Currently, there are two different approaches to the problem of building open-source-based NAPs. (The definitions of the terms OpenAP and "OpenNAP" tend to be a little confusing, since they are often used interchangeably to mean either a wireless access point that's "based on open-source software" or one that's "open to public use." For purposes of this column, I'm going to finesse the issue by avoiding both of them.)
One tack involves converting older 386- and 486-based computers that would otherwise merely be landfill candidates into NAPs based on Linux or on various flavors of x86 UNIX. The other is aimed at replacing vendor-proprietary firmware in commercial NAPs with specific Linux-based code.
Both approaches have merit. The former is both environmentally responsible and particularly affordable, which makes it attractive for schools and non- profits, as well as Linux hobbyists and other shut-ins. The latter lends the well-known power of open source development to notoriously bug-ridden vendor firmware, and it also adds all the capabilities of Linux to their products.
The problem with the first line of attack is that the cost savings involved aren't all that great, once you factor in the price of a PCCARD-to-PCI adapter; and the adapter is needed to add an external antenna to the AP. IBM does makes a native PCI Wi-Fi card that will let you skip one adapter, though that drives the cost up more.
...the right project may well be a truly open-source 802.11 Network Access Point. The hardware necessary to actually vet the code isn't cheap, but it should only take one or two bench units to test likely release candidates...