My concern for streaming is latency.. WiFi latency as it is pretty good, point to point. Wireless over Cat6 would be no issue. In theory user controls should work fine, a 1-2ms lag isn't THAT bad. But when it gets to 5ms or greater, it's a concern, especially if that is not the absolute max but more like the average.
But I'm primarily concerned with frame-grabbing and streaming over wifi. Obviously this would be done at something below 720p@60hz, and then interpolated to something higher using some instruction-intensive combination of lossy and lossless compression schemes. I just wonder if that can all be done in under 16ms. I mean, taking compression out of the mix and then aggregating to some color palette between 16 and 24-bits in depth would require somewhere around 1Gbps throughput. So we need substantial compression, and that will certainly (already does) create considerable loss in quality.
At present, I think even establishing a 150mbps wireless connection (guaranteed operating performance) is iffy, especially in a densely populated area. So we'd need at least 10:1 compression on a 20-bit color depth. I think you can immediately get 2:1 by dropping to 30fps and interpolating between frames. You can then probably get to 4:1 by then dropping progressive scan and going with interlacing (NTSC style) and then providing some baseline to interpolate back to a progressive frame for display. I think you could then probably get to 8:1 by using even basic RLE on small blocks (MPEG-style), so you're only sending the changes that have occurred -- but this is tricky because you can't depend on it. A game like Halo might have 98% of the screen pixels move frame to frame, so there would need to be a codec that understood motion and could compensate for that, rather than handling the picture in microblocks of pixels.
This turned into a technical paper... sorry..
Anyway, there are concerns that I have of the streaming technology. If they could instead run the games locally, and then have the rendering done remotely I think it would go a long way. Then if they could use the GPU to compress the data first, rather than rendering it out and then compressing it, that would be a tremendous help too. The games shouldn't be "streaming" from the PC to the steambox, they should be cooperatively working together to perform the intensive work PC-side and the non-intensive work STB-side.