Re: configuration: writable MIB modules versus NETCONF/YANG modules

On Thu, 20 Feb 2014, Andy Bierman wrote:

> IMO if a WG is designing a NETCONF/YANG solution for some feature, it 
> should include monitoring and notifications in NETCONF/YANG as well. I 
> think SNMP makes the most sense for a WG if no configuration is 
> standardized.

My view is that we should try to replace SNMP over time, and netconf/yang 
should be used for monitoring and configuration. I wish SNMP write would 
go away, mainly due to beforementioned security problems. I am not sure 
the current implementation of netconf transport for yang makes sense for 
bulk collection of data, and that's perhaps something that should be 
rectified over time. However, I fully support the initial proposal to 
encourage more netconf/yang usage over SNMP and other configuration 
methods.

Question is, since it's hard to make the working groups create SNMP MIBs, 
it's going to be even harder to make them create yang models (at least 
initially), how is this going to be achieved?

Also, doing models by means of publishing RFCs is kind of heavy process 
approach to achieving a structured model of management. I wish we could 
come up with something a bit more lightweight that would support including 
model extensions within the drafts proposing new behaviour.

For instance, today an author can request protocol number allocations from 
IANA etc. Wouldn't it make sense to have a similar model for extensions of 
yang models?

-- 
Mikael Abrahamsson    email: [email protected]