[Netconf] Information from chair phone call on 24 September

[David Partain <david.partain at ericsson.com>]

Hi,

This mail summarizes the conversation that the 4 co-chairs of
NETCONF/NETMOD had with interested parties on default issue.
This mail is NOT intended to re-open the question about default
handling but to confirm that what we believe is a reasonable way
forward is a decision that the working groups can live with.
Silence will be considered consent!

1. There were a number of comments during the call that indicated
that we need text that says that defaults are taken into
consideration in validation.

2.  The defaults question: No fundamental changes to the model
are needed.  That is, a server MAY choose not to send back values
that are the default as defined in the YANG model.  Instead, we
are going to try to address some things that will tighten up
potential inconsistencies (to "SHOULD level", not "MUST level")
in dealing with defaults.  There are cases where you don't
necessarily know what a non-answer about an object with a default
clause means.  In particular, this means looking at upgrade
rules, dealing with leafrefs and instance-identifiers.  We should
try to address the most common cases without trying to cover all
possible corner cases. We need specific text, which Andy and
Martin have been asked to address.

Again, we are not re-opening this discussion by sending this
mail: we wish to confirm that this is the right way forward.

Some expressed a wish that there be a simple, short explanation
of how defaults are handled in general so that one doesn't need
to be privy of the entire YANG history to understand why things
are the way they are.  If you believe that this is useful, please
offer text or, at a minimum, point out where you think text is
missing.

3. The NETCONF WG to check consensus on whether 'with-defaults:'
should be a SHOULD to implement.

4. In conjunction with the 'with-defaults:' work, the NETCONF WG
can consider adding an XML attribute in a report-all response
that tags default data as such to help NETCONF applications.
(This discussion is already underway.)

5. Juergen agreed to help with some examples of XPath expressions
refering to config / non-config / garbage data.  (I think I got
this right.)  If others believe that other examples are needed,
please suggest that on the mailing list.

6. We need to start thinking about (and writing about) the issue
of config vs. statistics vs. operational data.  This is something
that needs to be given serious thought and energy.  This will be
done separately (and after) the YANG work is complete.

7. We will try to put text in the architecture draft as a sort of
"placeholder" with respect to #5.  Text has been suggested on the
mailing list before and that should be put into the draft.

Thanks.

David