I was thinking along the lines of a serial input method and also a simple UI to feed it sanitized data for its config. maybe I'll go along in that direction.
how about this: a 2-level auto config. suppose you have (or create) an open wifi for a short time period and let the embedded system connect using that well-known ssid. as the ssid would be pre-programmed into the small system, so would an ip-addr and a path. is that asking too much?
yes, it is, in some ways, but would enough people be willing to set that little bit up in order to have a more auto-configurable small system environment?
if the device has a display of any kind, it could have a pre-programmed serial # or even generate something and then display it. a mac addr is ideal since we assume this is a wifi-config problem and you have to have a mac-addr to be a tcp/ip node, these days. that could be part of the url that you go 'fetch' from.
you fetch a file and that has your access control info in it. if you do this over a secure transport (ssl, etc) you are ok, but of course, low end systems won't have crypto ability. but for a short controlled duration, on a captive network (as captive as you can make 'open wifi, lol), it may be ok. it saves you from having to enter data ON the device, it lets you reconfigure things as much as you want and the device simply has to reboot to start the cycle all over.
if the device is remote or sealed, its not practical to require a hard wired serial connection. its a wifi device and so you can assume it has some kind of wifi and ip ability. maybe its ok to 2-stage it by booting up on a public wifi (that you provide), get credentials, disconnect the public wifi and then use the private credentials to get on the 'real' wifi network.
for $30, you can buy battery powered wifi access points. a rasp pi could also be a private 'open' access point as well as a config file server over http.
comments on this approach?