I think this is the result of the evolution from something like
Passive adversaries (typically External) , Active adversaries (almost always Internal) to Passive/External, Active/Internal
Also, it should be considered that, if an adversary reroutes traffic through itself in order to sniff, but makes no changes directly to the data being transferred; that adversary is ACTIVELY affecting the data stream in question, and is INTERNAL as part of data's route, but is also PASSIVE and EXTERNAL in regard to the packets' respective Application Layers in use (HTTP and IRC seem to be the two examples worth specifying here).
Also important: "A global passive adversary would be a person or an entity able to monitor at the same time the traffic between all the computers in a network." Your suggestion of "global external adversary" would only be possible in this sense if no Tor Nodes, ISP servers, etc. were needed to monitor all traffic network..... unless your meaning is external of the Tor network ...... but, as we can see, this brings more possible confusion. Also, the idea of a global passive adversary encompasses the threat of a "global external adversary", unless I am simply unaware of an existing example of an adversary actively affecting a network from an external position. Time based fingerprinting by an ISP server, for instance, is only active external of the Tor Network (making it passive and external in regard to tor traffic).