eap Mailing List: RE: Re: comments on draft-groeting-eap-netselection-results -00.txt

[Bernard Aboba (aboba]

> even automatic processing might be complicated if you have to
> carry a 10mb file of <SSID + BSSID> identifiers and their services with you.
> this also requires that you register your <SSID + BSSID> identifier pair
> somewhere.
>
> what do you think?

The SSID is a non-unique identifier.  This will affect all schemes
that attempt to use the SSID as an identifier of a network configuration.
It does not matter whether the schemes are dynamic or static.

In particular, there are SSIDs that ship by default on APs.  For those
"default" SSIDs, the SSID isn't just a non-unique identifier with *some*
potential for duplication;  duplication is the intent, making the SSID
meaningless for network identification.  One potential mechanism for
dis-ambiguating "default" SSIDs is to use the BSSID ot distinguish them.
However, the implicit assumption here is that "default" SSIDs are not used
in large networks, but rather in situations where only a small number
(usually one) AP is deployed.  Thus the SSID + BSSID combination may
uniquely identify a single AP network.

If this assumption does not hold, a host of problems arise.  But these
problems will also afflict dynamic as well as static techniques that rely
on the SSID as a means of network identification.

The solution to this problem is probably to utilize another mechanism with
guaranteed uniqueness to identify WLAN networks.  However, given that the
problem is fundamental to 802.11, it seems likely that 802.11 will wish to
become involved in the solution.  The recent "straw poll" indicating a
desire to standardize Network Selection within 802.11 is a likely
indication of this.