The concept is quite self-explanatory: there are connectors that are triggered from SN, connect to event sources and fetch the events - these are the pull connectors.
Push connectors have "listener" components, on the instance itself, or on a MID server, to which 3rd party application can connect and send events.
So in one case the "active"/triggering component is in SN in the other case it is the 3rd party application.
Similar to how there are REST Messages (SN calls 3rd party end-points) and Scripted REST APIs (3rd party components connect to SN).
When talking about source, it is a little vague without context, but I'm assume you are talking about event sources, in which case are a means to bundle event processing rules and link that rule bundle to incoming events, whether pull or push.
When talking about destination, I'm not sure what you mean.
