Internet Engineering Task Force 103, 6 Nov 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF103-NETMOD-20181106-0900

Description

NETMOD meeting session at IETF103
2018/11/06 0900

https://datatracker.ietf.org/meeting/103/proceedings/

A

B

B

To capture what's being said, and importantly, you might at least go back and check that we comment.

B

There or anything that you actually didn't say.

B

B

Think, oh yeah, well I think this. This governance are disposed to this. How we govern our process here and coming Foundation and the main takeaway is a thing you say here.

B

We are reporting what we're saying we are also setting it. The echo please use the microphone and state your name. Please.

C

Keep going around and there's a.

C

Lot so we have two sections this time. Thank you for making it to this one first thing we have.

B

Set up the agenda so that we're covering all the drafts.

C

That's on this day,.

C

D

A little bit of a rearrangement.

B

If you are at all interesting in the versioning discussion, please come the requirements. Document is has been a product of the design team and we really want to make sure we have good agreement and working group on the requirements that are documented and make that a working group document where we go with that in the long term, we'll figure that out we'll figure out whether we end up taking it to full RFC or not.

B

The important thing is is that by making a working group document, we are getting agreement that those are the requirements that the solution is going to adhere to. So it's a really important discussion, so please come participate in the next session. On that.

B

Alright, so since the last meeting what's happened, we have one new RFC, that's awesome to actually get an RFC out and we have three that are in the eye with the ihe for publication. There's been a good amount of discussion on ACL model scheme amount. I. Think we're close, that's pretty important to get out, of course, I'm biased, because I worked on that, but it's gating other work. So we really want to make sure that that is finalized. I think there was a request for some feedback from the authors is Martin here.

B

I, don't opens oh okay, so Martin. If you are listening, I know there might be an outstanding response necessary on that document.

B

Okay, since the last meeting, we have adopted a few documents and were going to be talking about them now, I believe the artwork folding has already been republished as a working group document. So it's now draft IETF I made we made up these slides before that was was done. We will see. We expect to see the yang instance data submitted real soon. Oh.

B

So I just looked at the author and said you know: what are you gonna do your job and his response was cherries. You haven't done yours because we have to approve it, so we apologize for that and that will happen by one of us during this session.

B

This is more just for awareness, general awareness. We have some old liaisons, we didn't get anything new, but it's good to know that these are out there. In particular, the invitation to participate in the monthly conference calls is a pretty substantive invitation and we wanted to remind the group that that's out there and, if you'd like to join those discussions, particularly if you're interested in the OEM topic with it's an opportunity to contribute to a different group and help align our work with their work or their work with our work and I.

B

Guess that's the end, because I'm clicking there's no more so we're going to move on the next.

C

B

Click to the next tab, yeah and there's a full screen.

E

F

E

Okay, so so quick states update on history drafts and they're actively almost done so there's not much more to be 70.

E

It's my pick of working-age scroll, okay, so the first draft means face more interested. Extensions draft has been no update since the last ITF so effectively think that this draft is done. So let's call it to the chairs and I think we struct is done. My question is whether we should hold off progressing this a further until we've got the songs face. Vlan lot, all done so I'll just cover that as well, and then we'll come back to that. So next slide.

E

Please Joshi works okay, so the the 70s VLAN model, one again, there's been no updates its last ITF effect again. This was also done. This is weighty on two things: one is the LTP na-yang model in their structure, the way that they augment the network instance container they've lost that little container. So all the LPN configs sits directly under the network instance. I wasn't quite sure that that was right or not so query that with them and and that only affects the examples but I think it'd be good to try and get those examples correct.

E

So then the plan I think for the altitude model was that it's going to go to law school fairly soon in Bess, so I think I'm. Just waiting on that close on that issue, and maybe that resolved this week and then the last one is effectively.

E

There was some changes made to the model to make it simpler and I wanted to get approval from the I Triple E 8201 should be eighty to one working group on those changes, so I've emailed them yesterday, not as a form of the aides of informally asking them to have a look at that and give comments back.

E

I have had someone comment back, pointing out a mistake in that model in one of the Wednesdays and identity, so I will fix that so needs to be one update for that and I'm hoping that effectively once a to get one agree that they're happy with the changes, then we can go to work with my support as well. If we.

B

Had talked about formalizing that liaison at the last meeting, what do you think is the informal adequate or should we do something more formal and I'll note that the I Triple E is meeting next week. I, don't know if there's anything when it try to accomplish informally getting a liaison done in that time, I feel a little quick, but so.

E

I think because he's been discussing before I think your formal should be fine. Unless they come back and say, look, we need this be formally discussed and I, don't see if John's in the room or Glenn here but I, know John, certainly ITF this week. I will try and speak to John messenger he's the eighty to one and vice chair just to try- and you know tribute along effectively this relatively minor changes. It's not like anything huge is just sort of simplification in the model, but she's going to check the other half of the names.

E

So if they are and I think through the iesg process, I suspect at that point they'll also check. They do got one but they're happy that's what happened with one of the other drafts and so I. Don't think formulas is required unless they you asked for one is mine and.

B

You participate in the coordination meeting, they're gonna be yeah, I think so you'll bring it up. Yeah.

E

Various points, yes, sir, what various places yeah yeah thank.

G

B

Any other questions yeah just for your own information. We we one of the biggest questions, is what we do to get this so that we can last call it. So it's good to hear you're making those steps so as soon as you think, you're ready you'd like to know that well.

E

The yeah I guess the one question to you is: we could work. We could potentially not this one with the previous one I. Could we could do last call on that one now probably's do to them together. Yeah.

B

If we're really close, it would make sense to do them together. Obviously, we wouldn't want to hold anything up. You know, you know. If we come to the next meeting and we're in the same place, you.

H

B

Okay, one technical comment: just looking to your slides about the good reveal to VPN one.

B

So no, no, this is it. So is there anything you want to raise this with the the best group and where do you stand I've.

E

Raised it with the authors of that draft before it's ITF I like a couple of weeks ago, and they said that they were happy with the current structure, so I'm just gonna, wait to see how well that goes to current.

B

Structures is there the way they have it documented or the way you have it in the example, but.

E

So Mike my current example matches what they're doing okay, but it's in that example. When you look at the example think yeah they're not quite sure, that's as clean as it could do and often is on their side. So it's it's only informative rather normative. Okay,.

B

But the for the chairs is there any action you want us to take to help to try to align things where you're happy. No.

E

I'm happy you see their choice. Okay, actually I'm I need to update you. They change their mind. Okay,.

E

I

J

Morning, Kent I'm, presenting yang bed extensions action, Martin put together these slides and just click through them quickly, there's only two issues.

J

The first is the naming issue, as you know, we're trying to migrate the rest, cocking Davis extension from the RFC 80 40, and, and so we were originally going to call this extension also yang beta, but I think some felt that using the same name was leading to confusion and a different name might be better, so some ideas were kicked around currently structure is out there I'm we're open to other ideas, as I know very minor detail point, but now is the time to throw out ideas. I think just simply beta would also be good.

J

B

Are you asking for objections or you're asking for concurrence.

J

Well, I'm actually yeah either. If anybody has any opinions, a strong opinion either way. But if there's no objections, then I guess we'll move forward with structure, because that's what's on this slide and there's no objection in the room, let it be noted and second and last issue. Actually this is a holdover from before, as as well Martin posted to the lists, and there was no response so we're here discussing it in the room.

J

So there's no agreement on the issue as to whether a structure must have a unique top-level name or not, and of the syntax for augment yang data to resolve this. We proposed on the mailing list that the structure would be encoded like any data that is with a container so, for example, if you're defining in your yang the structure, modifies description, datastore error info with a leaf reason, then you can see that the instance document would contain it's like a container. So it has the outer envelope modified.

J

Subscription data store error info, whereas that the current, if was the current yang data structure, you would just see reason as being on the outside. So this is the proposal. It's a it's actually, a pretty large divergence, I feel from the current gang data structure and in particular, I think many of you are aware that the zero touched raft is making use of that ability for the Duffy that the gang data, the name of the yang data itself, it's not showing up in the in the in the structure of the instance data document.

J

So if we were to do this, I'd have to update the zero to strap to accommodate it. But beyond that minor detail, it's probably well. I actually feels probably better long term approach better to be explicit to have the had the name of the yang data or structure show up in the instance document directly as opposed to it. Not any.

J

Are there any objections to to this approach.

