[DNA-BOF] Re: link "hints"

[Bernard Aboba <aboba@internaut.com>]

My understanding was that the relevance of L2 "hints" to DNA was for
determining whether a change is made in the point of attachment, and
if so, whether a change of subnet has also taken place.

This is in contrast to other discussions such as in ALIAS, where L2
information could potentially be used for a much wider range of purposes.

If this perception is correct, then it allows us to focus the conversation
somewhat.  For example, in IEEE 802.11 we have:

a. Information in the Beacon and Probe Responses.  This can provide the
host with the capabilities of potential future points of attachment.
However, it doesn't reliably provide information on the subnets.  Since
VLAN assignment can be dynamic it might even be argued that such
information might not be appropriate in the Beacon/Probe Response (unless
the SSID to VLAN mapping is static).

In some Virtual AP implementations the same BSSID is used with multiple
SSIDs, so a host attempting to determine whether it is reattaching to the
same "point of attachment" probably needs to use the combination of the
BSSID and SSID, rather than just the BSSID alone.

In future (IEEE 802.11k), the Beacon/Probe Response might provide
information on neighbor capabilities as well as capabilities of the
advertising AP.

Note that in some implementations the BSSID (WM MAC Address) is the same
as the DS MAC address; in others the DS MAC address is distinct.

b. Association/Reassociation.  IEEE 802.11i has two association
exchanges (an insecure one using 802.11 management frames, and a secure
one using the 4-way handshake).  In general, where both are available, the
completion of "secure association" is used as the "hint" that kicks off
DNA.  Today the "secure association" provides the SSID, but in future
it might provide prefix/subnet information as well.

c. Radio measurements.  This can include a wide spectrum of things under
consideration in IEEE 802.11k, including measurements of signal strength,
S/N ratio, packet loss, jitter, etc.  While many of these things might be
relevant to roaming or load balancing decisions, they don't appear
relevant to DNA in that they don't provide "hints" relating to the network
topology of a potential point of attachment.