Hmm. I friend of mine claimed that requesting trade with another player in World of Warcraft crashed his router. When he needed to trade, he always insisted that the other player should initiate the transaction.
I was tempted to discard this as pure imagination, but this was a smart and knowledgeable fellow who worked as a programmer, so I'm assuming he had done some investigation before he came to this conclusion.
I know too little about networking equipment, but I can see how certain byte sequences (timed right?) would have some magic meaning. Given enough traffic you're likely to end up sending just such a sequence eventually. Perhaps certain versions of the WoW client spat out just such a magic sequence for the particular router he happened to use?
If you were using a dial-up modem and sending raw data -- extremely common -- then sending +++ATH0 would cause most modems to hang up.
If you could get that sent over to someone else, their session would be abruptly terminated.
(Hayes patented requiring a no-data-sent time between the +++ and the ATH0. Avoiding the patent but being otherwise compatible introduced the vulnerability. In-band signalling is usually bad.)
This won't work with the line coding of modern network protocols. The key part of this story is T1 using the older AMI coding that is susceptible to loss of sync from the right data pattern.
Once at work I tried to upload a file to our fileserver to share it with some coworkers and it failed. Other files uploaded fine. The file wasn't very large, didn't have any funny characters in the name, or anything that would conflict with a reserved name, or anything like that.
After some experimenting I was able to figure out that there was a particular byte sequence that simply could not be sent via the ethernet card in my computer. I changed ethernet cards and then I could send the file.
After a lot of searching I eventually found a few discussions of this, and an errata list for the chipset in my ethernet card that said that a particular revision of the chipset had an error in the checksum implementation that would compute an incorrect checksum for a particular bit pattern.
I was tempted to discard this as pure imagination, but this was a smart and knowledgeable fellow who worked as a programmer, so I'm assuming he had done some investigation before he came to this conclusion.
I know too little about networking equipment, but I can see how certain byte sequences (timed right?) would have some magic meaning. Given enough traffic you're likely to end up sending just such a sequence eventually. Perhaps certain versions of the WoW client spat out just such a magic sequence for the particular router he happened to use?