Re: [netmod] operational knobs

[Andy Bierman <andy at netconfcentral.com>]

Romascanu, Dan (Dan) wrote:
>  
> 
>> -----Original Message-----
>> From: Phil Shafer [mailto:phil at juniper.net] 
>> Sent: Monday, November 02, 2009 4:42 PM
>> To: Romascanu, Dan (Dan)
>> Cc: Juergen Schoenwaelder; Randy Presuhn; NETMOD Working Group
>> Subject: Re: [netmod] operational knobs 
>>
>> "Romascanu, Dan (Dan)" writes:
>>> It was a political decision which resulted in parsing errors :-)
>> But couldn't the political decision have been reversed when 
>> the tide of "not so simple variables" began to flow in?
>>
>> Thanks,
>>  Phil
>>
> 
> It was not that easy, and the success of the protocol was part of the
> reason. It took a few years to realize that CMIP / CMOT will never
> happen. By that time SNMP was working at least for some applications the
> IETF needed, with security being the principal flaw the whole community
> was told to focus for another few good years. You can consider SNMCONF
> and SMIng as tentative to finesse the issue without rocking the boat of
> the protocol. And then somebody said X-M-L !!
> 

There were some of us that were tried to get SNMP to
change so that SNMP Set would be relatively implementable.
The 'gumball machine effect' in SNMP, and the lack of complex data
structures in SMIv2, made distributed control tables a nightmare
to code in the agent.  CLI accepts 1 command at a time. SNMP
accepts an arbitrary number of leafs from an arbitrary number
of commands, and reconstructs commands as they arrive in Set PDUs.

Something as dirt-simple in NETCONF as <make-toast> is a nest of kludges
in SNMP.

For awhile, many OPS people did not really believe operators when
they said they liked CLI because they could read it.  Think OID.
Then go as far in the opposite direction as possible.  Some of the
many benefits of XML -- it can be read, cut-and-pasted, and emailed.

But we don't need to replay the XMLCONF discussions that led to
the formation of the NETCONF WG.  SMIv2 action knob --> NETCONF <rpc>.
End of problem (for now).


> Dan
> (apologizing for historical soapboxes)
me too!

Andy