I believe I may have found a basic bug in the dynamic routing algorithm used in LAN products, specifically in Wi-Fi devices.
This comes as a result of having four pairs of Wi-Fi links fail. Yet when examined by the supplier there was no fault to be found. Since these Access Points showing the problem are from three different manufacturers, this bug may be common, hence this page.
All-network (255) ping.
Commonly LANs have IP numbers in the range, for instance, 192.168.1.1 to 192.168.1.254, with a netmask of 255.255.255.0. The Netmask defines the first three of these dot-separated numbers as the IP number of the network and the last number is a number which is allocated to each machine on the network. My purpose here is not to explain this allocation - there are many other sites that already do that!
You will notice that I did not include 192.168.1.255 in the above list. That IP number is very special: when a ping is issued to that IP number, every machine on the network should answer.
Typically, windows machines do not answer this call. Mac OS and RISC OS machines do, as do most Routers, Printers, Wireless Access points and most other devices.
Now Wireless Access Points (and some other routing devices have clever dynamic routing algorithms that learn which IP numbers are on the wired LAN site and which are on the wireless side. This is to avoid broadcasting indiscriminately across the slow wireless link traffic to IPs that are not connected wirelessly.
So exactly how would this learning algorithm respond to a 255 ping? Clearly if all machines respond the router quite simply cannot possibly learn the response - and if it attempts to learn the 255 call - it is going to get confused.
In seating up my wireless links I used this ping, and I am fairly certain that it was this that caused the wireless link to fail. It seems likely that the algorithm got confused and was left decrementing its routing weighting, when it should have been incrementing. So some indefinable time after using this 255 ping - the network failed!
Quite clearly the 255 ping must be treated as a special case and the dynamic routing must not even be allowed so much as a sniff of its shadow!
Once the problem is realised, the proper cure will be obvious to the vendor of the chip containing the routing algorithm. The way for the user to avoid this trap is also clear - do not use the ping 192.168.1.255
Since finding this problem in early December 2006, the problem has not occurred since. but the I have carefully avoided using the dread 255 ping!
If you have had experience of this problem please contact me and let me know the circumstances, and the make/model of equipment.
In this case, I re-uploaded the GUI at both ends - that had worked after the 255 ping. Not so this time. I must try uploading the configurations that I have saved: maybe the routing is saved in the configuration memory.
So I swapped the two WAPs, north end to south and vice-versa. Success! Or at least, so I thought. The newly restored throughput lasted less than 1 hour!