On Mon, Jan 31, 2011 at 4:15 PM, John W. Linville <linville(a)redhat.com> wrote:
For roaming "seamlessly", simply using the same ssid and
encryption
info won't matter much if the AP's aren't bridged. Most consumer
APs out of the box use a routed NAT configuration. This means that
when you switch between APs you will get a different IP address and
connection disruptions will result. To avoid that, you need to make
sure the APs are bridging between wireless and ethernet, and further
you must ensure that both APs plug into the same bridged segment
on the ethernet side. Somewhere on that network there needs to be
a single (or several coordinated) DHCP servers so that the same IP
address is equally valid on the wireless sides of either AP. In the
best cases the APs will runn IAPP or something similar to smooth the
wireless handoffs between the two APs, but in practice that is not
entirely essential.
In my case the APs are connected to a single LAN on the wired side and
the dhcp server is run on a Fedora machine such that the ip address is
tied to the MAC of the wireless NIC on any one machine - hence the ip
that I am given and maintain within the house is the same whichever AP
a wireless laptop connects to.
As for the inital AP selection...you assert that it is illogical to
select an AP you know over a stronger AP you may not know. I'm not
Not quite - in my case both APs have been registered on the laptop and
it is certainly possible to define two separate "connections" in NM
both with the same ssid and encryption plus password (in my case
WPA-Personal with AES encryption) - but with different bssid where
each corresponds to the hardware address of the AP in question. So
even if both are defined to a specific laptop in the list of "known"
connections to NM the laptop always appears to attempt to connect to
the one it was last connected to - and where that signal is weak, with
the near AP being both already known to the machine and strong the
connection icon seems to spin and is very iffy about whether it might
connect or not - but the logic would say it should now the near strong
and already known connection would be the one it should attempt to
connect to - surely? So the near strong signal is certainly not one
that is unknown to the laptop's NM list in this case.
I would be less concerned about the roaming but certainly in this case
the initial connection should go to the near and strong signal as it
is already defined in the system - but it appears this is not the
case!
sure I agree. So long as the known AP remains serviceable, there is
no
particular reason to switch to another AP simply because he registers
a stronger signal. In a perfect world you could easily determine
which AP is "better" simply by the signal strength of a beacon,
but in reality a number of factors can effect real world "better"-ness.
I would only consider the AP selection issue a bug if NM insists on
using an unreachable AP even when the other is reachable. In either
This is exactly the case - and the main source of irritation - and I
believe others have verified this behaviour on their own systems.
case, if you want to roam between the APs you should ensure that
they are both bridging the wireless connections to the same ethernet
subnet.[1]
Which is indeed the case at home - I can't vouch for whether this is
the case on the system at work but I could find out!
John
[1] There are other potential configurations for roaming, including
mobile IP and the like. If you (i.e. anyone reading) feels like
trying to explain them in an email then feel free. :-)
--
John W. Linville The truth will set you free, but first it will
Would be great to see these issues tackled....
--
mike c