Re: possible DoS attack

On Friday 12 April 2002 17:24, Aleksey Sanin wrote:
> I suggest to remove phrase "to identify such attacks" :

By that, I meant if you get hit once (a request consumes X cycles before 
timing out), there's no need to honor a request from the same person to 
consume another X, but I agree that simpler is better. So the last sentence 
now reads, "Consequently, implementations should be able to restrict 
arbitrary recursion and the total amount of processing and networking 
resources a request can consume." Everything else (identifying repeat 
offendors, dynamically reseting/adjusting the thresholds, etc.) is an 
implementation detail! <smile/>

Received on Tuesday, 16 April 2002 15:59:54 UTC