(Apologies if this is an old subject; I'm new to the list, but have
been playing with wlan stuff since 1997.)
In an adhoc network, IBSS mode seems a natural choice, since
infrastructure mode is unsuitable, and IBSS mode is in theory standard
and has the possibility of acks and even RTS/CTS (for those of you
that think that's helpful :-).
So, if one wants to have some large number of nodes, and configure
them so that any that are within range will exchange packets
(broadcast/multicast as well as unicast), a plausible choice is
set every node to the same "SSID for IBSS creation", say "FOO".
set every node to the same "IBSS channel", say 3.
set every node to "create IBSS".
If the cards are all near each other, they end up with a BSSID formed
from one of the cards (the first to be set up), and all other nodes
join that IBSS. This is the canonical case for IBSS mode, and is the
vision you get from reading the standard.
I wondered: what if I set up two such nodes, far apart enough that
they can't hear each other's packets, and then bring them together?
>From reading the 802.11 spec, it seems that packets with a different
BSSID will not be received (at layer 2); this is the normal network
separation in infrastructure mode. So, will the nodes not be able to
interoperate, since they are in their own IBSS?
I tried this, with one node having an Lucent/Orinco Gold card, and the
other (both NetBSD 1.6.1) a PRISM2.5 (IBM T30 builtin). The BSSID for
the PRISM2.5 card was formed in the canonical manner (set 0x20 in
first byte, copy rest of MAC), except the middle 2 bytes were
changing, and the Lucent BSSID was exactly normal (0x20 set in first
byte, other 5 the same as MAC).
Despite having been set to "IBSS channel = 3", and showing current
channel 3 when first configured (several km away), after driving
closer to the first node (but still out of range - 200m away), it
showed channel 6. When I got within range, I found that the
stationary node (with a Lucent card) had adopted the BSSID of roving
PRISM, and moved to Channel 6.
This is quite odd; it seems like the PRISM card might have been
scanning even though it should have been 'creating'.
I repeated the experiment with the roving node having a Lucent card
(stationary still Lucent). As I brought the roving node close, it
switched its BSSID to that of the stationary one. I observed no
changing of channels.
I don't recall the standard speaking to the issue of "merging" two
IBSSs. But it seems clear to me that IBSSs with the same SSID should
join together somehow - at least because that's what I want to have
happen.
Does anyone know if this behavior is universal with all cards that
support IBSS mode? Specifically, how about IBSS mode on the ADMtek
and Atheros chipsets?
Can anyone point out anyplace in the specifications where this is
addressed?
Thanks,
Greg
--
general wireless list, a bawug thing <http://www.bawug.org/>
[un]subscribe: http://lists.bawug.org/mailman/listinfo/wireless
