Networking Kid / Addresses, Names And Routes

Why The Same Name Reaches Different Machines

Ask for the same name from two cities and you may be answered by two different places.

A popular service is rarely one machine. It is many machines in many parts of the world holding the same content, and the name translation step is where the choice between them is quietly made. When a device in one region asks for the address behind a name, it can be given the address of nearby equipment, while a device in another region asking the identical question is given something different. Both answers are correct. Neither is a trick or a misconfiguration. The name is a pointer, and a pointer can point differently depending on who is asking and where they are asking from.

The reason is distance, which we will come to in detail. Content served from equipment physically closer to you arrives sooner, and no amount of capacity anywhere in the world can compensate for a long round trip, because capacity and time are simply different quantities. Serving many copies from many places is the only real answer to that problem, and steering each person to the nearest available copy is most naturally done at the moment the name is translated, before any of the real traffic has been sent anywhere at all.

This explains behaviour that otherwise looks like a fault or even a conspiracy. A site can be plainly broken for you and plainly fine for a colleague in another city, because the two of you are not talking to the same machine at all and never were. It also explains why asking somebody far away to check a site is genuinely useful information rather than a courtesy: their answer tells you whether the problem belongs to the service as a whole or only to the particular copy you happened to be sent to.