E

Comments, Rob Wilson Cisco, so my only comment is whether that's going to make it not quotes not what's so nice to read effectively in the instance data, so I think we used to see in context how yang and it's gonna be used and whether that will actually mean that when you like augment extra data, I come exactly where, as some use cases of it, but I wonder whether this is going to make that more ugly. In terms of like the messages you're seeing whereas before it wouldn't start, smiling concern I. Think the.

J

If the insta-tox mean you get is probably I identical in the sense that the the Arthur at the game designer can construct it's still before, they would have to do themselves put a container if they really wanted it to show up a you know a container with yang RS cocking data. It would be the outermost of a container, followed by whatever you want inside and now you don't have to put that outermost container. It's implicit. My question is so, but with augment yang data, I think that's what you were saying now.

J

You need to reference this and how would that show up in the x-pac.

E

It was more than you're saying you get the container for free now or what are you doing well that container to be there at all it's my question, I think to give a years case, for when the container would not be needed at all, I'll need to look uneasy. Look at the draft and have a look so might be think we're different use case. So.

B

My some misunderstanding at the value I see here is, is that you can augment the what's now call the structure before you couldn't augment the structure you get all the augment the instance right. I think.

J

That's part of it, but it also this gives you so there's a there's, a concern that the for, how should I, say, ambiguity showing up if there's, multiple top-level structures, having the same name and with the prior approach, it's actually possible that two different yang designers may accidentally I think through an augmentation, have the same name show up in the same namespace and that was causing a problem. But but with this it no longer you it teases apart. The ambiguous ambiguity is disappeared, so I think both both were motivations.

J

So Rob you said you would want to see more, so I think you would like to take this to the list and have it um I'll.

E

Have a look at the draft and then I'll like onto this on the list: if I have an issue it or not, okay, I try to remember all the struck the background of the history. Wasn't this issue I think: was it Andy raising it that you had this days of space? Yes and that actually he checked over because I know Andy's lunch everywhere.

D

E

Yeah I guess I wasn't necessarily and less important is concerned, as inland I thought realistically, would this really be an issue or not and I'd wonder where this is going to be harder to use in some way? That's my concerns. It.

K

Rashad from Cisco sorry I haven't read the draft in more than a few weeks, but maybe a question for Rob I, don't understand his question, because if you don't have the outer container, you don't have a context. How do you know what in what context? Who is because I think Rob you are questioning whether that outer container is needed right.

E

It might be I'm confused with the old man case, the sir big case, where it feels like you, don't want this container necessary to be there if.

J

You recall in 1848 the description statement, forgetting data said that the it had to resolve to a single container or a container, and- and so this actually codify that you remember in the co2, strap kind of worked around that language and said well, it does resolve to a container, it's just a choice, a choice of containers, and but this actually would remove that ability to even do the choice of demeanors. You would actually you know so now you really just have that top level named structure.

J

It just makes it a little bit more explicit, I think what was intended by the RFC 80/40 description statement.

L

J

Think I'll go and read the draft before asking more okay. Well, certainly with both leaves, will confirm them on the list, and that is the end of the presentation. Thank you.

M

One comment: Kent's presentation, the.

L

Yankee instance.

M

Data draft starts with gang date that container whatever yeah. We are happy.

F

Yeah I'm sorry ever get larger.

M

And instance late: the draft bow is adopted and I'm very happy to see that.

M

So, just very quick recap: we need a young instance data offline, sometimes because maybe we did the data already in design time before any device. They actually implemented it, or maybe we need data that would come from a gang server like you, yang the library, but you don't have the relevant network notes available. If you think about someone designing a management system that can handle ten different types of notes, you will not have all ten notes available and configured correctly, so you need instance, data, sometimes.

M

These are the some of the use case, and these are just use cases. They are not the normative in any way that was clarified in the Lakers draft.

M

Most important is documenting server capabilities there in the net comp there's already a draft about documenting notification capabilities. That's rather adopted by net concept is already using this concept, but some companies, including mine, is already using such instance. Data for pre loading default configurations, for example, access control. You need to you, could preload access control rules which part of the big day the complete data leave four available for different groups and roles, and there were some other use cases as well, which I I'm sorry not to mention here. Yes,.

N

Deluca I think that this document goes into quite some length explaining all these use cases, but I believe there are other use cases.

N

For example, we use instance data in rfcs as as examples we often validate instance, data with tools, which means they must be in some kind of file or konradin to different formats, so I believe it's not necessary to provide more and more use cases and because I think that it's quite undisputable, that instance, data can exist on their own, so I would suggest to shorten this pattern simply mention some some examples, because I don't think that anybody is arguing with the need of of having some and the fact is, for example, in our FCS we had to use with XML examples.

N

We had to use some get replies, or things like that which is sort of stupid. So I think that it's it's quite clear that something like this is needed, and there is no need to have several pages of use cases or something earlier.

M

There has been some debate now, I'm happy it's over and I'll try to look into shortening this while at least listing the use cases I think is useful, even if, in a shorter format,.

N

I also think that, in fact, in the light of the discussion of young data, I believe that there is no reason why we cannot have just a standalone instance data that do not belong to any server or whatever. And then this is a use case just for using yang without any yang data extension directly on on validating such that use cases, because in this case, I don't see any reason for for doing this. Young data referent having a separate augment statement and think extension and things.

M

Like this sorry, what do you mean by stand alone, so I think you needed a.

N

Yang model to define.

M

N

Some data, like the yang data extension, is trying to address, and this is just some data that I need to validate no.

M

Because, yes, if you think about an example in RFC, yes, you are right. If you think about that's a pre-loaded set of configuration of access control rules, then you want to say that this is the exact set of data and I need meta data connected to this.

M

Actually, this set of data so that I can identify when it's updated when it's who's, the author and a few other things yeah. This.

N

M

N

This drop so maybe.

N

O

Haitian jet I'm learning well I'll agree with her on that. We may not need more examples. I think it's important to understand where all the instance data can be used not just in a part of RFC, but also in how it gets packaged and distributed along with disk began definition itself. So, yes, I would agree that we may not need more examples that we need to at least understand where all the onion streams data can be used.

M

But I think it's definitely at least worth mentioning each of the use case. So if I left out an use case, then I'm sorry for it and then please tell me.

M

M

So the basic idea is unchanged: reuse, existing formats from net comp and rest, confer a get operation or request. Actually the reply for the get and decorate it with instance, data which might not be needed if you are just given showing an example in a IPF draft or idea RC, but often for the other use cases. It is very important to have this instance metadata added.

M

We have the two formats and there is a need, some time to document both config or true and config force data. We are thinking about the C camp alarm draft there. They will have a list of alarm alarm types that an network node can send, which is read only but documenting it and potentially using it for populating the server with that data. If they're still important today to document read-only data and yes bar partial data sets- are also burned.

N

Again, I think that in some cases you can have a data model that contains both configuration and stated, and in some cases you just want to provide configuration, for example, as I understand it. This is not possible in your approach. I think that it should be I, don't know. Why do you think it's not possible because then you can run into if some stay data somewhere is is mandatory and it's not it's not contained in the instant data that you provide, because you just want to be configuration then it can be reflect invalid. So.

M

Sorry, no partial data sets are allowed and in the draft it is stated that that means that any mandatory mineral amounts constraints might be violated. We're allowed to be violated. So if you want them, you don't provide the mandatory part, so be it configuration or or state I.

N

Will probably ask in the mailing list and I can.

M

Please have a look at this data. I think three things I'm.

N

Talking about something a little bit different, it's not about minimum.

M

Let's talk about it, but.

N

M

It should the plan is to avoid configuration and and state, and if you want oh, then you can provide only configuration or only state or that's a full configuration with 30% of the mandatory state yeah.

M

So the draft was adopted by the workgroup and the adoption was rather than relatively close to the closing date for the draft submission and there were some national holidays in hungry, so I missed that deadline, but now it is submitted but I think today it's pending or the workgroup chairs with sexually very late submissions, I think.

B

M

You very much yeah.

B

M

Yeah, it was my fault, but then I needed your help now, anyway. The last form of that last, the private draft that was stored, is really the same as what what will be stored as the first version of the net and the workgroup draft. The name of the draft has changed so that it really experts to criticize that this is about the file format. This is thought about how to use it or and mandating any of the use cases.

