[AGENT++] Agent++Win32 and Set Operation Failure with Multiple Variable Bindings

[Frank Fock]

Hi Dave,

"Only" Agent++Win32 needs to be changed. What needs to be changed are
the ExtensionDLL::*_set_request methods. Each method needs to collect
all VBs where the OID is within the DLLs subtree and then passes that 
collection
to the registered DLL instead of only passing the current subrequests VB.
When collection the VBs the respective subrequests need to marked as
"done" to avoid reprocessing when the AGENT++ API advances to the next
sub-request of the original SNMP request.

I will try to implement these changes for the Agent++Win32 1.5 release.
I hope this can be ready for release next week.

Best regards,
Frank



Dave White | Networking wrote:

>Hi Frank,
>
>Our problem is with a Windows Extensible Agent DLL.  I realize that this is a
>proprietary Microsoft protocol, but Agent++Win32 has implemented it with the
>benefit of AgentX++.  It seemed to me that AgentX was processing the requests
>one VB at a time.  However, if AgentX attempts to group VBs already, then any
>fixes should be limited to Agent++Win32 alone, correct?  This is a problem
>that we need to correct.  Do you have suggestions or ideas about what
>functions/methods needs to change?  Any other ideas you have about what needs
>to change, and how, would be helpful.
>
>Thanks,
>Dave
>
>_______________________________ 
>Dave White
>Sr. Software Engineer 
>  
>(650) 357-3980
>Dave.White at efi.com
>
>-----Original Message-----
>From: Frank Fock [mailto:fock at agentpp.com] 
>Sent: Monday, May 09, 2005 11:19 AM
>To: Dave White | Networking
>Cc: agentpp at agentpp.org
>Subject: Re: [AGENT++] Agent++Win32 and Set Operation Failure with Multiple
>Variable Bindings
>
>Hi Dave,
>
>Do you refer to AgentX subagents or Windows Extensible Agent DLLs?
>For AgentX all VBs of a SET request (served by a particular subagent )
>are always passed within one AgentX PDU to the subagent. So this is
>compliant to RFC 2741.
>
>For Win32 Extensible Agent subagents, VBs are always passed individually
>by the Agent++Win32 master.  Since this "protocol" is not AgentX I see
>not why it should be compliant to the AgentX RFC.
>Passing the VBs at once should be possible though, however this would
>make the current framework a bit more complex. If you need such a solution
>urgently then please contact me by email.
>
>Best regards,
>Frank
>
>Dave White | Networking wrote:
>
>  
>
>>Hi All,
>>
>>It appears that Agent++Win32 passes variable bindings one at a time to
>>extension agents during set requests with multiple variable bindings.  Does
>>this violate RFC 2741?  "When processing an SNMP Set request, the master
>>agent must send all of the VarBinds applicable to a particular subagent
>>session in a single agentx-TestSet-PDU." (p. 60.  See also page 64).
>>
>>I am working with subagents that expect to have a list of variable bindings
>>when processing set requests.  If the subagent receives variable bindings
>>    
>>
>one
>  
>
>>at a time, the operation fails.  Is this a problem with the subagent,
>>AgentX++, or Agent++Win32?
>>
>>Regards,
>>Dave
>>
>>_______________________________ 
>>Dave White
>>Sr. Software Engineer 
>> 
>>(650) 357-3980
>>Dave.White at efi.com
>>
>>
>>
>>_______________________________________________
>>AGENTPP mailing list
>>AGENTPP at agentpp.org
>>http://lists.agentpp.org/mailman/listinfo/agentpp
>>
>> 
>>
>>    
>>
>
>
>_______________________________________________
>AGENTPP mailing list
>AGENTPP at agentpp.org
>http://lists.agentpp.org/mailman/listinfo/agentpp
>
>  
>