[Nemops-interest] Re: [Nemops-workshop-attendees] Re: NEMOPS Workshop Report

[Wes Hardaker <[e-mail]>]

So I think in this discussion we are crossing the streams.

1. The conclusion in the meeting was that work was still needed in many
areas, and Michael's conclusions meet that -- until everything supports a
common way to do things, no one can exclusively use that "common way".
[this concept is protocol independent]
2. Having said that, we all agree [I think] that nothing works exclusively
everywhere, and Michael said he doesn't use Netconf/* because of #1 and #3.
3. When teaching someone how to administer *one* device, you probably don't
need a network protocol at all.  The average SOHO user does not ever need
something like netconf or any other protocol, as the whole process of
"device creates a temp wifi, you connect to it, configure it as needed, and
reboot and probably never touch the config again" is all they will likely
ever need.  Anyone touching a single switch or router: just the CLI is
likely sufficient.  This also seems to match Michael's case.

So, the question in my mind is: *if* we have extremely wide deployment of
protocol X (let's say netconf/*) then "would that change your mind into
using it?"  (with an implicit assumption of probably need to configure a
minimum number of devices to make the education hit worth it).




> Today, RESTCONF does not seem to be a superset of what SNMP MIBs can do
> today.
> I'm not saying that it true, but it doesn't seem that way.
>

Oh, and 4.:  monitoring != configuration.  So RESTCONF is not a superset of
SNMP, as it was designed to do something that SNMP was failing at:
configuration.  I gave a speech 15 years ago (or more) about why we should
have one protocol not 2 or more for configuration and monitoring because
implementing and deploying multiple protocols is a pain.  I apparently
didn't get my point across since "here we are".

-- 
Wes Hardaker
USC/ISI