The DA 10-17 though, being a fisheye, will produce a much _wider_ angle of view than a 10mm rectilinear lens (e.g. the Sigma 10-20)

It'll be 15mm-equivalent: the crop factor on Pentax is 1.5x (e.g. 10mm -> 15mm) - 1.6x is Canon

The larger the front element, the less likely there will be vignetting... or some other such undesirable. I'm not an expert on such things, but some review site explained it all in their review of the (77mm front element) Sigma 50mm f1.4

The 17-50 is new, so may not be listed as available yet:

http://www.dpreview.com/news/1002/10022014sigma17mm50mm.asp

It's because of quickshift - none of the QS lenses have their focus rings turn during AF, so what you said is also correct

Basically (and I'm cribbing from that other thread here!) with QS lenses the focus ring is not actually directly connected to the lens elements. Instead it turns them via a clutch as intermediary (the same reason AF doesn't turn the focus ring - the AF motor has no physical connection to the focus ring). When the lens reaches one of the focus limits the clutch disconnects and so the focus ring can keep turning (whereas it wouldn't, for example, on an FA or DA-L lens if I understand this right).

Also bear in mind the Tamron 17-50mm f2.8 (about EUR 350), which is supposedly (I don't own either lens) competitive in wideopen sharpness to the Pentax 16-50.

Hehe I continually have this issue - my interest in photographing plant life is coupled with a complete inability to name anything other than a daffodil

The Canikon-mount ones have HSM (and quick shift), while the Pentax-mount one doesn't which is probably the reason for a lower value-for-money score (the Pentax version costs the same despite having fewer features).

As regards the original topic, I think the one failing of Photozone is the tendency to give a final (low) score to a lens the reviewer knows to be faulty (ie non-typical). Compounding this is the fact that the fault is not mentioned alongside the score. The more honest options would either be to have the lens serviced/replaced, or to state that the scoring process was compromised by a fault.

The most noticeable is actually the bottom-right corner, and it's fairly rounded which is what made me think it was the edge of the filter holder rather than the effect of the filter itself.

Not to be a geek-evangelist, but for the sake of preserving a £2500 piece of kit have you tried a Linux dual boot? Such drivers as there are on Linux are often created by independent developers (rather than the hardware manufacturer) and hence can (theoretically) be made available for 64-bit independent of the manufacturer's desire to 'encourage' upgrades.

EDIT: This thread on the Ubuntu forums strongly suggests that the Coolscan IV does indeed work.

That's a very nice shot, but there seems to be some severe vignetting in the extreme corners: I assume this is caused by the filter holder rather than the lens itself?

Returning to the OT, I only have the Sigma and hence can't give an objective comparison. However the Sigma has a reputation as being the sharper lens, and I love mine to bits.