I have not tried the technique, and don't really have any plans to at the moment, however the technique looks interesting from the paper.
I noticed Little Big Planet used a similar method to gain transparency, although without the DSF, it seems they use stippling, but take advantage of the way the PS3 handles 2xAA to get an extra layer.

It certainly has a practical solution to the transparency issue, and if you are interested in gaining full flexibility at the cost of some overall image quality, it is certainly a viable solution given the right hardware. This seems to be a case of using the most appropriate technique for the target hardware, and of course it also needs to be based on the requirements of your game.

At the moment I do not really have any time to look into other rendering techniques, I'm trying to get a good shadowing system done, and some other core utilities for my engine. At the same time I'm dreading the migration to XNA 4.0 HiDef.

I read the paper for IR when it came out, and after reading through Matt's article on the topic, I would have to say that just like LPP, it seems to be a case of the best technique for either your game, or target hardware.

Having transparency built into the technique is quite nice, assuming you are willing to take the image quality hit that you get from a 1/4 light buffer. You could try using buffers 4x your back buffer size, but with modern resolutions this leaves you with giant surfaces, and on the XBOX that will lead to a lot of EDRAM tiling.

It may be something to look at in future, assuming nothing newer/better comes out. When it comes to transparency, at the moment I am more interested in Order Independent Transparency, although that needs more than what XNA provides at the moment.

Michael

I will have further information on reconstructing position when using a fullscreen quad (like in the sample) next week.