Hi all,

We're starting up a small-scale project which aims to simulate/emulate
some SNMP capabilities of network elements on an IP network. We're
doing this as a master's thesis, and initially we're looking at
different ways to replicate (a very limited subset of) the
functionality of a real-world NE.

More specifically, we're looking to build an agent which can send any
traps/notification PDUs that a given NE can send, and also reply to
SNMP-GET-REQUESTSs with a valid (but not necessarily realistic)
SNMP-GET-RESPONSE. We're mainly concerned with SNMP v1 and v2 so far,
but extensibility into v3 in the future is wanted.

The ideal way would of course be to parse and load the pertinent MIB
files and build a model from there. Other methods could include
sniffing actual IP traffic and extracting the needed information from
there, and also also doing a simple SNMP-walk of the NE.

What we'd like to ask the experts on this group is what you think of
the feasibility of these three methods.

We've come to realize that the biggest problem in the first approach
is in the varied quality of vendor MIB files. Parsing a random MIB
file seems to be no trivial task, since even the commercially
available options have trouble loading everything we throw at them.

Is there any particular API that you would recommend for this task?
Currently, we're looking at Adventnet, Mibble, NetSNMP and SNMP4J. The
preferred language for us is Java, so that gives Adventnet an
advantage I guess. Does anyone have any experience in exposing
e.g. Adventnet's parser to a wide variety of MIB files?

On another note, would an SNMP walk of an agent always be able to
provide us with enough information to replicate the different
notification PDUs sent by that agent?

Best regards,

Joel Hansell
Jonas Lindquist

--
Google doesn't seem to be working. When i press the "i'm feeling
lucky" button i'm left unsatisfied -- MrSaturn