https://datatracker.ietf.org/meeting/123/materials/agenda-123-netmod
https://datatracker.ietf.org/meeting/123/session/netmod
WG Chairs:
Lou Berger (lberger at labs dot net)
Kent Watsen (kent plus ietf at watsen dot net)
WG Secretary:
James Cumming (james.cumming at nokia dot com)
Time: Monday Session III 14:30-16:30 CEST (12:30-14:30 UTC)
Room: Patio 2 |
https://datatracker.ietf.org/meeting/123/floor-plan?room=patio-2
Available During and After Session:
Notes: https://notes.ietf.org/notes-ietf-123-netmod?both
Slides: https://datatracker.ietf.org/meeting/123/session/netmod
Zulip (chat): https://zulip.ietf.org/#narrow/stream/netmod
Video/Meetecho: https://meetings.conf.meetecho.com/ietf123/[TBD]
Onsite tool: https://meetings.conf.meetecho.com/onsite123/[TBD]
Drafts (TGZ):
https://datatracker.ietf.org/meeting/123/agenda/netmod-drafts.tgz
Drafts (PDF):
https://datatracker.ietf.org/meeting/123/agenda/netmod-drafts.pdf
Datatracker: https://datatracker.ietf.org/group/netmod/about
ICS: https://datatracker.ietf.org/meeting/123/session/[TBD]
Available After Session:
Recording: http://www.meetecho.com/ietf123/recordings#NETMOD
YouTube: https://youtu.be/GiPUJU5VT7Q
Jabber Logs: https://www.ietf.org/jabber/logs/netmod
Duration: 10 mins
Discussion Leaders: WG Chairs
Mahesh Jethanandani: Comment on the immutable flag. Enquiry about the
status and this meeting indicates that this is nearly complete.
Balazs Lengyel: As the immutable flag will not become part of the model,
this is not what 3GPP asked for.
Duration: 5 min
Discussion Leader: Scott Mansfield
Drafts:
https://datatracker.ietf.org/doc/draft-ietf-netmod-sub-intf-vlan-model/15
Drafts:
https://datatracker.ietf.org/doc/draft-ietf-netmod-intf-ext-yang/16
Scott Mansfield: Plan is to update to a new version soon after the
IETF123 meeting.
Druation: 10 mins
Discussion Leader: Joe Clarke
Drafts:
https://datatracker.ietf.org/doc/draft-ietf-netmod-yang-versioning-reqs/11
Joe Clarke: Switch bullets 5 and 6 on slide 4
Reshad Rahman: We have additional comments on top of the YANG doctor
comments that needs addressing on module versioning.
Duration: 40 min
Discussion Leader: Rob Wilton
Draft:
https://datatracker.ietf.org/doc/draft-ietf-netmod-yang-packages/05
Ramkumar Rajagopalan: Can you explain what the YANG package compiler
looks like?
Ignacio Martinez-Casanueva: Are you planning to include deviations and
augmented-by here?
Rob Wilton: No plans to do this. There are complexities with adding
these into YANG packages.
Kent Watsen (as a contributor): There is effort to do full-include, will
this effort consider this as well?
Rob Wilton: We need to review this. Can you email me please Kent?
Lou Berger: I think you would end up with the same semantics,
implementation time vs definition time
Nigel Davis: We have found issues where mandatory items, turned out over
time to be conditional. We ended up with almost everything being
conditional.
Rob Wilton: We had this on the module level, not the package level but
it was too complex. We concluded that for the first version we should
keep it simple and perhaps consider this later.
Nigel Davis: We have a number of sitauations where we need to express
capabilities. Each one seems to point towards a solution to this
problem. Each one points to another definition and you then just follow
the tree.
Rob Wilton: In the YANG push work we've considered capabilities to which
leaves support ON_CHANGE. This is also a very complex problem and out
of scope of the work we're doing here.
Nigel Davis: You could have all options as conditional and then have
another thing that defines what are used and what are mandatory.
Ahmed Elhassany: From the package you can have a translation to YANG
library. Does this give us augmented-by?
Rob Wilton: You cannot put augmented-by into a package because you don't
know what a package includes.
Ahmed Elhassany: This might be an issue for people who depend on
augmented-by in YANG library. This is an issue and we'd like to have it.
Rob Wilton: Lets talk offline about this but I don't believe this is an
issue.
Draft:
https://datatracker.ietf.org/doc/draft-tt-netmod-yang-config-templates/00
Discussion Leader: Robert Wills
Slot length: 10 minutes
Ramkumar Rajagopalan: This approach seems odd to me, why not make this
an action?
Rob Wills: Open to alternatives, an action is a transient thing whereas
a template is a part of the state of the datastore.
Ramkumar Rajagopalan: Is the content an anyxml node?
Rob Wills: The template's content is anydata, with an ability to specify
patterns for list keys, but don't see anything that captures exactly
what you could put inside the template content.
