There was a news story doing the rounds recently about some French researchers finding a "flaw" in the IEEE 802.11 wireless networking standards.
According to academics at the Institut d'Informatique et Mathematiques Appliquees in Grenoble, there is a drawback to the Carrier Sensing Multiple Access with Collision Avoidance (CSMA/CA) protocol that governs when a node can access the communications channel. That drawback means that in a shared 802.11b wireless LAN environment, a user with a slow connection to an access point (AP) can also slow other users' connection speeds down to a crawl.
Now anyone who knows a little about 802.11 wireless networks ought to be aware that they share bandwidth among all devices attached to the same access point.
This is an inherent property of networks that allow all nodes to fire off data packets whenever they think the coast is clear - alluded to in the name Carrier Sensing Multiple Access. The familiar Ethernet standard also does this, but on a wired LAN, two devices that try to send at the same time can quickly spot this problem - known as collision detection - and take appropriate steps.
With 802.11b wireless technology, it isn't possible to detect collisions in this way, so the protocol tries to avoid them, hence Collision Avoidance. It does this by using short handshaking messages to check that the channel is clear, and that the target device is listening. After sending the actual message, the device then waits for an acknowledgement from the target. If it doesn't get a signed delivery note, it sends the packet again.
This strategy minimises the number of lost data packets, but on the downside it takes up vital bandwidth to do all this handshaking, which is just one of the reasons why 802.11b seems slow even compared with the old 10Mbit/s Ethernet.
The CSMA/CA protocol's benefit in a wireless environment is better resistance to interference and the other horrible things that can happen to radio waves. Throughput may drop as the signal gets worse, but there's more chance of the message actually getting through.
But it should be obvious from all this that if there's one device on the network that's having problems sending packets, other devices hooked up to the same access point will be held back by the lame duck.
Now, it didn't take me very long to work all this out once I had possession of all the facts. I didn't need a research grant, and I'm not an academic. So should I be running the IEEE? Of course not. I'm certain the IEEE was very well aware of the implications of using this protocol before it ratified the standard. The benefits of robustness and lower cost will have been weighed against the bandwidth-sharing problems, before a decision was taken. From a user perspective, the best solution is to ensure a decent signal strength over as much of the required wireless coverage area as possible and ferret out any "shadow" areas.
On a final note, it's interesting to recall that the failed European competitor to 802.11b, HiperLAN/2, was intended to use a Time Division Multiple Access protocol that gets around the collision problem by allocating a specific time slot to each device. C'est la vie.