M

I very much hope that some of them will be mandated or recommended, and that I assume and maybe I'll even work on it- that there will be an that they're connected draft or that written beside describes how to use this. The document server capabilities actually yeah the net cough notification capabilities. One is one such draft.

M

There have been some changes since the last IETF clarified. The scope is the most important and we added one parameter for the to tell and define the data store where this data is intended to go. That might not always be necessary, so it's an optional parameter. Let's say if it's paid data, then it's trivial that it goes into operational or interstate. They take you on on an MVA, so we don't depend on an MBA at this point.

M

At this point so open issues, some people stated that we might need more metadata, potentially metadata for each target yang module and this terminology target yang module. So the yang module that defines the structure and content of the instance data the real data that is put into inside this wrapper.

M

It depends on what kind of additional metadata we need. We might. We might need this birth target module metadata. The big question was: how do you define which version of the target yang modules you are using, which does version of the target yang modules? Is this set, of instance, data good for- and there is no good answer for this, because the versioning theme has not finished its work. Yet definitely is just saying that revision date X, that's too strict. It will be up the modules will.

M

The update is for a number of reasons, taking the most easiest, just waz Mattox, and we still don't want the instance data to be updated for every of these, so my proposal is indicating, in the instance, data the revision that was used when well the instance data was last updated. If you combine that with that ability, information for the taja yang modules, then you have a very good view of where and this could be used, but giving anything more specific I, don't think, is possible. At this point.

J

Kent as a contributor, and how would you do that to indicate which were very nauseous if.

M

You first open issues have a list of target modules which would not be needed that much because namespace is already dictated. But if you want to include the revision date of the taja tiang modules that include list of taja yin-yang modules in the metadata, including their revision date, I, don't see any better solution than that.

M

The neck next question came up: yes, but the whether this instance data set, or instance, data file can be used might depend on the features implemented by a specific implementation specific on network device. Yes, that is true, but then documenting revisions features deviations.

M

We are getting into a lot of details and we don't want to I just say: don't want to replicate the full yang library inside every piece. Every instance data file I see that that you might try to say that use it together with that or you rather say that if you get a bit more instance data than you bargained for, because you don't support some of the features then follow the compatibility rules and ignore that yes, vogelson.

E

Cisco so I think miami-dade. Why raising someone else I think my aim here is to be able to say any service routes of loading instance data document and have enough context in that document itself that it knows what schema to uses and what features need to be there. So that's that's what my aim is so that you don't have to have extra metadata information around the instance. Data to know be able to interpret that I.

M

See but I think revision and the comb I hope very vision and the compatibility rules, meaning that you ignore whatever data is there but ya.

L

M

Know it should be now if we say that that's not enough, then I would say then rather say always say that for every instance, data file you need associated another instance, data file about Ganga library and reference that so don't reproduce. The complete yang library information in every instance, data files, yeah.

E

I, don't do you need to, but and again I'm, not sure it features whether you need to or not you just assume that all features are available, maybe might be sufficient. It's just that's. Thus the aim is to say look, I don't want. Somebody else had to say all to interpret this file. I need to have this extra information. I.

M

Assume a yes and I assume at this point that you can write your instance. Data for all features supported and if some feature is not supported actually on the network device, then compatible compatible, ignoring or discarding of the additional data will work yeah and there was another requirement that metadata. Yes, we defined the Draenei set of metadata, but there might be other parts needed, so augmenting must be possible and I hope that yeah, Ken and Martin will solve that problem with yang data or yang structure. Draft.

M

E

I had one other comment actually going back to the issue. The latter raised. I actually agree with Lada. That would be better to have the use cases section per se. Shorter I think that it's still quite detailed for the individual use cases and I'd, rather those either move into the appendix or it just says, look these are things and how in Stata could be used without going into much detail as if you write another draft on server capabilities, you've now got text into places describe what those are.

E

The conflict potentially I, sort of feel that if we start our draft, then that text move there. You just say this is an example of how it can be used without having so much detail. So don't think it's key to the document. I think the key thing is really. How do you actually document instance? Data I can.

M

Make it throw it through I can make it shorter and the other question.

M

Thank you for you.

B

Yeah, so just so, you know, what's happening with the keyboard. Doesn't work on the laptop, so we're sort of stuck with the view that it's in.

P

So one of the benefits are major in one of the benefits of doing what the chairs are asking submitting. The slides in time is that the slides are out of date, so this is our fairly simple document on handling long lines in artwork, and it's now a working group draft.

P

That's just just leave it right.

P

Yes, brilliant this, so I can't read any of the slides. If that's too small and that's yeah.

D

P

Of stencil I tell you what how many people are watching online. Put your hand up if you're watching online right so I upon me for a couple of slides going over of what's going on that to set context, I think that's: okay, because we've just adopted.

P

So what we know is the the format of drafts is a 72 character. Limit line feeds sometimes have meaning in particular when they show up in XML and JSON. They actually have semantics, and that combines with the desire to include examples in our documents where those examples can be automatically validated and not confused users who try and sort of cut and paste the examples into real use.

P

That means that we have to do something to hand or line feeds that are necessary in our examples, and we've actually got some examples that are in our documents that have gone out where people have had their way around this or written special cases and I know from bitter experience. A lot of time is spent trying to construct examples that actually don't overlap the 72. So the other thing we need to know is that lots and lots of different tools and development approaches are used by different people.

P

So there's never one size fits all and if we scroll forward, so what we're trying to do is come up with a notation, a current, a way of handling these examples with line rats, so that there's not confusion and we'd like to be able to automatically rap lines so that you can write your example in one nice legitimate piece of code and give it to a tool that will wrap it.

P

But we also want to support creative people, who'd like to do the rats themselves and using indentation, and we also want to support Native character, escape characters so roll on what we've done since last time at ITF 102, we were at zero. Six of Kent's individual draft wrote these slides at zero, eight, which was the version that was adopted. What we've changed?

P

We clarified the scope a little bit, it used to say internet drafts and, of course it should apply to rfcs as well, and we've clarified that the primary intent is for formatting examples and code. You might use it for other things, but I think if you started to use it for tables and for real figures, you would cause confusion.

P

We we clarified that we are coding in Pascal and not in C, so the columns count from one this is caused us immense stress.

P

We changed the header line format it used to be it used. The natural format of the header line used to instruct you how to wrap by the length of that line, and that was getting a little confusing.

P

So we've gone to an explicit value for that. We've moved to a a slash at the end of the line and a slash at the start of the rap line, which I think was Rob's idea. So thank you and that that's quite nice and the chairs I'm working out how to do the slides. This is cool.

B

Remember we only have hands hey.

L

If I really know, we did.

B

Figure out how to go full screen, they can make it out I've.

P

Only had a working group secretary to do it all for you.

P

What else did we do? Oh, we clarified tabs.

P

We added some text to separate the manual and automated unfolding. We put in a forward reference to XML to RFC version three, because that's got a source code tag as well as the artwork tag that we currently used, and we beefed up the examples.

P

So that the current solution is pretty simple, when you fold automatically you get to the line length that you want, you insert backslash newline and a backslash as a start of the next line. So automatically unfolding is really easy. You get to the line end of the line, find a backslash find a backslash on the start of the next line, and you can just strip out the characters and off you go because somebody might have done automatic in certain manual indentation. You also strip out any white space that you find before the continuation backslash.

P

There are tricky cases that that arise, and what we have said is that, if you're doing automatic, folding and you've come across a tricky case, it is best just to throw an error and say I can't fold this, rather than we make this solution really complicated and handle all of those cases.

P

Tabs, as I said, we've clarified this vertical tabs probably will never see them. But if we do, we fail to unfold to fail to auto fold.

P

There was lots and lots of dis about what to do with horizontal tabs, and we ended up with this text, which basically says note that you're not supposed to put tabs in artwork and source code anyway and when you're automatically folding. Usually you must either convert the tab into spaces or throw an error.

P

We said, we've stated there. Other use cases are not out of scope, so clearly we're net mod. We worked on on young stuff. There probably are other use cases where it's useful or to all sorts of code like things and we we do not rule them out and we hope that we are applicable to them, but our focus is clearly within the working group.

P

We have a few small, open questions. Some of the examples I think are a bit clunky, but that, if that's the biggest problem we have where we're doing just fine and the authors will work to polish them. I have a question: there's a scripting in the appendix sound and I believe that somebody's pointed out an issue to you Kent. So in result it's been resolved. Well, there you go, Kent is so efficient.

