How 802.11b clients handle roaming appears to differ from vendor to vender. Here are some brief notes from the MailingList as we sort it all out. Please add what information you may know. -- AdamShand
JimBinkley writes ...
Lucent cards behave acc. to the Cisco spec ... although I have seen very quixotic behavior. Traditionally that sort of handoff is a form of "starvation" I suppose, although "going down with the ship" might be a better name for it :->. As a hack, reassocation seems to occur (call it "starting over") if you hit the lucent cards over the head say by doing something like:
# iwconfig eth0 essid ""
MichaelCodanti writes ...
- This is from the Agere readme in their Spring 2001 software release for their APs:
Load balancing is supported; this feature is permanently enabled. This feature attempts to balance the load over the available overlapping cells. The Access Point maintains measurement of the load in its cell and provides that to the Stations. Stations use this information in their roaming decisions. The better the comms quality to potential Access Points, the bigger the impact of the load is. This feature functions only if the Access Points support it.
JimThompson writes ...
- Its all up to the Client (STA). I know that the Cisco NIC firmware used to hold onto an AP association even if it meant dropping the modulation rate (from 11Mbps to 5.5Mbps to sometimes 2Mbps). I haven't tested same lately.
- The Cisco cards still do this if you aren't using Cisco AP's. Cisco lets the AP make the decision on when to move the node based on signal strength and load. It's quite nice if you are in a Cisco only environment because you can be in a room with 2 AP's and a large number of people and the AP's will load balance themselves.. (Provided they are setup that way).