The sigma is definitely one tick better
than the 10-22
. As for the 17-40/4L, I was always a bit disappointed by it, especially for an L lens. The numbers don't lie
, either - edge performance is pretty poor at wide angle, where the edges are more important than on, say a fast portrait lens.
I guess there's a reason the 17-40 is priced relatively cheap for a L lens. It's more for APS-C users who want a better normal zoom
than the kit lens than it is for a pro shooter on FF.
Not really, especially not since they came out with the 17-55 IS. Which point did you really find fault with in the comparison between the Canon 10-22 and Sigma? MTF is better for the Sigma in the middle, but similar or worse at the edges. Both are quite high and well enough. Canon has lower vignetting, but looses to the Sigma in CA measurments, but is still low for an UW lense.
The 17-40 isnt not that bad except in the extreme corners when used at 135-format sensor, but then, most ultra wide angle lenses have a really hard time there. Considering what it has, and is less then half the price of the 16-35/2.8 II, it clearly has its uses at its pricepoint. Beyond weather sealing, both boque and color-balance is things that goes into being an L-lense. But yeah, the distorsion could be better, but nowdays, most photo editors can use a lense-profile to correct for that easily enough. Same goes for vignetting, although I usually add to the vignetting manually in the post-processing.
All of the above lenses perform more then good enough to do the job thats needed, the question is just the pricepoint and what other features you need. Even though tests are good, they arent always a good indicator if it works in practice. A good example is a lense that may be very sharp but at a few aperture settings, its quite worthless because the boque it produces make me want to put out my eye or something.