Kent Watsen: You mentioned you are tracking the requirements. Do you
have a sense of how many are being addressed at this time?
Rob Wills: I don't have a specific number at this time.
Kent Watsen: It would be good to see if there are any gaps here. Also,
you mentioned there is no current method to edit metadata. I think we
should rely on edit-config delete and merge operations and not introduce
new mechanisms.
Rob Wills: I will take this to the list to ensure we capture it all.
Robert Wilton: You have not got the template appearing in the intended
datastore. You should know whether your template has been applied.
Rob Wills: There was not a consensus in the interim about what intended
should look like? Are you saying the template should appear in the
intended?
Rob Wilton: Yes.
James Cumming: Does the conditional validation based on whether this is
in a template or not force code changes on existing clients?
Rob Wills: Yes
James Cumming: Given applying apply-templates="t1 t2" what is the order?
Rob Wills: This would apply t2 and then t1 with last write wins.
Aihua Guo: This is generic but we already have YANG models that already
create their own templates. How do we do conflict resolution in this
situation?
Rob Wills: It would be confusing to use both the generic method (beiong
proposed) and a module that has its own templates mechanism.
Reshad Rahman: The use of metadata that changes the impact of the
configuration, was this done to reuse the existing edit-config RPCs?
Rob Wills: This was the main reason yes, also, the template is part of
the configuration, so edit-config seems natural.
Reshad Rahman: What would the client get back?
Rob Wills: Are you saying that at the point the client commits it needs
to know where that came from?
Reshad Rahman: Yes
Rob Wills: The authors will look at the traceability, thanks
Don Fedyk: I've used YANG lint extensively, have you give thought to
tooling to test this?
Rob Wills: We haven't given this much thought yet.
Mahesh Jethanandani: What happens when you delete a template? Does this
remove from all the applications?
Joe Clarke: +1 to transbility information for where did the template
come from?
James Cumming: If we remove a template application does this mean over
time you end up with apply-templates="" on all nodes? Or should you
servers remove this metadata when it equals ""?
Rob Wills: Yes, the draft hopefully says this, but if it doesn't it
should
Lou Berger: Both this document and the next has the issue of (parseable)
nodes
Draft:
https://datatracker.ietf.org/doc/draft-netana-nmop-yang-anydata-validation/00
Discussion Leader: Ahmed Elhassany
Duration: 10 mins
Rob Wilton: Are you adding data to YANG library?
Ahmed Elhassany: All we are saying is look for the namespace used in the
YANG library. You check every message at runtime.
Reshad Rahman (Speaking as NMOP chair): You said this was started in
IETF119 but this is a -00.
Ahmed Elhassany: Yes, this was our mistake on our part, we took it to
the wrong WG? The drafts are linked in datatracker though.
Lou Berger: Poll, is there interest to work on this
** Poll 1 **: "Is this a problem that should be addressed by the
WG?" Yes means I understand the problem, No opinion means I need
clarification on the problem statement
Lou Berger: ** Results **: Approx 50:50 split between Yes and No
opinion, my takeaway is that clarification of the probelm statement is
needed
Rob Wilton: The YANG push specification already says that any pushed
data must be valid.
Benoit Claise: Can you restate the problem
Ahmed Elhassany:
** Poll 2 **: Should the problem stated on slide 4 be worked on by
the WG? Yes means I understand the problem, No opinion means I need
clarification on the problem statement
Lou Berger: ** Results **: no change in results - perhaps prepare a
draft that clearly states the problem and leaves the solution to a
future revision
Kent Watsen: I was going to ask for the same.
Draft:
https://datatracker.ietf.org/doc/draft-rtv-netmod-yang-subtree-replacement/00
Discussion Leader: Rajesh T Venkateswaran
Duration: 10 mins
Kent Watsen (as contributor): You called it replacement rather than
deprecated, should this not be called that?
Rajesh Venkateswaran: While this is deprecated people should use this
but when it's obsoleted it would be removed.
Kent Watsen: You're using a string but this is actually a tuple. Can we
make this a structure somehow? The replacement might be multiple
entries, how would this be represented.
Lou Berger (as chair): We do not have agreement on the problem yet so if
we can table solution discussions for now?
Balasz Lengyel: We also have a solution
Reshad Rahman: Does the extension apply only to leaf nodes or also
containers etc?
Rajesh Venkateswaran: Containers, lists etc
Deepak Rajaram: Why do we need an extension vs instance data?
Rajesh Venkateswaran: So it can be parsed.
Lou Berger (as chair): I would like to poll on the problem statement.
Yes means I understand the problem, No opinion means I need
clarification on the problem statement.
** poll 3 ** should the problem on slide 2 be addressed by the WG?
Lou Berger (as chair): The results show that there is generally interest
in this topic, but some want more information and no objections.
Rob Wilton: To address extension vs. instance data, this is because we
need to evaluate at design time.
Balasz Lengyel: It is important to evaluate this in design time not
runtime.