No subject

view checking etc. Subagents just register the region they are responsible
for and answer any query that gets forwarded to them by the master. If a
query can not authenticate itself or tries to make access outside its view,
the master won't even forward it to the subagent. There is no authentication
on the AgentX-Protocol, so the master should only listen on the loopback
interface or use a unix socket to limit access to local processes only.
Traps are simply forwared to the master which then uses TARGET-MIB etc. to
determine where and how to send it out on the network. The subagent does not
know anything about that.

Hope this helps.
Henning