I'm a bit embarrassed to admit this, but I'm not even sure what to do to test a lens.
Testing for absolute sharpness is difficult unless unless you have some lenses of the same type, but you can roughly compare to the iso crops from the digital picture. Adjusting afma is easier (or more expensive if using focal pro) - but looking for decentering is really easy and that's what I did above: Shoot a test chart so that fine patterns at the edges and corners, then compare the relative sharpness of the outer areas.
This one is very good for this purpose: http://www.bealecorner.org/red/test-patterns/EIA1956.pdf
At what point are people chasing unrealistic and unnecessary expectations?
Good question, that's why I attached the crops above. But for even for €670 (which actually is also a lot of money to me, I could nearly buy another 60d for that) I expect the lens to perform at the top 1/3rd of the potential - and it's frustrating to see that one side of the lens is visibly sharper than the other.
At times, the 17-40L has been notorious for that (a few years ago many people complained about one side being worse and every single person always mentioned the same side so I wonder if there had not been some bad runs on it a few years ago). I exchanged my first one. Pretty huge difference.
Thanks, that probably explains it (my 100L and 70-300L are just fine with the first copy), I hope the next 17-40L will be better :-o because since the lens isn't the sharpest on the block I'd really at least get the max. possible lens performance. On the 100L a small difference wouldn't matter because that one is incredibly sharp anyway.