P

The precise format of the header line might need looking at, but it's that's again a really trivial thing. We had some discussion about how to tell an auto forwarding tour, how to wrap and currently we we default at 69, and we have a minus C flag that allows you to set a different value. I think I think we're probably happy with that, but we just want to check at the moment. If you were Auto folding, you would do no indent at all.

P

We could add a flag that also tells you how to indent, if anybody's got a strong feeling on that, we might otherwise not, and we possibly need to talk more to the XML to RFC version. 3 team so yeah the poll ended we got adopted, we need to tidy and polish.

P

The authors think that we've been around this loop long enough and actually we'd, probably move forward to a working group last call pretty fast, so I'd encourage you all to be looking at it now, so we can get it right before that I proposed that we should bounce this off some of the other working groups that we know do a lot of Yang just so that they're all aware of what's happening since they're. The people that write young models, perhaps their input, would be handy, Robert.

E

Robertson Cisco, so one question I have, is you have a mark at the beginning of the folder data to say the photo that is following there's no marker at the end of the folded data. So how does that all know when it finishes where it's gonna finish? Is that obvious.

P

Well, so yes, and no so when you, when you hit a full deed, not good point yeah, so yeah in the source of course, you'd hit the end of the artwork or sauce carry tagged so yeah. That would be done. But yeah, that's a good point. Actually, when you don't want to plow on trying to unfold all the because.

E

P

Might have blank lines.

E

And they don't know in the actual data, so don't use black lines. The two stops I think you might need it can't.

J

As a contributor and co-author, but we we kept that explicitly out of scope, the framing of when the end of the artwork is. We took it out of scope for this. It was discussed in the list. How does it work? Well.

E

J

It's the xml-based. If you're purchasing XML, you can just extract the so.

P

Ken no he's talking about the finished document, so you've got a text file in front of you. Yeah.

Q

Angles: you're you're.

P

Trying to work out how to unfold I know.

J

There's two answers: if you're looking at plain text, there's one answer we there, which is that we don't support it and we took a scope if you're looking at the XML, you can extract the netted directly to.

E

The that doesn't make sense, so you can't unfold it's in a text. Humans, I'm holding, doesn't work that doesn't matter, but.

J

It's not our problem self, it's a different problem. There are other artworks that are not folded and they.

E

Also, don't have.

F

Indesign one taxes.

J

It's not this problem. Okay,.

F

Yeah so I think there's a there's a place here where the some intersects, with how the ITF sees text which is text, is able in in the in the modern structure text is rendered output. So it's not so much. That um text exists as the a structure for it for anything beyond the convenience of the reader right and you actually sort of have to reach back towards the thing that actually has the structure in it, which is not the text document to to to imply anything other than that this is rendered output.

P

So I agree with robe and disagree with to the three chairs.

F

P

I'm reading the text file I reached the thing that says what you now have is wrapped text and I, and this is how you interpret the slashes and then I get to the end of that piece of wrapped text and I'm in normal text and I counter. Some slashes I will interpret that to mean that I have to in that that it has been folded, and so, if I was a chair at this point, I'd say: let's take this to the list.

A

B

Have three minutes more to talk on you.

P

Don't want to hear me talk for three minutes more.

J

Kent as a concealer just to add to that in XML to RFC v3, if you use the source code tag, it will automatically insert the code. Begins code ends framing around our work. That's rendered in plain text format. So again, this draft doesn't need to add any additional framing for the end of artwork. So.

P

Certainly the the begin, the begin code end code would handle the case weirdo.

B

Get it is this inviscid? As you said, this is a good topic for the list. I think it's a very reasonable comment and clearly we have some differences of opinion that we'll explore and see if we can get to some consensus and and then agree that Kent is wrong, but can.

E

I, just give one other example where this came up was I've quite eyes, but to use this for the tree output effects of the Yankee output and not necessarily generator automatically I, have pie airing or p.m. when it generates it bill to put this stuff in to wrap the lines and again there you wouldn't have the source code. Encode normally just put those snippets in inline effectively so I'm another case where it helped.

C

J

G

Waits, oh hello, everybody, so this concerns the comparison of an MVA dinosaur's draft. This is also a draft that went from that since last IDF has been adopted with the working group draft we've only very minor updates on this. So again missing. Just as a recap, this defines an RPC that allows to compare an MVA data source to report only the differences without meaning to about the entire data source and yeah. We just posted the the the working group version of this, which basing is pretty much pretty posting up your individual and crap there.

G

One of the items that was discussed on the list they are not yet reflected on. The draft is to add support for the possibility of different encoding schemes, use different encodings to report back the differences, and basically the agreement is that there will essentially be a leak that will allow for specify which encoding it is to be used.

G

So, just as a sorry.

N

Christian, what do you mean by an encoding just chase some XML or just the difficile format, the dip format so.

G

It's all the deform and gets reported back, because I mean busy. The the thing is: a is a yang model, a yang, our pcs are.

L

Really basically,.

G

Whichever whatever that contrast, um you use that.

N

Suggests, in order to avoid confusion to use another word than encoding, because this is written.

L

N

Way and actually I would also support in the future, may be developing some like more more interesting format of the diff, which is something what can suggested I believe in the mailing list, because in some cases, what you have here, the yang edge it can help you to transform some data into something else. But, for example, if a user wants, if a client needs to decide whether to use the original version or the new version and maybe switch between them, then it's. This is pretty much useless. So.

G

We get to it in a moment actually on the different point, but clearly the first item is basically to allow to have different possible incoming that's the first thing and then the second one is that you, which one off basically manatee mandated essentially in this trap we get to this in one moment. So if we can just be further for other questions until these, as chair generally I.

B

Try to suggest that we avoid options that are not strictly necessary or have a strong use case. So if the problem is, is that as a lot I was talking about, we have a format- that's not really suitable for our general use cases. Then we should fix that.

B

If the problem is, is that we really have different options and we want to allow for that than sure different types is fine, but we should make sure we have a strong use case for different options, because if, if we really have these options, that leads to interoperability issues, we.

L

B

One implementation that does one thing and other implications does another, and if one's the client wants a server, they can't talk to each other. So we only want to go to those that option if it is a strong justification for it. So.

G

That's a crux of dishes are actually let me just jump forward before going back in, since we are just on this say sure. So, basically, right now or the ref specifies is used, is to use the nth moment essentially as this, but this does not include values from both data stores, which would, however, sometimes be useful.

G

Why there's a basic issue that was raised on the on the list and then therefore he won in addition to just support you option to have your own could add a new form provider format, there's a quick, but they should be an additional format we define. There are basically two suggestions and I guess the first one was to define additional format that is not based on yang patch.

G

If such an addition format is defined, of course, they have some additional follow-on question that will need to be answered subsequently, namely which form it should be the default if any, but because we will have gang patch, presumably and this and is their format that is mandatory to implement and facing again this the whole issue about, inter probability versus flexibility and implementation costs. It comes into play here with this. With regard to that to to that argument, and of course the question also well of this format, opening should be part of this draft.

G

So then was the first suggestion since then, it was also a second suggestion to simply basically actually augment basically augment the yang patch edit list, with an old value, any data and essentially address the fall butthey having a better yang patch format. If you will and then use this simply as a one that is being defined in this trap here and be providing the augmentations there, this does appear to be the preferable solution, but obviously that is something that that we need to discuss this very basic display.

G

The reason I say it may be the preferable solution is because it ties very well with the existing yang patch and because it avoids having multiple formats and support to find it. As.

J

This drought, but.

G

Anyway, there's this one thing: where what we should get some feedback comments and.

J

Yeah I can't, as a contributor sure, I'd say it's preferable and II threw it out as the suggestion to make. You know how we could send yang, patching and I said that it was on the path towards another direction that we might want to go towards, but otherwise I don't think. We really had any sort of important group consensus around that and then, as chaired sort of echoing what Lou said or building apartment, I.

J

Think the motivation for the previous issue supporting multiple formats Phil's motivation for that when he posted to list was because he perceived that the current format wasn't really a deaf. It was like what Lada just said: it was actually just transforming from one to the other. Yes,.

A

And what were their dips.

J

Yes, yeah and so Phil's was defensively, thinking that well, this means we'll have to do something later on. So that's my lesson sure we have the abilities or multiple but to lose point. If we actually pick the right thing now, we don't have to have multiple. The first issue disappears completely. You can just do the right thing now and so for this opportunity.

