Stefan, actually I was going to argue right until the end of time but have to admit that I agree with you, like Gallen Rowell wrote:
"An equally important trait is the size of the rat...... The rat refers to the voracious creature gnawing at a person's stomach from inside that drives him or her to repeatedly leave the comforts and security of civilized life to challenge him or herself in the natural world".
Now, getting back to the whole technical topic, almost all the test I see tend to be presented on landscape oriented websites but ALL seem to ignore or leave aside several vital issues to the landscape photographer (at least for me
1). Long exposures (battery issues): I do mountain photography, that is several days (quite a few sometimes) far away from any power supply. The solar packs add too much weight considering their effectiveness. About 2 or 3 long exposures with my 5d2 in low temperature with the consequent dark frame drain a complete battery, not fun at all. I can leave my 4x5 or almost any mechanical camera ticking for hours in below 0š with no side effects.
2). Long exposures (highlights and DR): just a couple of quick and simple tests I did a while ago to calibrate my developing (all epson, I didn't drumscanned these samples):
Number 1 is a long exposure (more than 20min) made more than an hour after the sunset. The same shot on digital (at least with the 5d2) would have the sky completely blown to a thermonuclear white with absolutely no hope of recovery, I don't have the digital sample at hand but anyone who tried it, know that this is true. It can only be done on digi with some really heavy filtration, impossible to use in many cases.
Number 2 is a long exposure made in similar conditions, the same story here, I could leave my 4x5 ticking there for the whole night and walk away with a correct exposure.
3). Contrasted scenes: I'd compare the performance of digital and film in really contrasted scenes. For example a flat scan of a mountain forest scene shot during the winter (with temperature that have me swapping my digital batteries after a while outside). The sun is directly exposed with no filters. As I mentioned above, these are epson samples, I didn't drummed them yet. On the neg, the only thing blown out is the sun itself! Even the patches of snow on the PP have texture. And I have other scenes shot over glaciers in high mountain with the same result. And no flare!
4). Big prints or big upsamples of completely organic nature: Almost all the tests I see out there are usually done in some parking lot with some leaves on a wall, or cars, or other regular, monotone, man made objects. These are pretty easy to blow up, even a 6mpx DSLR can provide enough horsepower to upsample a regular, predictable surface to a huge size using the current interpolation algorithms. But landscapes are shot in the wilderness and usually are about the nature, not some license plates. I'd like to see how the digis hold up an interpolation to a 4000dpi scan size (from a 4x5 or a 8x10) of an organic, chaotic scene, full of textures, leaves, irregular patterns and completely random objects distributed all over the scene. In my tests, the digi that I own, that is a 5d2 is completely pathetic in this situations when compared to scanned film. How an MFDB would perform for example in the following situation? I shot this scene in the jungle some time ago using fine grain slide film, it has all the features I mentioned. Sorry, once again, it's a fresh sample and I have just the epson version at hand, but here is the full scene anyway in a pretty straightforward scan
And some random crops from my 11000 pixels sample scan. No PS sharpening applied, it's what it would look like on a 90cm print viewed from close distance and unsharped. And a drum scanner would add an additional truckload of info, and crisp info
I guess that a test published on a website that has the word "landscape" in its name, should at least try to walk through common landscapes situations and use real life scenes before drawing some conclusions