They are tapped into the hubs of internet communication as well as most (if not all) of the major webmail based systems (think Google, Hotmail, Yahoo).
As to the how they are taking feeds directly off major internet routers (the vast majority of traffic will go through a major router at some point particularly if it is international though it's quite possible for a packet sent from one side of your country to another to go international as well).
So yes they do have email content capabilities (if you look at the actual slides they also have a sophisticated filtering system, they can do stuff like "show me emails from iran with word documents attached containing IAEO").
This system is absolutely terrifying, it genuinely is the work of a dystopian sci-fi author from 30 years ago.
----
If you want to get right down in the trenches email is SMTP and POP over TCP/IP (normally), email is fundamentally a human readable text protocol which makes it trivially easy to parse (this was kind of the intention after all) so once they have the captured stream reconstructing the mail is not much harder (if any) than writing a mail client.
You can see an example of SMTP if you open a console/shell and type "telnet smtp.gmail.com 25" and then when it has logged in type HELO the response is just plain text.
As to the how they are taking feeds directly off major internet routers (the vast majority of traffic will go through a major router at some point particularly if it is international though it's quite possible for a packet sent from one side of your country to another to go international as well).
So yes they do have email content capabilities (if you look at the actual slides they also have a sophisticated filtering system, they can do stuff like "show me emails from iran with word documents attached containing IAEO").
This system is absolutely terrifying, it genuinely is the work of a dystopian sci-fi author from 30 years ago.
----
If you want to get right down in the trenches email is SMTP and POP over TCP/IP (normally), email is fundamentally a human readable text protocol which makes it trivially easy to parse (this was kind of the intention after all) so once they have the captured stream reconstructing the mail is not much harder (if any) than writing a mail client.
You can see an example of SMTP if you open a console/shell and type "telnet smtp.gmail.com 25" and then when it has logged in type HELO the response is just plain text.