much more to do with how you use the lens than with a one-stop difference in aperture on that lens. Finally, d) most people who consider themselves "concerned" with "isolating subjects" just as often don't have enough depth of field as too much. I see the evidence all the time, in pictures, all over the place, everywhere. Here's just one (a sample from the E-P1 and 17mm lens). Just one of millions...do you see that this photograph doesn't have enough depth of field for the subject? (The only reason I can pick on this picture is that it's a sample of the new lens wide open, and is not presented as something that's aesthetically intentional. I try not to pick on actual pictorial pictures by actual amateur photographers who might be hurt or offended by having themselves used as an example of what not to do.) O, if I could but count the number of out-of-focus dog's noses I've seen because the earnest but hapless photographer was trying to blur the background....
The problem in photography from 1840 until about 2005 with rare exceptions (fashion photographers and their once-fashionable 300mm ƒ/2.8s come to mind) was to get enough depth of field. The problem was almost never too much. And then came the theoreticians of the shared armchair and their picayune calculations of sensor size and diffraction and all else that's much talked of in numberless fora but wrong....
Best advice: unless you're doing technical work, and have extremely strict conditions and/or requirements to deal with, and you thoroughly understand the issues from a practical as well as a (real) technical standpoint, then ignore everything you come across on this subject. Most of what's out there is 99.75% complete chrome-plated bulldung. (And as you can see, the persistence of the error—and the knowledge that it's never, ever going to go away—makes Mike a mite testy.) —Mike