J

G

Is a third case, though, maybe knows a responder that yeah that we would have to define multiple formats in this trap. Of course, the question would still be. We might still maintain keep the hook that would allow somebody to. We could define just one format here in this draft and there's baby the one that mandatory but keep the hook. It allows somebody to find their own additional provider from it if they wanted to do with this additional leap, the bus that was discussed actually on this one.

G

Maybe you can just very briefly go to the previous slide that which maybe shows the the model structure and/or the format string that was discussed. This is basically whoops. Sorry.

G

This thing in red there Betty this this little format leaf that would basically guide this up there. Okay.

G

Cantaloup you're making no.

B

Finish your point: okay,.

G

All right anyway, so that's I, think this is where we are at. This is really basically the the main issue that needs resolving at this point other than that I think baby. What is in the graph? It's it's, it's fairly straightforward! It's not a it's, not a big draft and obviously yeah I, think having Tilly actually having one format and not a selection of formats is I would also the least thing is clearly preferable because we want to achieve interoperability with this.

N

Value again, I would say there are some advantages to really permitting different formats. One is that this work can be can be delivered, at least with some way of doing this death, which already exists. Otherwise you probably have to wait until this mule death format will have been developed, which, which can take some time. I, guess and second thing- is that we may not get the new D format right just immediately, and then there might be people that develop a smarter format later and then it can be just exchange for it.

N

So that'sthat's I would say more flexible and I don't see any problem. If, if T, the format can be somehow signaled using some capability or some playlists, and it.

L

N

Be it should be OK for for for both parts to agree on something? Ok,.

J

Can't as a contributor, so I I agree with your second point. Actually, if it you're right, the DIF format we select may not be optimal and it's good to have a way to at least version right. If we had versioning of the yang patching v1, which could be one version, v2 could be that's a different way of doing it, as opposed to just having one version that internally had an ability to select different format, so I think having an out of being able to support a different version later.

J

It's a good idea for the first issue, though, that you raise I'm just wondering about the urgency of this say: I do question that I mean I, understand that nmda came out and it's a good idea to have a spiffing support. But is there really urgent see that this matters? If we did, if we're delayed by an extra quarter or two I mean? Can anyone speak to this.

N

Oh yeah, we have been discussing some in in relation to my previous draft about West Conference actions. We have been discussing some kaitos staging repository and this before Matt ordered if this capability would be useful for this. So of course this saying that after big a store is still sort of remote thing at the moment, but I don't know. So are you suggesting that this draft is is written without any any any concrete option for specifying that? But if or all that, we just right away this work until the the diff format is develop.

N

Iii I think it can take several years, maybe two, because people will have different, which is of course we know this, and everybody knows what a diff is so I expect quite heated discussions and so on in in the plain UNIX diff. We have several different formats that can be somehow regulated in this as well, so I'm a bit afraid afraid it can take a long time. So at least we can do some some work, and if we have some pluggable before mat, then we can work on that later and still.

B

N

B

So I heard what I heard Kent suggests is that we try to come up with the right format and then, if that seems to be taking too long or we can't reach agreement, then fall back to what's what we have available. If maybe I read a little too much into what you were saying but I think that's a sensible course is: let's try to do the optimal. The optimal is a single solution. Let's try that and if we can't get there then fall back. I would personally with chair.

B

Have any hat I would be opposed to saying that here's this one mandatory format, but you can do whatever you want late I would say: here's this one mandatory format and that's it because we already have a way of extending models. There's augmentations! If we come out and find something better, we find another one that could be done. That extension can be done as an augmentation, but I think we should start out with one format whatever that one format is.

B

If we can, if we can define what we agree as a group would be a right format, let's use that one. If we can't, let's use patch, but let's at least try to get the right one, yes,.

N

So shall we set a deadline on how long it can take to develop this format and if we don't succeed and me so the backup to do this, because I think it can still take quite.

B

So, according to the authors and the working group, don't come up with a proposal we'll have our answer that it creates patch. If the authors come forward to the pros Oh or someone in the working group, it's a working group document. Anyone can contribute if we come up with a proposal and we have some good agreement on it. It could be very quick, but we really have to see what the what kind of contributions we get in and yeah we could revisit it in in one if' and maybe at the next IETF.

B

We have a great solution, a great proposal that we're all happy with, or maybe it's obvious that we have to stick with patch can't.

J

As a computer, or actually as chair sorry I'm, just adding on to that I think, there's a caveat that there has to be good-faith, effort right. They just didn't run out the clock. I mean there's, you know certain amount of time until 104 I mean actually has to be discussions on list. If there's no discussions and we get to 104 and we're just having the same, you know that's not fair right.

F

Yeah Joel yeagley um I mean it's sounds like we have a milestone, but like all milestones, it's essentially aspirational and if it looks like we're not actually on the way to achieving it, then we probably need to throw it out and do something else. At that point,.

B

Great I think I think that's reasonable. If we got great proposals between now I'm 104, we'll see where we are with that, if we don't we'll, probably go down the path of just using so.

K

Rashad Cisco question to do when you were saying we find the ideal format. If we do do, we still go with the format leaf or you saying we do the augment later on, if needed. If there's a better format.

B

K

B

A big fan of we in a standard putting a opening the door for a future option when we don't know what that future option is. If we, if we knew, we had the future option, and/or multiple formats, and we think that we need it, we have a good use case. Absolutely do it, but if we.

K

B

We, if we don't, they can worry about it later. We build. There are mechanisms so they're extensible, so we can do that extending life so.

G

You just look there so you're, basically asking the same theory that the earlier suggestion to add this format leap. We should essentially remove that I would miss this. This decided in here on this thing, so baby we've be bidding just go back on Muffy I.

B

Think you should have one format. What the one format is we have to work through. It.

G

One moment and not the extension of Hercules all right.

B

Okay, correct now.

G

B

All sorts of back, you know alternatives that may come up when we find that we really do need multiple formats. That group says absolutely need them will support that. There's alternate paths we can head down. We have to see where we go and what kind of contributions we get.

E

Robinson Cisco I, don't talk about formats at all. I'm I would like you to maybe add in filtering by origin metadata as one things you could filter on, and the specific use case that I'm concerned about is that the as its described now that big he provide is a raw diff between two dates stores filtered by the schema between those two data source.

E

But if you had, for example, configurations coming in dynamically DHCP addresses that sort of thing, then they'll always flag up as being a difference that you may not be interested in you may actually be wanting to know is what a might of the configuration have I put in by running or intended hasn't been applied. So I think if you filter by origin metadata.

E

May then they'll talk to that question, or at least look B at answer that question of what configuration I put in is not yet been implied, because that might be what you're? Looking for? Okay.

G

Yeah there was actually the item was one of the most motivating items back to C, so I could even different dozens of also arise from okay, yeah point by taking.

E

Any other comments, so just clarify, say: I wasn't requested so I'm, not asking for the origin, mixed data and the output to know where the changes come from. I'm asking you on the input to say: I, don't see these changes, I, don't see the differences that are due to learned configuration I wanted to see ones that come from intended, so so like a filtering program to rather than necessary now for one, but that might be different question as well. All.

G

J

As a contributor I think, the output format should include a version number so that if we will do ever version, the the patch format I'm. Sorry at the diff format that and someone were to obtain it out of a tab and they know which version of the dip formatted is looking the version number they had.

G

J

G

Itself, yeah, why would we need it if we, if you mandate, if you say one format, is mandatory I, don't and we don't be another coding, a hook to a tool to extend this avoid I. Don't understand why we would require that we.

J

Can take this to this okay.

J

Thank you. Thanks.

G

Okay, good, so this next okay.

B

G

I can just say yeah, okay, so the next thing is in update this video, not a working group draft but I guess maybe moving to the individual drafts now, and this concerns the smart filters for push updates. I can title is probably a lot of problem statement. It is not filter for push updates.

G

This we posted an update to this draft is one of the dress which have been somewhat basically in a little bit dormant stage by the way we're working through the yang push sweet, subscribe notification support tool to get to, but as those obviously nearing the completion, it's time to do like taking up some of this other variant. So again, as a as a recap and gang push, we provide filters that allow clients to select which nodes so to subscribe tool.

G

However, the filtering that is based on values of those objects is currently not covered in yang push. However, many of monitoring applications are actually based on observing values, so they would want to know points okay, I'm interested in this only if one, since it is in a certain range or above a certain threshold and so forth.

