The 802.11 standards are moving far ahead of Moore's Law. (Image from Linksys.Com.)
The first 802.11 standard ran at 1 Mbps. The second, 802.11b, ran at 11 Mbps. The third, 802.11a, ran at 54 Mbps. And now we have the 802.11n proposal, which promises service at up to 100 mbps.
While all these standards are backward-compatible, your speed will be that of the slowest component. Thus, the current moves to embed 802.11 into PC motherboards may be premature.
When people say "I Got Wi-Fi" they're not used to asking "what kind."
They usually mean 802.11b, the 11 Mbps standard using the 2.4 Ghz. unlicensed frequency range. There is a standard called 802.11g that could run data at 54 Mbps in the same frequency range, but unless everything on your standard runs g you're going to be disappointed.
When you make something part of the main system you've put a limit on it. Put something better on, as a peripheral, and you not only have redundancy, you might have real problems, radio waves running into one another with different configurations.
Thanks to low demand the 802.11 field is moving ahead very fast. Most of the home networking gateways I see on the market are 802.11g, and are rated at 54 Mbps. But if PC makers embed 802.11b in PCs, none can use that speed.
In some ways this is a nice problem to have, until customers find themselves with equipment that has limited compatibility. That could reduce demand further.
What we need, I think, is more imagination on the part of gateway producers. We need more advanced "network molding" technology -- cognitive radios and better antennas -- that will let users define more precisely the limits of their home networks and eliminate dead spots.
What we need, is a better definition of "better," one based on user experience rather than feeds and speeds. Once we get that, and have people willing to pay for that, we can justify good prices for networking equipment, and won't be as tempted to make networking a mere feature.