Re: [netmod] IETF-95 Meeting Notes

[Lou Berger <[e-mail]>]

On 4/4/2016 10:39 PM, Ashesh Mishra wrote:
> All,
>
> Please review the notes from today’s NETMOD WG sessions and make
> revisions if we missed any comment that was made during the meetings. 
>
> http://etherpad.tools.ietf.org:9000/p/notes-ietf-95-netmod?useMonospaceFont=true
>
> ONLY add comments that were made DURING the meetings. :)
>
> Thanks,
> NETMOD WG.
>
>

Kudos to Ashesh and all who helped out contributing to the session and
the notes. Here's the actual initial raw notes...


                                        NETMOD  Agenda For IETF 95
                                       
                                        Monday, April 4th, 2016
                                        15:50-17:20  Monday Afternoon
session II
                                        17:40-19:40  Monday Afternoon
session III
                                        Room: Atlantico C
                                       
                                       

<ACTION> Update presentation links in Agenda (@Chairs)

0   Title:        Agenda & Intro
    Draft:        n/a
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-0.pptx
    Presenter:        Chairs
    Notes:

    Lou: New IPR rules. Chairs will request IPR disclosures from
authors/contributors at various steps.

    <Kent reviewed WG status>


1   Title:        YANG Summary
    Draft:        n/a
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-1.pdf
    Presenter:        Benoit Claise
    Notes:

    Mehmet: Do we have crieria defined with which we can analyze which
models are relevant for standardization?

    Benoit: Two answer.

    1. L3VPN service model (L3SM) will help determine with server YANG
models are required

    2. The other WG will have to decide for themselves


    Eliot: Should we hold progress to wait for OpState and Mount
compatibility?

    Benoit: They need focus and need to be resolved.


    Dean: Merge pyang 1.1 into main branch. Have an option to complile
with 1.0 or 1.1.

    Lada: That's what's being worked on.

    Charles: Is it possible to tie pyang to ID-to-nits or to xml2rfc?




2   Title:        Opstate & schema mount update
    Draft:        n/a
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-2.pptx
    Presenter:        Chairs
    Notes:

    Kent: Open items with requirements doc: Data node locality,
simulatenous access to intended and applied config, telemetry data.

    John Messenger(?): Crystalizing requirement document helps focus the
work, but maybe not everyone will agree with it.

    Lou: Requirements are good/right, but they are not the full set that
will be captured in the full solution. We'll know what's really right
when we implement it. At this time, we believe requirements document is
good enough to proceed with the solution.



3   Title:        Schema Mount
    Draft:        structual-mount/ysdl
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-3.pdf
    Presenter:        Ladislav Lhotka / Martin Bjorklund
    Notes:

    Chris: What is the difference between schema and a model?

    Lada: Good question, but I'm not going to answer that now. (?)


    Jason Sterne: In the logical device, if we want to put interfaces
and ACLs to mount point, can't we add references from ACL to interface?

    Lada: The sub-schema would contain both, interfaces and ACLs. So you
will be able to refer between the interface and ACL. Can't refer from
logical device to parent device.

    Eric Voit: There is also alias-mount. Will be discussed later, but
is not covered by schema mount.

    Andy: Concern that schema-mount throwing away YANG statements will
result in invalid models

    Lada: Within a logical mount point, one can refer to other leafs,
but not leafs in another root/mount point.

    Andy: I am fine with that restriction.

    Benoit: Is it possible that we'd have a deviation in a model and
another one in the mount?

    Lada: ???

    Dean Bogdanovic: Support schema node identifier option instead of
mount-point extension to yang.

    Lou: Different use cases (??) where mount-point fits better.


    Andy: Anydata tells clients that there may be nodes

    Lada: We need to discuss this more.



4   Title:        Needing an extensible Mount syntax across Schema,
Alias, & Peers
    Draft:        http://tools.ietf.org/html/draft-clemm-netmod-mount
    Draft:       
http://tools.ietf.org/html/draft-voit-netmod-peer-mount-requirements
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-4.pptx
    Presenter:        Eric Voit
    Notes:

    Dean Bogdanovic: Terminology is a bit different from Lada's presentation

    Eric: Shouldn't be any difference.

