> It does. The only complications is how to find a good way of merging > the access logs from two or more proxies to build the request graphs. > It is not always trivial to identify which requests on a upstream > proxy belongs to the downstream request, but by keeping the time in > synch (i.e. NTP) and sorting on time you can most likely make it > without to much fuzz. > > Regards > Henrik
Ok, this is making some sense to me. I'm thinking that we sort based on time, then there should be some millisecond - second discrepancy between each forwarded cache request and then we can follow the path until a HIT or MISS (go to ISP) is encountered. To further make things a little safer some access.log checks can be made, such as making checks on content, and url requested, or an algorithmic check from each node - by keeping a handle to the previous node to make sure it requested the same content, url requested and for the return path, the same number of bytes. All the best, Phillip Lucs
