Re: [Netconf] [netmod] meaning of default

[Andy Bierman <andy at netconfcentral.com>]

Balazs Lengyel wrote:
> Hello,
> I have the feeling that we should make with-defaults MANDATORY for all
> nodes using YANG.
> (Currently report-all is the only mode that is mandatory to support in
> with-defaults.)
> Opinions?

I was thinking that, until Phil explained to us
that a system-created node (not a YANG default)
is an explicit node.  Every defaults style MUST
return s-c leafs.

The leafs that can get left out are YANG defaults.
Here are some open issues that impact the reliability
of the meta-data the client needs to derive the proper defaults:
I think MUST implement with-defaults is a more user-friendly
solution than the following:

   * Agree that using revision statements is mandatory

   * Agree that the server MUST advertise all its modules in the <hello>

   * Agreed that import/include by revision SHOULD be used in vendor modules
     and MUST (+) be used in standard modules.

   + I am not convinced mandatory import-by-rev is a workable solution.
     I think it will lead to a rats-nest of version dependencies.
     Instead, YANG should specify that the most recent revision date supported
     by the server MUST be used if no revision-date is provided.

     E.g, A -> B -> C, but now A wants to import the new version of C,
     but also wants B to use the same version of C; otherwise
     typedefs/groupings in B are out-of-synch with those in A.

     If vendor(A) is not vendor(B), then updating B may be difficult
     or impossible.  In the IETF, it will mean republishing RFCs
     just to change 1 import-stmt, which could turn into a
     boiler-plate/process nightmare.

> Balazs

Andy