2005-06-28 22:22:35 +02:00
|
|
|
|
|
|
|
|
Sparse notes on major existing problems/plans
|
|
|
|
|
|
|
|
|
|
* Re-design SMI model aiming at more consistent design
|
|
|
|
|
and better performance. One problem with current model
|
|
|
|
|
is that MIB variable and variable instance is not clarly
|
|
|
|
|
separated. Scalar vars and table cells should share the
|
|
|
|
|
same mechanics. For example, MibVariable may be a subtree
|
|
|
|
|
with a single leaf -- MibVarInstance; MibColumn would also
|
|
|
|
|
hold a ref to MibVarInstance then.
|
|
|
|
|
|
|
|
|
|
Linear search for subtree constributes to poor performance,
|
|
|
|
|
|
|
|
|
|
Possibly, MIB tree should be a single-level dictionary of
|
|
|
|
|
MIB objects.
|
|
|
|
|
|
|
|
|
|
Subtree registration at MIB tree objects may be obsolete
|
|
|
|
|
|
2005-07-27 17:55:42 +02:00
|
|
|
Another issue showed up in MIBVIEW object model: MIB module name,
|
|
|
|
|
MIB object name and object instance should be clearly separated.
|
|
|
|
|
Dedicated, simple API for table cells mangling should be provided.
|
|
|
|
|
The whole thing should be simpler, including API.
|
|
|
|
|
|
|
|
|
|
* Trap application code is still 90% finished. Also, examples & apps
|
|
|
|
|
needed.
|
|
|
|
|
|
2005-06-28 22:22:35 +02:00
|
|
|
* At MibViewController, do not index intance objects
|
|
|
|
|
|
|
|
|
|
* Implement 'oneliner' API for v1arch
|
|
|
|
|
|
|
|
|
|
* Implement prettyIn value format parser at TC
|
|
|
|
|
|
|
|
|
|
* Disallow SNMP access to USM objects
|
|
|
|
|
|
|
|
|
|
* Make sure objects shutdown methods work alright
|
|
|
|
|
|
|
|
|
|
* Maybe MibInstrumentationController should be async what
|
|
|
|
|
implies callback function interface.
|
|
|
|
|
|
|
|
|
|
* Re-work ACL to get rid of linear search there
|
|
|
|
|
|
|
|
|
|
* Implement SNMP trap and proxy apps; complete example apps
|
|
|
|
|
|
|
|
|
|
* Check for memory leaks esp. on failed requests
|
|
|
|
|
|
|
|
|
|
* Implement partial context name match at isAccessAllowed()
|
|
|
|
|
|
|
|
|
|
* Should managed objects be accessed through some dedicated,
|
|
|
|
|
non-SNMP-based API? A sort of backdoor which is basically
|
|
|
|
|
intrum minus transactions and AC.
|
|
|
|
|
|
|
|
|
|
* Clean up cache management code at MsgAndPduDsp and MP
|
|
|
|
|
|
|
|
|
|
* Sort out whether pyasn1 obejcts should probably have either clone() or
|
|
|
|
|
subtype() methods.
|
|
|
|
|
|
|
|
|
|
* Implement SNMP engine-scope debug mode logging
|
|
|
|
|
|
|
|
|
|
* Convert symbolic constants into integers for performance and typing
|
|
|
|
|
reliablity
|
|
|
|
|
|
|
|
|
|
* Expire discovered snmpEngineIds at mpmod/rfc3412
|
|
|
|
|
|
|
|
|
|
* Implement by-transportTag search at secmod/rfc2576
|
|
|
|
|
|
|
|
|
|
* Review PDU components coercion code at proxy/rfc2576
|
