netconf implementation guide

[Andy Bierman <ietf@andybierman.com>]

Hi,

I think eventually we will keep collecting clarifications,
maybe even enough for a NETCONF Implementation Guide RFC.

Are there any WG members willing to work on a
NETCONF Implementation Guide (Informational) RFC?
Is there community interest in having such a document?

If so, the first step is collecting implementation experience,
FAQ material, document clarifications and errata, etc.


==========================================

2 more issues that aren't very obvious
(No, nothing more is going into prot-11;
NETCONF Version 1 is in the can.)

1) Error processing hierarchy

There are some errors that need to be processed before others.
It is reasonable for an agent to stop looking for certain types
of errors after particular errors have already occurred.

For example, if the target of an edit-config is an unknown
configuration name, or the user doesn't have access rights to
invoke the edit-config method, then the <config> element sub-tree
may not be analyzed for errors, regardless of the 'continue-on-error' flag.

It is therefore possible (e.g., big edit-config request) for a manager
to receive an rpc-reply before it finishes sending the matching
<rpc> request.

2) Edit-config processing

The merge order is data model dependent.
A NETCONF data model definition which allows
multiple instances of an element must indicate the
merge order somehow.

When processing a replace operation, it is important
to know if a sub-node (in the data model, not the PDU!) is
 1) mandatory
 2) mandatory-with-default
 3) optional

If (1) then missing sub-elements are errors.
If (2) then missing sub-elements with defaults are filled in before
  checking for missing elements.
If (3) then this is not an error, but any corresponding nodes
 in the target that match the missing nodes must be removed

A NETCONF data model definition must indicate the <edit-config>
replace behavior somehow.



Andy



--
to unsubscribe send a message to netconf-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/netconf/>