G

The reason to not include filtering based on values and yang Bush was we did not want to stretch implementation complexity further and push is already a fairly big draft, so we felt that putting these types of advanced capabilities in there would yeah we're busy unduly overboard this or this especially follow-up follow-up verb.

G

Also, actually, when going into values, one comes actually in looking at the use case. There is also clear that you get very quickly actually on a path that leads you to also requiring potentially state, because one points me the regards to to lose being interested in in values above a certain range between this leads very. This leads down the path of what supporting the fish or crossing regards, but you would want to be notified once the threshold has been breached, and then you don't want to keep getting notified. You want to only get notified.

G

Perhaps if it falls below out of all- and I still has the reasons threshold- which again would require you to maintain state and again, this would add complexity anyway. So so there is a gap. If you want to have this type of functionality and this map filters drop. Basically, it tries to address this gap in terms of updates. So, as mentioned busy, the graph has been there for some time, but basically one thing that we did add actually pretty close to general Walker, who actually added this part, is to define the first set of filters.

G

Those are value filters that compare compare objects against the using an operator, so greater than great, listen and so forth, and as many of the filter is then to push the update of an object only busy if the value of Beijing matches the comparison so busy if it passes the filter, it's not yet, if spread short crossing alert in the sense that there's no state it doesn't maintain. If you have the next push update in the next period and it matches and again the third cost, you would pity we send.

G

We send this update and you don't have it at all or any things of that nature, so in terms of basically taxonomy that we also defined earlier. Basically, the the filter that has been defined now. I has been added this between the first item. The first bullet I can filter based on and based on values and the sake for filters aggregates of time of the other items and so forth.

G

Those are basically still TBD so in terms of next steps yeah, but he so clearly be the so this is the very first sir printers we want to grow the supported, filter, pipe Ubuntu and stay also. Stateful filters specifically could deal with the pressure of motion alerts, so the other type of reporting, okay, also, the high water, no watermarks and again assess interest of the working group that we should greatly go ahead to define a solution.

G

Falls problem, there was interest in the past again, given the yang focus on yang push getting those things out, we waited there a little bit but keep in the state of readiness. Now it may be time to take this up again and perhaps once these other drafts have gone through the shepherding reviews and I nice G, then maybe at that point in time we may also ask for adoption absolutely.

G

R

Yes, Michael from Holly, okay, I have real document and I. Think I have several question for sizing. This is useful, but I. Do you think you need to define some more generic method to support different mechanism? Not just a limited. You know young push that maybe it's more useful. You mean I mean different. Some generic master to force a smart filter.

G

Well so I think baby well, I guess, for it comes what you could go, Falls, really the two categories, one thing, basically, where you have updates- and you want to do something based on the values of those things so you're not and I, think that one fits actually very naturally with yang Porsche. It is an additional capability that builds on top of that.

G

However, your point, a key thing where your point comes in is when you are not just filtering in terms of passing, something or or blocking something which is what a filter does, but potentially aggregating this information to more things and clearly busy. There were also the question: okay, should we have things that that define in average sure or or that that sum up things across a across the list? It's about? There are crazy, each really item that you could take there. Those things go actually a little bit beyond yang portion, yeah I would agree.

G

That is actually one one question to whether those things should be defined here and perhaps not because one thing that is exactly in in in gray and in sntp MIPS. There was this concept of Arman. Many people here will remember solve Bailey. That would not be a substitution of Arman. Alright, somebody in Arman, you would have busy these these aggregates the top ten lists. Whatever it's about you, you could define of different things, so we would not go they do that direction. That would go so for those types of filters or pre-processing.

G

It's of all about have you. There would be outside the scope of this left. Hey.

R

You mean you'd like the most clear scope. Okay, another thing: okay, I think me there's some kind of work we proposed even man German the unit module and the way also some blog a maybe similar to his walks and to find some filters and reference, some males definition and I think. Maybe we can discuss more and more.

L

R

And maybe improve each other and church fun summer, 800, each other okay. So.

B

I think you have some slides on a related and you're actually up next yeah.

R

B

You're setting yourself up, and so maybe you'd be good to switch over to that, and you can talk to not only with the slides you have prepared about, but how you might generalize them to support this case as well, and then we have another document after that. That also might be related depending on where the authors want to take it. So it seems like it might be to consolidate discussions before we move on to that. Let's just see how many people in the room have read this draft the current draft.

B

So if you've read this draft so that that's a few, but you know not not a huge amount, we'll ask the same thing of the next two drafts: Michael don't go away.

B

If it's quick ask if not we'd like to maybe have more discussion once we've heard the next two drafts I.

E

Just guess it's a couple, quick comments, I don't know so one of them means I quite like this draft I, like the idea of filtering the in a stateless way and for the stateful filters, I wonder where those things aren't better putting into the data model directly and handled separately. I think you gave some examples. Other cases like Arman, for example, I.

L

E

For the things are those matter maybes best for that into the data model that having generic mechanism the last one is I wonder whether if this does progress, whether it should be done in net model Netcom, because that's where the other drafts are.

G

M

Think, as author of the Netcom camera or notification capabilities drafts, I was wondering your idea is that let's say someone who implements these filters: will they implement it for everything or will they implement it? For everything that supports push change or where is the subset of that? Well,.

G

So the well the should.

B

G

Into the cave, okay, you know your question: basing does it impact the capabilities? Okay, I'm, not sure I think this is I mean if you mule is really busy, is an additional filter that you would apply on top of yang portion. This is bathing in where this is position now.

L

G

You would you would basically apply the filter to the output from the yang push and maintain the things on that face off. From from that perspective, it should be applicable across all so buddy having differentiated capabilities to say, I support the filter for object a but not for a big B I. Don't think that it would be required here. Maybe we go into some of the stateful stuff might be different. Sorry about some people, the stateless filters also busy the first super filter is defined.

G

Now I would think this is you know that would just be generally applicable. Let's pick this up on the list, all right.

R

Good morning working groups, tasted Michael from Hawaii and today, I want to introduce John Titor model for even amendment. Ok, why we want to propose this document is because we want to support some of provide some maximum to support closed loop at make auto mission, management of data collection and Network automation. Our dealers to monitoring young instance data, and if this incident data have medicine configured condition and then it can trigger some action.

R

For example, a single notification out, you know, do some reconfiguration, so he proposes document and defined under the model for an abstract, even measurement and provided ability to help user to monitoring the young instant data, to define design the condition to the furnace and treat a triggered action.

R

Ok, this riser prior to water, our it means a provide. What is on, he went for my wheel, okay, the even here's. Something just happened. We mentioned interesting. Our triggers are by some use and are we also think the evens may be very large scope definition. Many scenes can be defined eyes event, for example, unfold or learn, or some changing a network state or some network security security threat.

R

Lock test so we're seeing maybe we can found the mechanism and help the user to automatic measurement says even and do corresponding action, and here is even examination. Disanto and slider present several event, for example, if our some surveys want to be implemented and they need to set up some panel here, we defined even like okay, we set up at her colonel from a site a to site, B, and so the tacit may be.

R

Okay, here is a young instance at a node at a no-name is a to be, and if this attano we set up successfully, so it can trigger a confirm the action. The action is okay, we need to enable some function. Satellites are monitoring the tano's ATVs, and this action may be also in will set off. Another event, for example, even be is to monitoring the section, sir. The tunnels lost faculty number and the tuck is okay.

R

This is a panels, namely the ATP and monitor, is his lost packets and if the lost packet number cause some earth protrude.

R

Then it kinds may send a notification and also trigger another action, locks out Sacre, corresponding value for examples which have and another even maybe monitoring, Kano's state. And if the tanoda state the panel down it can send a notification to other toe to the current under customer. Okay, say, say: just a no down and maybe trigger not a ones. They even told him. Okay, you need to do some active assembler over.

R

Okay, this right, it shows our solution, which defines its youngest module.

R

Okay, this young male reversal have two three blocks: it have even even it to to present even name and rent hard to distinguish different, even and even type can easy extension and you can a piss on your interesting to define what kind of even are you interesting and you want to management and another. A second block is a trigger. They defined a set of triggers from condition such as frenetic trigger or some filters filters.

R

We based on the event young us or even need to borrow their conception of their of their triggers condition at the find some three candle filters, such as existence, if some young instance appears sin and trigger, and maybe notification and sugar of recuperation and the bowling.