    Lada: Don't think there's any difference.

    Dean: Okay.


    Chris: Are the alias and peer mount client driven (does the client
control them)?

    Eric: peer mount is not signaled today. Defined in the mount point.
Mount server is the source of the info that is mounted from the client.


    Kent: with regards to defining an extensible syntax, this needs to
be discussed on the list.

    Lou: Is the peer/alias mount requested by the server or the client?

    Eric: he element doing the requesting of mounted data is the
client.   This is where the mountpoint exists.  The server doesn't have
to know that the data is being represented as Mounted.



5   Title:        Guidelines for Authors and Reviewers of YANG Data
Model Documents
    Draft:        http://tools.ietf.org/html/draft-ietf-netmod-rfc6087bis
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-5.pdf
    Presenter:        Andy Bierman
    Notes:

    Issue: "the anyxml statement MUST NOT be used to represent a
conceptual subtree of YANG data nodes

    Martin: "Must not use anyxml" is okay in 6087-bis. These
recommnedations are more strict than the language allows.

    Lou: personal preference to not having options, thus MUST NOT is
prefered

    Issue closed => leave the text unchanged


    John:???

    Andy: not in the draft yet


    Andy: Should --ietf pyang option be used on examples?
Maybe/maybe-not. If not, then use <EXAMPLE BEGINS> ... <EXAMPLE ENDS> tags

    Benoit: Should the <EXAMPLE BEGINS> tag be suggested to other SDOs?

    Mahesh: Other SDOs submit yang models as separate files so this
issue doesn't always apply to them.

    Kent Watsen: things "example" refers to XML instance documents,
maybe a better word can be found


    Andy: Requesting recommendations for other conventions that should
be in this draft.


    Acee: is this going to be in this document or another

    Andy + chairs: this one

    Mahesh: I prefer in a BCP


    Rajiv: Recommend a proper structure for a yang model (graphical
representation + guideline). Also, recommend what is contained in the
operational-state.



    1720:         Beverage and Snack Break


6   Title:        QoS YANG Model
    Draft:        http://tools.ietf.org/html/draft-asechou-netmod-qos-yang
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-12.pptx
    Presenter:        Mahesh Jethanandani
    Notes:

    Kent: Does this document replace the 03 diffserv document?

    Mahesh: Yes.

    Lou: Why was the model renamed?

    Mahesh: Original model excluded anything L2 related. Idea is that
this model should be extensible to other layers/features.

    Dean: Common abstraction of QoS is difficult because of different
implementations from vendors. Need more active participation from other
vendors.


    Benoit: SUPA has YANG model for policy and does such abstraction well.

    Mahesh: Havne't reviewed it. Will do.

    Dean: There's no way to extend what SUPA is doing for policy for the
QoS model.


    ??? from Ericcson (?): There's a QoS model in rtgwg.




7   Title:        SYSLOG YANG Model
    Draft:        http://tools.ietf.org/html/draft-ietf-netmod-syslog-model
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-6.pptx
    Presenter:        Clyde Wildes
    Notes:

    Kent: Is there a desire to extend this to support TLS?

    Clyde: Current draft supports this. Data elements to support key
exchange not added.

    Lou: What's left for this draft?

    Clyde: Benoit asked to make change to examples (?).

    Lou: Is the document ready after these changes?

    Clyde: Yes. 




8   Title:        YANG Model Classification
    Draft:       
http://tools.ietf.org/html/draft-ietf-netmod-yang-model-classification
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-7.pptx
    Presenter:        Dean Bogdanovic
    Notes:

    Jason Sterne: Why do we use extension vs augmentation in  "vendor
specific extension"

    Dean: May have multiple identities that not all vendors support.


    Kent: how much left until ready to Last Call

    Dean: it's ready now, assuming the WG is okay with the changes made
to this version of the draft


    Lou: what about going deeper and add more catagories (e.g. OAM)?

