The sRAW mode looks interesting, but I'm skeptical. The fact that it lessens the amount of pixels by a factor of 4 seems bogus. CCD/CMOS Digital cameras (excepting the Fovenon sensors) use Bayer interpolation on groups of 4 pixels: 2 green for color AND luminance info, 1 red, and 1 blue per cluster, because each pixel has a filter only allowing one specific range through. Bayer interpolation is used to reconstruct a full-color pixel for each, even though only one wavelength range was detected there.
Cutting the pixels by 75% is a terrible idea, and is in no way equivalent to full size RAW. For heaven's sake, that means the nice new 10.1 Mpxl sensor just became a 2.5 Mpxl sensor! I was sure Engadget misread it, but I have read it myself now - several times - and have no idea what Canon is playing at.
Perhaps somebody can enlighten me. Nikon has compressed RAW which reduces the filesize by about half - losslessly - so I just don't get it.