RE: Help/guidance on L2TPv3 MIB draft

[Andy Bierman <abierman@cisco.com>]

At 05:33 PM 9/29/2002 -0400, Natale, Robert C (Bob) wrote:

>Hi Andy,
>
>I think you have missed the point of the
>discussion around AgentX at the time it
>was developed concerning MIB design.  There

I don't think it's a good idea to force a MIB designer
to use a table when a scalar is needed, just for the
sake of AgentX.  Why would multiple sub-agents need
to register for a scalar anyway?  Monolithic vs.
sub-agent design has nothing to do with whether or not
a MIB object is designed to have one instance or many instances.

Andy


>are some MIB design approaches that are
>preferable for (open standard) extensible
>agent implementation in general.  Enabling
>multi-sub-agent support of sub-agent-specific
>configuration (or similar) attributes --
>normally represented as scalar objects in
>monolithic agent designs -- via the tabular
>approach that you seem to consider "more
>complicated" (?) is one of those straight-
>forward (my assertion) approaches.  Our hope
>was to promote such MIB design choices going
>forward.  Improving MIB designs along these
>lines *might* become more important if the
>SNMP community decides to extend the scope
>of MIB functionality as part of an attempt
>to rejuvenate SNMP[v3] as the management
>protocol of choice in ever-wider arenas.
>
>As per its original design intent -- and
>the Applicability statements in section 4.2
>of RFC2741 -- AgentX can certainly support
>MIB designs that implement a monolithic
>agent approach to scalar objects...in which
>case a single-sub-agent handles such objects
>(which might or might not be a limitation
>for any given application...and wouldn't be
>for any MIB that is expected to be implemented
>only by a single sub-agent with a given
>master agent scope).
>
>Cheers,
>
>BobN
>- - - - -
>> -----Original Message-----
>> From: Andy Bierman [mailto:abierman@cisco.com]
>> Sent: Sunday, September 29, 2002 4:18 PM
>> To: Natale, Robert C (Bob)
>> Cc: Jed Lau; mibs@ops.ietf.org; agentx@dorothy.bmc.com
>> Subject: Re: Help/guidance on L2TPv3 MIB draft
>> 
>> 
>> At 03:46 PM 9/29/2002 -0400, Natale, Robert C (Bob) wrote:
>> >At 8/19/2002:08:26 PM, Jed Lau <jedlau@cisco.com> wrote:
>> 
>> I have a strong objection to this suggestion to change
>> a MIB design solely to make AgentX implementations easier.
>> When AgentX was chartered, there was an understanding that 
>> sub-agents would be transparent to NMS applications (command 
>> generators).  Asking MIB writers to use a more complicated
>> design is a hack. It makes the MIB more complicated
>> for everybody, for the sake of a corner case. That's not
>> good engineering.
>> 
>> Andy
>> 
>> 
>> 
>> >Hi Jed,
>> >
>> >Please review this MIB for compatibility with the
>> >IETF standard SNMP agent extensibility protocol,
>> >AgentX (RFC2741...and RFC2742 can also give you
>> >some additional insight into AgentX operations).
>> >
>> >The fundamental issue is that scalar objects are
>> >potentially problematic for AgentX. They should be
>> >avoided by putting them into a table indexed by
>> >some form of "AgentEntityID" object (an exercise
>> >left to the implementer for now :-().
>> >
>> >Note that the issue of AgentX compatibility for
>> >MIB writers will be addressed in a forthcoming
>> >Internet Draft that I will submit...perhaps with
>> >an extended applicability statement for AgentX
>> >(Sections 4.2, 4.3, and 4.4 of RFC2741 do a good
>> >job of the basics already)...for consideration by
>> >the relevant IETF WGs. My hope is that consideration
>> >of this Draft will lead to adding a statement or two
>> >to the standard "SNMP Framework" boilerplate to
>> >document these requirements for MIB writers.
>> >
>> >Guidance from Bert re where to discuss this
>> >issue -- AgentX compatibility guidelines for
>> >SNMP MIBs -- will be appreciated. (The AgentX
>> >e-mail list is still operative, but the WG is
>> >closed pending further work when its time to
>> >move to Full Standard status.)
>> >
>> >Cheers,
>> >
>> >BobN
>> >- - - - -
>> >>Hi,
>> >>
>> >>I am one of three co-authors of the L2TPv3 MIB draft. I 
>> attended the IETF
>> >in
>> >>Yokohama in July, and I got this e-mail address during one of the
>> >plenaries. I
>> >>understand that somebody behind this e-mail address might be able to
>> >provide us
>> >>with some guidance in developing our MIB.
>> >>
>> >>The first draft of the MIB is 
>> <draft-ietf-l2tpext-l2tpmib-base-00.txt>.
>> >We'd
>> >>like to get some feedback regarding our organization of the 
>> MIB tables. The
>> >>L2TPv3 (Layer Two Tunneling Protocol, Version 3) MIB 
>> borrows a lot from its
>> >>predecessor, <draft-ietf-l2tpext-l2tp-mib-04.txt>. However, 
>> it introduces a
>> >>new framework that, we hope, allows us to better modularize 
>> the various
>> >>components of the tunneling protocol (e.g. pseudowire, 
>> payload, transport,
>> >>etc.). The intention is to allow future enhancements to the 
>> protocol (e.g.
>> >>additions to the list of supported payloads or transport 
>> types) to occur in
>> >>scalable and modular manner.
>> >>
>> >>I intend to go into further details and ask some specific 
>> questions in a
>> >>subsequent e-mail. Should I send follow-up e-mails to this 
>> address, or
>> >should
>> >>I interface with a specific individual?
>> >>
>> >>Thanks in advance.
>> >>
>> >>Jed Lau 
>>