R

Bowling nation, its corresponding young incident data equal to some parameter of rush hood, a twelve trigger codes upon the action and the thrush food threshold filter is help you to compare different and different value offer between this sample and previous sample. If they detect, if it crosses their hood, it will trigger cause bonding action, and the third part is action, action and prizes. Three kind of action: firstly, can, since the location and another you can do this election and do some reconfiguration walks and also it can read triggers another event.

R

It can help to do some closed. Loop automation, closed-loop management.

R

Here is some gap between evened document and previous walks of charge. Young push, and here Sara table shows Awards difference between the young push and even even too young and the trigger condition. Chris Young push. How to condition for say, is a product push and unchanged coach, and here we found even the neighbor rate of answer kind of futures existent bowling attribute and the notification over there are two master can support certification, and here our document also can support automatically reconfiguration and trigger another event.

R

The next start, the next step would like to solicit more comments and import solution document. Okay, all.

B

Right, thank you in the interest of time like to ask you to work with the other authors and see if there's ways to align your work and report back on on list. Yeah I know that some are more. Some drafts like alex, is more focused on a particular problem, but it may be that this type of approach is more encompassing and move in support boat. Yeah. Thank you. Can.

G

I make a quick comment. Please yeah! So.

B

G

So yeah one I I think. Basically, this is a in.

L

G

Of consolidating I think this is a very large driver, especially the scope is larger, so I think they complement each other right, because one thing is clearly what you want to do on the events, because the whole acting part, the automation parlance of both I, think that is clearly something that would be. You know that is new in this draft. That would not that should not go into any short.

B

But including an action of notification right so in yours was about notifications right. Your aunt is.

G

B

G

So to say, if filter is.

G

B

And you guys can talk to each other and decide that you know what the suggestion is is utter. You know it is not applicable and and come back and say that that's okay, all.

G

S

Morning, everyone, my name's Tim pasa Allah, who introduced this draft an MVA base event this a job that had to be around for a while that if the early was a presenting call for ties- and it is based on discussing in nas meeting, we are going to move this job to nano mode. So here we give update of this job, so this job that actually would define three an MB base event.

S

The first two event actually to indicate the start and end time of congregation that apply from intent and the third a event actually is to keep track of. The in can apply conversion later that somehow convergent a that is not applied some company that it may be inspired from antenna, so we can actually to keep track of these intend apply the results. If it is not applied, we will actually give the reason and also we will give a list of object that is not applied.

S

So the change compared with the previous version. Actually, we add two new event: praise them the many nice discussion, and we also do some polishing restructuring for this draft and we narrow down scope to focus on the MB specific event, actually only.

M

S

The computer that is applied from intend also, we consider federal handling case in our young model with number in some cases before is fair to apply the conversion from antenna. We were actually capable warning report, the their every formation using some some building block.

S

So this is the model where the model look-alikes actually very simple model. Actually, we define the event event event notification. The first line is intend supply start. Actually, these, we also, you know, carry some session information. This is a session information inherited from the medical for session start that defined in obviously 6470 and for intend applied, and we add additional information to report the completed status, to indicate whether it is got a success or whether it is fair. So we use the error option that we find in the rest, come for specification.

S

That is actually a restaurant ITF, let's count young model and the most important part is an MBA in can apply the event notification. This is actually we can leverage using a media foe and to detect which opportunities apply, which actually is not applied, and we can actually to reporter the every information for summer coverage is not apply from in n, so here show the latest all the walking flow and also relate the pace.

S

Events occurrence and we can see the five men call for basically random has got to be defining, obviously 774 6470 and they can indicator the start time and end time. O Network obsession so for the antenna by the start and intend applied and actually take a place between the Merkava confirm, commit and letting have a session end. So for the intend of line start actually almost sent at the same time as a confirm comedian, so for the champion can apply actually can happen at the same time as the can be apply. The end.

S

So for some issue, where you have some outline of line, this kind with the cursor actually is about whether we need any amenities in canopied start, because the reason is actually in 10 apply studies sent almost as centum as a confirm comida. You know so also ETA is very expensive to send a is a startup event message. Every time you do the commuted, so we think maybe we have to option to address these. The first option is you pretest, the liberty optional and the turn off when it is no needed needed.

S

Another opportunity chopped this event. So this job there has been around for a while with singles, go, relax, sociated feedback and that's.

J

It can't, as a contributor can go back to the previous five yeah. Where does the intended applied event gets? That is, it is it after intended has been applied to operational.

S

Means so that actually is a sin at the almost at the same time as a medical confirm to me. No.

J

No I'm, not talking about intended place, start yeah I'm talking last one. Oh my.

S

J

Plight is that is that last one sent after it's been applied to operational yeah.

S

J

And to the question regarding in the usefulness of applied start event, I guess it would be useful for any clients that are not the one that you know. That's put the been initiated the commit yeah.

S

Right: okay, yeah.

E

Rob Wilson Cisco, so my one question is I, don't know whether the apply to start is useful or not as I. Probably it's okay, but it is another notification story, provide that same information. It's probably not worth sending to different notifications. At the same time, I didn't understand why there's two separate notifications for the completion of the event I was wondering whether you could just have your one, the same intended, applied and and also in the aim intended applied.

E

So do you need two notifications? There do they come same time or they separate right.

S

Now we actually separated we use the antenna star and to indicate the start and end time, but the original proposal. Actually, we think that we can use the one single event event notification to indicate at any time so yeah, that's that's the water. We design in this job in a corner table. Okay,.

E

S

E

Might be a might be more useful, just have one this is this is the end of that configuration operation and the result is either it succeed or it's farallon's failed. This is why it was not applying or something so merging. Those two might make more sense. I think.

S

Okay, probably.

E

Something say: Thank.

S

J

Kent again just quickly well, I think the picture is probably little arrow, uniquely needed with the arrow to another placement. Also, the names may be misleading. You.

S

J

Simply call it intended updated and operational, updated yeah.

S

J

As the events yeah.

S

Dad that make sense, thank you. Yeah.

B

So again continue the discussion on the list.

B

Well, the technical discussion: how many people would read the draft.

B

All right so a little more than the last one, but not much I think we need a little more review on this to move it forward and I'm, not sure, there's any commonality in this notification. With other notifications.

I

B

A little too different, yeah, so yeah one.

M

More comment to this I have seen systems which have some kind of auto commit feature where big transaction will be break broken down to many small comet transactions transactions. In this case. Sending so many notifications would be problematic so there month there is a need to be able to switch this feature off even yet supported or.

B

Established filters.

F

Jason Stern: can you go back to your flow diagram, we'd.

B

Like to move to move that to the list, just because stick that'd be fair to the other presenters who includes who.

S

B

Expectations: okay,.

S

Thank you. So the second way is the faculty for setting. So this humor again. Actually, we also moved Israel from their constituent and model. Everything is MD, related also is transported in independent.

I

S

That's so the goal this job. Actually, we want to introduce a new measure to resettle the youngest Orion to the separately for the contents and that you realize these go. Actually we define when you a beast option. We call that reset their stall in everything we define the factory data store. It's actually actually, this kind of optional data store so so reason why we introduced a new RPC overreaching. Actually, we investigated all the existing medical protocol regime already we have delete config and in this college, engine I believe config economy.

S

We said the startup take stock, but cannot reset their own intake installed for the discard changes, you can copy the running biggest or contents into the candidate, but actually it cannot be actually to reset the Canada guest or without the changes that the Iranian datastore. So does the dynamic data store? You cannot reset these dynamic data stores for rest, confer actually also apply here. You can always set reset yeah, so the new RPC which we introduced here is we use these a visa to reset contents of target datastore to use the factory for setting.

S

Actually, we use the data so as the input at the parameter, so when we were so so the messer we document as the factory contents that we have many ways actually the typical ways.

S

Actually we can use running start Canada as the the contents of the factory if the factory datastore can be defined or exists, but in some other way actually when to use a young instant data to document the contents of the factory datastore and that in some other cases, maybe we can define a set of the modules that can be load loaded by the factory, for example. But the problem is the different vendor may not have given a solution, so it's kind of improving in tradition specifically and know for also tenango datastore.

S

Actually, we like to set a chest empty.

S

So this is a new ticker store. We introduced the we call the factory dated datastore and for this datastore actually is rated only so it cannot be changed by the nanocloud reskin for operation all say RI by the it need to be reset by the young server in provision specific Amana, and it also actually is optional.

