Re: [netmod] call for consensus to adopt draft-wilton-netmod-intf-ext-yang as NETMOD WG draft

[Juergen Schoenwaelder <[e-mail]>]

On Tue, Dec 15, 2015 at 07:12:07PM +0000, Robert Wilton wrote:
> Hi Juergen,
> 
> Hopefully I can answer your questions inline ...
> 
> On 15/12/2015 17:08, Juergen Schoenwaelder wrote:
> >On Tue, Dec 15, 2015 at 04:48:21PM +0000, Kent Watsen wrote:
> >>The minutes for IETF 95 show that there was in-room support for adopting 
> >>draft-wilton-netmod-intf-ext-yang as a WG draft.   The minutes also show 
> >>that this decision would be confirmed on the mailing list, which I am 
> >>doing now.
> >>
> >>Should we move to adopt draft-wilton-netmod-intf-ext-yang as a WG item 
> >>and correspondingly add this to the WG charter as a milestone?  Please 
> >>comment by Tuesday, December 22, 2015 at 9AM EST at which time the WG 
> >>Chairs will gauge whether or not there is consensus to move forward with 
> >>the document.
> >>
> >This I-D contains some Ethernet specific definitions. Are we going to
> >define Ethernet specific things in the IETF or is IEEE ready to take
> >over data models for 'their' interfaces? In other words, what exactly
> >is the scope of this work?
> There is an informal IEEE 802.3 Ethernet design team (that has the 
> backing of the 802.3 WG chair, and that I'm part of) that is working on 
> YANG models for Ethernet interfaces.  The intention is that this 
> informal design team will become a formal IEEE 802.3 WG design team once 
> it traverses the necessary IEEE 802.3 WG processes (i.e. likely to be 
> sometime later on next year).  The exact scope of this project hasn't 
> been defined yet, and can't formally be defined until the IEEE 802.3 WG 
> agrees that we can do it, but my expectation is that the long term goal 
> will surely be to define YANG models to cover all of the 802.3 work, 
> although there may be various interim goals along the way.
> 
> In the interim, whilst we wait for the formal WG to be started, the 
> Ethernet design team are working on Ethernet models in Github 
> (https://github.com/8023YangDesignTeam/EthernetYang).
> 
> How does that overlap with draft-wilton-netmod-intf-ext-yang?  The basic 
> answer is that it shouldn't and arguably doesn't.  The only thing that 
> it defines related to Ethernet is three leaves related to MAC address 
> (configured value, operational value, and burnt-in value) that are 
> applicable to all interfaces that use Ethernet framing.  There are 
> various types of interface that use Ethernet framing but are not 
> Ethernet interfaces, nor necessarily defined in IEEE.  I.e. I'm thinking 
> of interfaces where a VPLS instances terminates to a layer 3 forwarding 
> instances, or where a pseudo-wire is terminated directly to a layer 3 
> forwarding instance.  But at the end of the day, if this part of the 
> draft needs to be defined as part of the IEEE 802.3 space then I think 
> that would be fine too - I don't think that it should really be an 
> issue, and I think that we can involve the necessary folks to ensure 
> that this bit gets to the right home.

Thanks for the background and explanation.

> >If we add something to the work of this WG, what will be realistic
> >milestones and how do we make sure we stay focused? Is there a need
> >for some prioritization?
> I believe that at least some of this configuration is required to 
> configure networks in a reliable way, so I would have thought that we 
> need to progress these models at the same time as the routing protocol 
> models.
> 
> On a related note, any VPLS or Pseudowire models defined by IETF are 
> basically going to be undeployable without 
> draft-wilton-netmod-intf-vlan-yang-01 (or an equivalent) being defined 
> because there will be no configuration mechanism to bind the 
> classification of traffic from a particular VLAN to a VPLS instance.  
> Note that I don't see that any models coming out of IEEE 802.1Q are 
> likely to help here.  This point was also raised in PALS at IETF 94 by 
> some of the folks working on, or reviewing, those models.

So what will be realistic milestones? There are many things needed to
configure a complete network using YANG models and we need to make
sure we are able to finish what we start with realistic milestones
(and realistic really boils down to have a sufficient number of
dedicated reviewers lined up with the necessary cycles available to
make the milestones happen).

It might help the WG chairs to not only see "I support this work"
statements but also explicit "I support this work and I will provide
the time necessary to review the drafts as they progress through the
WG" statements.

/js

-- 
Juergen Schoenwaelder           Jacobs University Bremen gGmbH
Phone: [phone]         Campus Ring 1 | 28759 Bremen | Germany
Fax:   [phone]         <http://www.jacobs-university.de/>