P: Ability to apply masks to individual Smart Filters

Yes, definitely the perspective transform is a great addition!

I came across this link which shows the differences in smart objects between CS2 and CS4 (but not CS5).

CS2 to CS4 Smart Objects

One of the things that really disappointed me with CS5 was the lack of developments on smart objects - and I very much hope Adobe rectify this in CS6 (which would be the only reason I would have to buy CS6, unless there are other major improvements in HDR, for example). However it may be that there are improvements in SOs between CS4 and CS5 that I am not aware of?? Perhaps performance improvements, for example? Anything in that area would be great!

I expect the differences in your relative file sizes compared to mine are due to a different kind of image - I've tried it with a file, both RAW and DNG, and in my case the SO converted to SO is still smaller in both cases than the duplicated SO.

»It would seem to me that mask linking should have been there all along - this is more in the nature of a bug-fix surely. As for transforms - scaling, warping etc - they were there in CS2.«
Not perspectival transformation, though.
That was a particular problem if one passed a file with perspectively transformed SO to someone using a lower version, because the transformation would get lost on updating the SO.

»I suspect the reason is that people aren't using them much because they are heavy on computing«
I suspect that many people are not using them because they may not not have taken the time to read the »What’s new«-section in the Help for many versions ... as you say, SOs are powerful, so the willful ignorance of some users baffles me.

»of course you can use a blurred copy of the smart object with a mask, but this adds a lot to the file size (for example, flattened file 280MB, file with one ACR SO 310MB, file with duplicated ACR SO 475MB ... no filters, adjustment layers, masks).«
I tested with a dng and the psd with the SO is 246MB, the one with two instances of the SO 344.7MB, the one with the SO converted to a SO 470.9MB.
But you are right that with certain masked Filter combination the nesting would seem best.

My main request was of course to have masks with each smart filter - not to complain about the enhancements to Photoshop.

However, since we are on the subject:
- still can't use the Lens Blur filter
- auto-align layers and auto-blend layers still cannot be used which means that panoramas, for example, cannot be done on ACR objects, nor can focus-stacking be done on ACR objects (or any kind of object for that matter). However these can be done manually so there can't be any technical reason for not implement them.
- can't use liquify and vanishing point (if puppet-warp can work as a smart filter then presumably liquify and v-p could also).
- can't do content-aware scaling (I guess this makes sense as CAS needs to add/remove pixels)
- can't do HDR on smart objects (not even HDR Toning!).
- there are quite a few other things that would be very useful in the Lightroom/ACR integration of Smart Objects (a simple one would be the ability to use the original RAW file in the SO and not the (automatically converted by Photoshop) DNG file).

It would seem to me that mask linking should have been there all along - this is more in the nature of a bug-fix surely. As for transforms - scaling, warping etc - they were there in CS2.

So even though smart objects have some of the new features added in new releases (for example Lens Correction), which is nice even if it is to be expected, the smart objects aren't much smarter than they were 3 releases back. I suspect the reason is that people aren't using them much because they are heavy on computing (I had trouble using them on 1Ds files and couldn't use them on 1DsIII files until I upgraded my PC). Perhaps this is one of the reasons why there are so few wish-list items for smart objects? Perhaps another is because people don't realise how powerful they are so they don't bother using them? Maybe a third reason is that Adobe aren't putting in the effort to make them really smart (which isn't very smart to my way of thinking!).

BTW Christoph - yes, of course you can use a blurred copy of the smart object with a mask, but this adds a lot to the file size (for example, flattened file 280MB, file with one ACR SO 310MB, file with duplicated ACR SO 475MB ... no filters, adjustment layers, masks). Also blurring is simple because it's unlikely you would want to go back to the sharpened SO - but what if you want to apply some develop sharpening with a mask then apply some creative sharpening with another mask, say? Doing this with a copy of the SO doesn't work ... you would have to rasterize it before applying the creative sharpening - better to nest the SO (which is also better file-size-wise: 385MB).

In case you mean a Gaussian (or other) Blur – would using a blurred instance with a Layer Mask instead of blurring the partially sharpened SO not also work?

Basically I support the request for individual masking for Filter applications on SOs, but your statement »you seem to have done an initial cut and now you are doing nothing to make it into a more useful and mature feature« seems not wholly appropriate when one considers the improvements SOs do have undergone since CS2.

Edit: To elaborate on the improvements: perspectival transformation, mask linking, ...