    Dean: customers have said that it's too complicated

    Lou: but the OpenConfig folks had a more detailed model




9   Title:        ACL YANG Model
    Draft:        http://tools.ietf.org/html/draft-ietf-netmod-acl-model
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-8.pptx
    Presenter:        Dean Bogdanovic
    Notes:

    Jason Sterne: Feels that interfaces shouldn't be part of base model,
doesn't think there is broad support


    Elliot Lear: there many be a need for other motdels to define
separate ACLs for input and output interfaces


    Chairs: call for if interface should be in base:

    6 prefer NOT having it in the doc at all

    5 prefer having it in, but as a feature

    2 prefer having it in the doc as required




10  Title:        Routing Cfg YANG Model
    Draft:        http://tools.ietf.org/html/draft-ietf-netmod-routing-cfg
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-9.pptx
    Presenter:        Acee Lindem
    Notes:

    <no questions>




11  Title:        Network Device YANG Organizational Model
    Draft:       
http://tools.ietf.org/html/draft-rtgyangdt-rtgwg-device-model
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-10.pptx
    Presenter:        Acee Lindem
    Notes:

    Dean Bogdanovic: Why are we putting the ietf keychain in top-level
instead of system managemnet?

    Acee: Would ACL go there too?

    Dean: It's HW specific.

    Acee: Devices put keychains at the top of the tree so that they can
be used across.


    Sue Hares: Put ACLs and filters at the same level.

    Dean: do you think they (keychain and ACLs) should be at the same level


    Benoit: how does it makes sense for the routing design team to
define a model that impacts models outside of the routing area?

    Dean: Splitting it up into different WGs would be complex and cause
the models to get watered down

    Lou: this draft is informational, should have no impact.


    Lou: do we want a structure or not?


    Chris Hopps: Benoit, would it be ok if we just talked to other
groups and kept the document progressing in routing area/wg?

    Lou: when we lost /device, we decided to make the draft Informational


    guy in red t-shirt (Jeff Tantsura?): would like to see it be
stronger than Informational

    Jeff Haas ???:


    Jason Sterne: RFC LNE and NI ... rest can be informational


    Chris Hopps: some think there should be no structure

    Andy B: don't go with absolute paths


    Lou: call for support (strong guideline vs Informational)

    Strong Guideline: 8

    Informational: 6

    Go away: 6

    Don't care: 1



12  Title:        Subscribing to YANG datastore push updates
    Draft:        http://tools.ietf.org/html/draft-ietf-netconf-yang-push
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-11.pptx
    Presenter:        Eric Voit
    Notes:

    Chairs: The details of this draft belong in NetConf.  Discussion
belongs there.



13  Title:        YANG Data Model for Configuration Scheduling
    Draft:        http://tools.ietf.org/html/draft-liu-netmod-yang-schedule
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-13.pptx
    Presenter:        Xufeng Liu
    Notes:

    Jason Sterne: have you considered any alternate (point-cut like)
approach of having a list that can point to subtrees where scheduling is
desired?


    Kent Watsen: Why do we want to standardize a grouping?

    Lou: It didn't seem to belong to teas wg. Scheduling was beyond the
scope of the te topology wg document.

    Jeff Haas: I2RS had discussion on calendar objects as part of
ephemeral state. This is a nice mechanism since the info is not ephemeral.

    Andy: All sibling nodes and sub-trees are activated/deactivated
based on schedule? How does this work with validation?

    Xufeng: The schema tree won't change. This is applicable to object.

    Jeff Haas: This does not affect the config. 




14  Title:        BBF YANG Models
    Draft:        n/a
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-14.pptx
    Presenter:        William Lupton
    Notes:

    Benoit: something about the guidelines draft?


    Balaz: can you put your mouldes on IETF github?

    Chairs: please discuss on list




15  Title:        Manufacturer Usage Descriptions
    Draft:        https://tools.ietf.org/html/draft-lear-ietf-netmod-mud-00
    Draft:       
https://tools.ietf.org/html/draft-lear-ietf-netmod-acl-dnsname-00
    Slides:       
http://www.ietf.org/proceedings/95/slides/slides-95-netmod-14.pptx
    Presenter:        Eliot Lear
    Notes:




Adjourn                19:40