S

Data store, that means actually the reset the text or abuse it can be performed that can be revoked without this data stock and for the managing of reaching actually unlike the staff of datastore. Actually, it cannot automatically program program. Pro problem can automatically to any other data store. Also, it is the static actually, but in some cases you may paste on the hardware chain which, based on the hardware availability- and you may actually make me- make some changes. They should pay. Some comments from Yokohama least actually actually has come the way these with this text.

S

So these are Colonel status. Actually, the chart move to the data mode. Actually, the channel Manor is we restructure. The document make it more solid and we also remove the nano comparability at anybody. Many discussion and we also add a class as the new consul and two minor open issue, actually is how are they related or is that how they relate to the reporter? Actually for restart? We already have system management module, it defined the system restart, we can reuse it, so we also think I'm.

S

It restarted is not in needed for reboot that we also believe is not needed it, but if any folks single is important where you can actually can reconsider, so we single team Java had sitting around for a while and had a lot of support on an engine is that we like to recall, request the walking to last adoption.

S

That's all thank you.

E

Listen Cisco, so I was just discussing here. The it might be useful to also talk about the copy config, our pcs, to find a net Kampf, which you could potentially use as well to copy the contents, either from factory datastore to running configuration, which may may not mean that you don't want to have your art PC, because you other reasons for wanting that be poor. To this describe saying you can also use this mechanism most equivalent yeah.

S

We also hope we also actually talk about that. If they were a little struggle, be possibly move to them net metamodel. We need new, clean, clean, album, negative transport independent. That's why we remove the copy configure occupation, but in the earlier version we didn't have a copy completion, compartment expansion or for faculty people setting yeah. Why did need to be extended? Why doesn't it just work today? You me just the review the existing copy configure.

E

S

Do you, what do you suggest you you, you want to document these in this yeah? Yes, I, why I suggest scoffing copy config.

E

Just works: okay,.

M

Copy configures I, remember, doesn't work because they're the target is a choice.

G

M

Extended and we wanted to have reset our datastore, because we want this to work even without an MBA. The once you have written said, datastore there's really no need. Yes, we should describe that copy. Configure is not updated because we don't see.

L

The need I, don't think.

M

We need there is me to include it.

J

I can't as a contributor, the RPC that you find reset factory, there's worse at it and just want to I think the draft should be clear that that's different than resetting the device to factory default yeah.

L

J

There's more things that could change whenever you setting a device to factory people, yeah yeah, okay,.

D

B

How many are interested in this topic at all?

B

That's a that's a reasonable number. How do you know have read this draft?

B

It's also a reasonable number less, but a reasonable number. How many people think we should adopt this document as the foundation as the starting point for our work in this area and keep in mind that adoption is the starting point, not the ending point. So changes are possible.

B

That's that's a very good number. So that's about the same number as the people who have read it any strong objections.

D

Q

Hi so I'm William, not tinder and I'd, just want to give an update on what we're doing with some yang work in the broadband forum. I think I last spoke at the Prague meeting, so it's just an update really um so just to quick outline. I won't go through that. Those are the things I'd like to talk about briefly. Moving onto the next slide. Sorry can I move this down a little bit or don't know.

I

Q

The previous one, sorry that one yeah okay, so the emphasis. What do we work on? So we work on addressing BBF requirements? You know, that's that's the thing that drives us. You know we're not trying to provide a general solution for the whole world but of course, we're trying to make, but we are trying to make what we use what we provide them generally applicable. You know in the context of particularly the access network.

Q

So that's a little bit of a contrast between us and what I might call a core sto the tones of Technology and wants to apply some provide a solution that will be used absolutely everywhere for all applications. It's particularly, as I said, the access network, the requirements that drive our models come from. I guess the key standards are are those documents that I mentioned there and also particularly now. There's this work called broadband access. Abstraction which Tim Care is very involved in, is producing also new requirements, but for young models.

Q

So we define stuff to address what we've defined and we also to find stuff to address what other organizations have defined, but where they're not interested in defining the model. So, for example, part of our models are actually for itu-t standards.

Q

How do we work? Well, we like to important augment other organisations yang, for example. Itf interfaces is definitely the core and ITF interfaces in Hardware definitely the core of our models, but sometimes we can't do that. There's no standard yang, or else there isn't and of young, but it's for some reason. We can't use it. Perhaps some it's too much for a big lump. There aren't enough optional features. It would too much. We have to be taken on board for it to be used, or maybe there is sort of another example of the same thing.

Q

Maybe some nodes are mandatory with that adjust not mandatory, at least for our applications or shouldn't be mandatory. You know in our opinion, so if that happens, then you know not not brought broadband forum because broadband forum, you know, can't work directly in another organization, but our members may well work directly in other organizations. So, for example, the two main examples of that in IETF at the moment are contributing to the IETF alarms work.

Q

We have a lot of alarm requirements and several of our members have been very active in in that work, which is happening in C camp and is nearing last call and also there's some requirements for IP fix that we have across several of our standards and we feel there's need for new IP fix models, and so some of our members have put in an individual submission for that which again I mentioned a little bit more detail at this afternoon's ops area session.

Q

There's a last resort. We might address our requirements internally, but as I say, that is a last resort.

Q

Okay, I won't go through this in detail, but that's just a couple of slides showing our active projects and on the rides in the status you can see, which ones have been published and I'm, not those hieroglyphics of a1 a2 and so on. They're just different amendments, that's just to show that there have been several different versions of some of those.

Q

The next slide just again, I won't go through, but some shows roughly what's on some of the main things that are covered by the models in those different projects.

Q

Moving on to publication, so we have a release registry that lists all of our published software, which includes all of the draft and standard yang. All the Yang is published to a public github repository a BB f1, but it's also in the young models, young repository as a sub-module, and it's also all in the young catalog. So we're really trying to make this available and such a you know act as a good member of the community.

Q

What are our external dependencies for as a matter of policy whenever possible? All of our modules must use standard external modules, typically as a starting point of building as the base to be augmented, and that means not that that means you know understanding how they're supposed to be used and using them in the letter and the spirit of how they're intended to be used. There are the the published ITF modules that were already dependent on nothing surprising there I think abroad.

Q

You mentioned the alarms and, and they need to say anything more about that on best practices, we have our own best practices document, which is essentially an extension of eighty four. Oh seven, we will incorporate aspects of other stos as we become dependent on them. I mean our Triple. E is going to be the next one, but we're not yeah. None of our modules are yet directly dependent on any anything other than IETF. Externally I mean intend to make these guidelines public, we'll publish them.

Q

Virus I took github pages on that same repository, I yang repository, but that hasn't happened yet so final slide, but you might be interested to know what we're doing about an MVA so we're still discussing it. Of course, the real world many of our models were defined and published before it came along, and we can't we can't control when people are going to switch when implementers again, a switch, so we do have to complete can continue to support both nmda and non MVA. I.

D

Q

It's a similar situation to what happens internally within ITF, of course, so the final sort of outer bullet sort of non-binding sneak preview of some of our thinking. Firstly, this might seem controversial, but we feel we still need to stick with the non MDA ITF interfaces as our as our as US or a base. Even though we know there is the I forget the number, the updated eighty four hundred and something we know. We are aware of that for the ITF hardware.

Q

Well, it was only over a draft that was non MDA, so we don't have a choice there. So obviously we've switched already to referencing the hard we actually deferred publishing some of our modules until the RSC was out so as to avoid referencing a draft and as and if and when. We update our modules that are currently using 7223 and also four modules. That are what we call standalone, meaning modules that we define, which are not dependent on on ITF interfaces.

Q

Variety of hardware, then we will where, for example, when, when we're updating a module, that's dependent on the interfaces we will follow, or we think will follow the same ITF guidance of defining a separate state module. But when their modules that are standalone, then we're still not quite sure what we'll do.

Q

We could define a deprecated state tree, but we're a little bit more inclined not to have two trees, but just put augmentations straight into the single tree where, where we we see the need of it in order to support some non mga usage, you know to mean, in order to support implementations that don't yet support an MDA.

Q

Okay, any questions.

B

You're gonna have one, but you walked away. Thank you very much appreciate the update. Okay, so we'll be this. At the end of the first session, my first net non session, we have a second one on Thursday in the afternoon.

B

Second, slot look forward to seeing you there and thank you.

A