Re: SNMP MUX, SNMP DPI, et al

["James R." (Chuck)Davin <[e-mail]>]

Unni, you may find ietf-draft-snmpsec-admin-00.txt helpful.

Return-Path: [e-mail]
Received: from nisc.psi.net by PTT.LCS.MIT.EDU via TCP with SMTP
	id AA15379; Fri, 12 Jul 91 15:21:53 EDT
Received: by nisc.psi.net (5.61/2.1-PSINet Operations )
	id AA28759; Fri, 12 Jul 91 15:11:28 -0400
Received: from nisc.nyser.net by nisc.psi.net (5.61/2.1-PSINet Operations )
	id AA28754; Fri, 12 Jul 91 15:11:26 -0400
Received: from Maui.CS.UCLA.EDU by nisc.nyser.net (5.61/2.1-)
	id AA03855; Fri, 12 Jul 91 15:09:03 -0400
Received: by maui.cs.ucla.edu
	(Sendmail 5.61a+YP/3.07) id AA24201;
	Fri, 12 Jul 91 12:13:36 -0700
Date: Fri, 12 Jul 91 12:13:36 -0700
>From: [e-mail] (Unnikrishnan Warrier)
Message-Id: <[e-mail]>
To: [e-mail]
Subject: SNMP MUX, SNMP DPI, et al


Hi:

I am in the process of specifying my own agent variables for systems management 
(under Unix, DOS-Windows and OS/2, if you MUST know!) under SNMP.  However,
 since some vendors (notably IBM, and perhaps Sun?) are shipping their own 
agent on their OS (e.g. AIX on RS6000), I find myself in a confused situation.
  Turns out, either my SNMP, or their (IBM's) SNMP will battle for control of 
the SNMP port on the machine.  So I need to multiplex both my variables (my 
sub-agent) as well as their variables (sub-agent) via the same port.  There 
are two published ways to do it:

1.  via MarshallUs SNMP MUX protocol
2.  via the IBM SNMP-DPI

Both of these have their advantages and disadvantages.  In addition, Jeff Case 
appears to have even BETTER ideas.  So does Karl Auerbach.  Is it a good idea 
to try to standardize on a SINGLE set of ideas?  Do we have enough experimental 
knowledge out there to decide which ideas are the optimal set?  

If anybody has suggestions as to which way I should go, please e-mail me, or post to the list.  

unni
[e-mail]