Minimally, shoot blue sky wide open to check for symmetrical vignetting (lack of that means decentering).
I've read the 70-300L doesn't show vignetting at all on crop bodies, so that wouldn't help :-o
Sorry - didn't notice the 60D in your gear list. Instead, you'd need to shoot a test chart or flat surface with detailed textures, align the camera so the sensor is parallel to the chart/wall, and compare the corners to each other to see it they're equivalent (not as sharp as the center, but as sharp as each other).
See http://www.canonrumors.com/tech-articles/how-to-test-a-lens/.
Thanks for the link! However, it basically says that you cannot check for overall sharpness - only for a broken lens and for for relative sharpness in one quadrant of a lens in comparison to another. Which is better than nothing, though.
I don't think it says that at all. You certainly can test a lens for overall sharpness, and several suggestions have been offered for doing so. Personally, I use a test chart that costs more than some L-series lenses; there are free versions that can be printed and will work fine. Likewise, there's the Imatest software which Mt. Spokane linked, but if you'd prefer not to spend the money on that, or like me, are a Mac user whereas Imatest runs only under Windows, Norman Konen provides some download links to macros (free) which run in ImageJ (also free, formerly NIH Image) to perform SFR analysis.
Copy variation is inherent to the manufacturing process, as made quite evident by Roger Cicala's (lensrentals.com) testing of many copies of various 50mm lenses. On a smaller scale, Bryan at TDP tested three copies of what is currently Canon's most expensive zoom lens, the 70-200mm f/2.8L IS II. Two of the three were sharper at 200mm than at 70mm, the third was sharper at 70mm than at 200mm. Canon's theoretical MTF curves suggest the lens should be sharper at the long end. Does that mean one of Bryan's lenses was a 'bad copy'? Not if you'd want to use it more at the short end than at the long end... But it's an outlier compared to the other two.
I think the point about testing sharpness of a lens that you buy was not that you can't, but that doing so offers no menaingful way to determine if yours is a 'good copy'. No two testing setups are identical, and to compare results across different testers/setups, they would need to be the same. Even so...say, for example, that you duplicated Klaus' photozone.de test setup quite accurately, such that you could compare the MTF50 numbers you obtained to his posted data. What if your new lens was slightly better? Would you think you have a good copy, and stop there? All you'd really know is that your copy was better than the copy Klaus tested - maybe that copy was one of the worst, so your copy would only be slightly better than the worst. Conversely, what if the copy you got didn't perform as well as the photozone copy? Would you return it and buy another one? What if Klaus tested one of the best copies - how many lenses would you go through? What if the next 4 lenses you bought were worse than the first one, but you'd already sent that one back, and of course you would not be able to repurchase that specific copy...
Going back to the original question, "How can I get / test a good copy of a lens?" Basically, there's only one way to do that - establish a robust and reproducible test setup, then buy 10 copies of the lens you want, test them all, keep the best one and send the other 9 back. (Then, the return postage with insurance would probably pay for a 17-40L). Short of that extreme which would be needed for meaningful comparative data, what you can realistically do is not determine if you have a 'good copy' of a lens, but rather, determine that you don't have a defective copy of the lens. If that's the case, then you have a copy that should meet Canon's manufacturing QC standards, and while there are obvioulsy variations in quality within that range, the a difference of a few 10's of LW/PH that you can detect shooting test charts is not going to make a meaningful difference when shooting things out in the real world. Still, I've heard of people who send every lens they buy into Canon service, and some report the lenses are adjusted and come back sharper.
Having said all of that, there is a test you absolutely should perform, and one that will make a whole lot more difference than finding the sharpest copy of a lens. Test the AF performance of your new lens. Assuming you're not one of those MF-only folks, if your camera AF doesn't play nicely with your new lens, that's going to have a real and meaningful impact on your shots, since shots where focus is off are rarely sharp. For example, my 70-200mm II IS delivers very sharp results with my test chart (where testing is done with manual focus, by focus bracketing). On my 7D it requires an AF microadjustment of +2, which means it's off by 1/4 of the depth of focus wide open - in most cases, that's close enough not to notice. But the same lens on my 5DII needs an AFMA of -6 - that's 3/4 of the depth of focus, and means that without that adjustment, many shots would be noticeably back-focused. Given that you have a 60D, where you cannot make those adjustments, AF performance will be very important for you to test. Just keep in mind that if your body is at one end of the range or the other, you might find a lens to match, but if you get a new body later, the lens might be way off. For me, a backfocusing 100L Macro on my T1i/500D was a big part of my move to the 7D, and I'll never buy a body without the AFMA feature.
AFMA
might not be such a big deal for this lens - f/4.-5.6 on APS-C doesn't yield a very shallow DoF, and deeper DoF masks focus errors. But having said that, I owned a 70-300mm f/4.5-5.6
DO IS lens, and it required a -7 AFMA setting on my 7D to correct the backfocus. You can see a typical pre-AFMA shot below, focus was not on the eye but rather the plane of critical focus was running through the back legs of the frog - this is a shot at 300mm f/6.3, and the backfocus is evident.
