Boxboy,
I've had cable modems installed in my house twice, both times the cable company insisted on sending a tech rep to my house even though I could have done it myself, and both times the tech reps were very very happy I chose not to use the offered USB network adaptors because they get a lot of tech support calls for USB network adaptors. From a pure techy and gamers point of view, USB network adaptors are sometimes seen as *bad* because USB requires the cpu to do extra work, and by definition anything that makes the cpu do more work is bad for gaming and offends a computer geek's sense of right and wrong. On some computers, a network adaptor on a USB port can use up to 20% of the cpu. It's not that bad in modern computers, but I had a pentium 3 700mhz that would spike to 90% or more cpu if I used a usb mouse and simply wiggled the mouse around on the screen. The hardware and drivers are better now, but I still don't like having extra usb adaptors sucking any cpu time if it's not absolutely necessary.
I've also seen cases where a heavily loaded computer (high cpu usage) will start dropping network packets when using a USB network adaptor. If you're going to be heavily tasking your computer, a usb adaptor might not be the best choice.
On the other hand, a usb adaptor may be used with a USB extension cable to let you place the antenna in a better location without the signal degredation you'd get by using a pc card and regular analog antenna extension cable. Lots of wireless networking geeks are taking those tiny usb wireless adaptors and putting them at the focal point of wire mesh chinese parabolic cooking utensils (used for cooking noodles and stuff like that), turning them into directional antennas with ranges increasing up to several miles. They love these things because the entire adaptor fits at the focal point of the antenna reflector and the extension cable is the digital USB cable, not an analog antenna cable, so there is zero signal loss due to long antenna cables. That means they can even use a powered USB hub and run a rather long USB cable anywhere necessary to get line-of-sight to other antennas without worrying about expensive shielded antenna cables or losing the signal. But those guys are going for connection range and not really worried about cpu usage.
I guess it depends on what you're going to use the computer for because both have their good and bad points.
breakbreak
Llama, I don't think you're making it up, but theory simply doesn't always hold true in practice. I'm not sure if it's even possible to improperly set up my wireless router's antennas... Put router on desk, point both antennas straight up, make sure I don't hide it behind stuff. It's not rocket science. I've fiddled with it while measuring signal strength and without a high gain directional wireless antenna it's simply not possible to get a good signal from my router to my wife's office only 30 ft away due to the signal passing at an oblique angle through two 6 inch thick brick walls. No matter how skilled someone is at pointing those two antennas straight up, the signal loss is going to be there and it drops the bandwidth down to less than half of the 1.5mBps cable modem I have. Even with the high gain antenna I installed pointed right at my wife's computer desk, the signal is still degraded to where a computer on that desk can't max out my cable modem.
An 802.11G connection along the same route would easily be able to fully load the cable modem due to the higher available bandwidth for a given signal strength. I do not understand why you fail to get this simple concept, but I'm done arguing with you over it. You sound like an engineer explaining to a pilot who just bailed out of his plane because the engine failed that it's not possible for the engine to fail because these documents *right here* prove it wouldn't fail, so the pilot MUST have done something to cause it to fail. I've been in that situation a dozen times in the F-15E... The radar quits working or the displays freeze or the moving map flashes up on the hud for a second, and always the engineers and maintenance people say that it's not possible, they can't figure out why, whatever. It's the difference between tech specs and the real world, and it's why in normal everyday usage an 802.11G connection will quite often be faster than an 802.11B connection even though it's ultimately utilizing a network pipe that is slower than the THEORETICAL maximum bandwidth of the 802.11B standard. And this discussion hasn't even gotten into the areas of network overhead, channel sharing, packet framing, and all the other nitty gritty details that chip away at that max theoretical bandwidth.