Does Microsoft sell this camera as a product?
No, the camera is a research prototype developed at Microsoft Research Asia. It is not  product of Microsoft and at the moment no plans for productization exist.

What is the cost of the camera?
The camera cost is approximately US$ 25K (2009). The main components are the main sensor (10K), Lens (4K) and XYZ translation stages (3x2.5K). For comparison, the highest resolution scanning back (only) that is commercially available today costs over US$ 60K and delivers “only” 340 mega-pixels.

What is the resolution of the camera?
Resolution is the ability to separate between objects. The resolution of this camera is limited by the lens and is approximately six seconds of arc. This is ten times sharer than a 20/20 vision. This is sharp enough to see 30 separate lines on a head of a sewing pin from a distance of one meter away, or separate between two pencils that are one centimeter apart from a distance of over two hundreds meters.
    The term resolution is more commonly (but inaccurately) refers to the size of the image in pixels. The maximal image size of our camera is little over 50,000 x 33,000 pixels or 1.6 giga pixels. However only one giga pixels can be optically resolved (the over-sampling is useful to prevent aliasing and help with the Bayer demosaic).

Can we put a sharper lens? 
Strictly speaking - yes, lenses that are designed for digital sensors such as the Schneider or the Rodenstock digital lenses have significantly better resolving power than the film lens we are using.  However, the image circle created by our lens is much larger than of the currently available digital lenses.  If we multiply the resolution of the lens by the size of the image it produces, we realized that our lens obtains much more information than the sharpest commercially available digital lens can obtain. Moreover, working at lower resolution means that larger and better pixels are used. That said, if we could find a lens that has better sharpness and a large image circle, then by using computation methods (super-resolution) we can obtain real resolution of 6 giga pixels from the same camera.

Can the camera be made any smaller?
Unfortunately, not significantly. Making a giga pixel SLR for example will require very small pixels (0.9 micron) and very high resolution lens (538 lp/mm) over a wide aperture (f2.8). Such lenses are not commercially  available. Having small pixels is also problematic as can be seen in this site http://6mpixel.org/en, and in http://www.clarkvision.com/articles/does.pixel.size.matter. The sensitive area of the pixels of our camera is approximately 9x4.5 square micron, this is 50 times larger area than a 0.9 micron pixel size. This affects both the light sensitivity  (by over than 5 stops) and the dynamic range.

Can we use image mosaics instead?
Image mosaics can create very large images (tens of giga pixels). They are usually created by rotating a camera about its optical center (assumed to be, but not always is, at the location of the iris in the lens) and then stitch the images together. There are however, several key differences between our camera and an image mosaic. The most important one is that in our camera the lens doesn't move - resulting with an accurate planar perspective image with uniform resolution that cannot be obtained by a rotating camera mosaic (projection is onto a sphere, gnomic projection to a planar surface does not have uniform resolution).  The second key difference is that the quality of a mosaic, usually made by a 35MM camera,  is bounded by the quality of the underling hardware, which does not compete with the quality of large format sensors. Finally, image mosaic is an imaging technique,  that can also be used with our camera to create an even larger images (though not longer planar perspective ones).

How "good" is this camera compared to, say, an 8MP camera or a 1080p television?
”Good” can relate to many things. Obviously if you resolve 8MP at an certain object or over 1GP at the same object without increasing the noise,  you will get more details with the 1GP image. However, it is not always true that more pixels means a better image, you can read more about it in the answer to the question “Can the camera be made any smaller?” above.  The term “good”, can also refer to the quality of the sensor, the lens, and processing. With this respect, there are good cameras with 8MP, which are comparable to our sRGB images. However, if we are capturing raw format images and process them off-line, our results are better and are comparable to medium format digital backs. Finally, the application also matters, for example a 1080p television and even new digital SLR, can take videos of dynamic objects whereas our camera can (and was designed for) capture static objects only. However, when it comes to static objects, in particular large 3D  objects that require high resolution and extended depth of fields it does this task better than any other camera one can buy today.