**
** BUGS file for the Haskell binding to net-snmp
**

The binding is new and largely untested; there are sure to be bugs and
many omissions.

** 
** SECURITY CONSIDERATIONS
** 

This code has not undergone any security review, and is very immature.
It would be inappropriate to use this binding where security is a
priority, without taking appropriate measures.

** 
** MISSING FEATURES
** 

This binding only supports a subset of the NetSNMP library's
"single-session" api.  Only synchronous queries are supported, and only
the get, next, and walk operations.

No support for bulk get requests.

No support (and none planned) for snmp traps.

Many SNMP / ASN.1 data types are unsupported as yet.

This binding only addresses use of the library for an snmp client.  No
effort is made to support snmp agent or proxy operations.

No support for snmpv3; currently only snmpv1 and snmpv2c with admittedly
insecure SNMP community strings.

This binding doesn't do anything with the library's log handling, and
the library's default action is to spam stderr.  There should at least
be a way to prevent this, and it should possibly suppress the logs by
default.

This code should be publicly hosted, perhaps on code.haskell.org.  It
should use a public trac.

** 
** BROKENNESS
** 

Looking up OIDs by name doesn't seem to work yet.  Once it's fixed
snmpwalk probably needs to be fixed as well.  It uses `isPrefixOf` to
determine when the walk is over, which won't work if the original
oid isn't dotted-decimal.

This binding's exception checking is spotty and amateurish.  Some of this
is because I haven't found any decent API docs for the net-snmp library.

There are too many things called 'session' in the library, and in the
binding:
-  the C struct snmp_session.  The 'single session' examples use *3* of
   these, with subtly different purposes, never fully explained.
-  data SnmpSession, the binding's representation of 'struct snmp_session'
-  data Session, the binding's abstract session container, holding
   the above and managing C memory (de)allocation

snmpWalk doesn't check for cyclical responses from an agent; reportedly
there are agents with this misfeature.  Currently if it happens snmpWalk
will happily follow the cycle until it runs out of memory or stack.

I don't know how to intelligently set the alignment parameter for data
types representing C structs in their Storable instances.  I'm using 16
bytes as a (hopefully) conservative setting, but isn't there a way to
detect and mimic what C uses?

The binding is currently one monolithic .hsc file.  It should probably
be organized into a few modules.