Internet Engineering Task Force 102, 18 Jul 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF102-GROW-20180718-1330

Description

GROW meeting session at IETF102
2018/07/18 1330

https://datatracker.ietf.org/meeting/102/proceedings/

A

B

Of kind of the old communicator keywords.

C

Good afternoon, this is the grower working group meeting at ITF 102 in Montreal Canada, if you're not supposed to be in Montreal something's kind of horribly wrong for those people that are speaking at the mic today, giving their presentation a couple of convenient things, you have your own clicker look, you can make the slides go back and forth. That's not actually doing what I wanted.

D

We'll fix that.

C

And edit, let's fix it in post, it's fine it'll be good anyway. Oh there's, also a pointy pointy thing stand inside the Pink's, our pink square. Please don't you have to talk into the mic exactly okay, so I'm Chris, that's job! You are you there's a note. Well, I can't get it all on the screen. Sorry!

C

Well joke in all right! It's teeny, tiny! Read it quick, there's a whole bunch of stuff! Oh sorry, it's not igf-1 I'll change that in post to all the stuff changing what I want to want to I. Don't want the 102 you said or pearl. There's blue sheets going around! There's one in the front here. Please sign your name! There's one somewhere headed that way. Please sign your name. That way. We make sure you get the right size room, not a stadium.

C

We're gonna need a minute sticker and a jabber scribe. It would be great or somebody volunteered.

C

Aha, Jonel jabber excellent. Thank you and notes. Okay, great you'll! Do this so just email at the end, your notes or you can use the ether tap ether whatever pad thing. Whatever you like. Let us know, we don't have a lot on the agenda today. We have five presentations, I think or six.

C

C

Okay, yep the existing drafts, the AAS cones. We sent a an adoption request and we sent an option reply: somebody's gonna post that back with the right name and the W key say: W Casey behavior, is in the same state I think we sent to the author's a please send in the replacement name. If we didn't I'll, do that, I'll make sure we did that. That's the two current in-flight drafts and then there's a bunch of potential things coming in the talks.

C

So now we have presentations, let's start with BMP local rib, who I forget who that is: aha, hello, yes,.

E

F

A question I figured I'd: ask okay.

B

F

The question is a bit of help and when I could ask this, maybe I should float the mailing.

E

F

But there's a lot of.

F

Bgp stuff, coming in from various other working groups to BGP mechanism like spring I, would like some sanity checks as the IVR chair. If, if you guys are able to take on something like that as a working group, I'm just mentioning it as a plea for help because I see all these drafts- and you know idea that they don't get out unless they have two implementations. But that's just the implementations, people I, don't know if this is saying work, I hate to say, I'm, asking for operational sanity that that's I.

C

Thought that kind of the agreement- that's sue Harris by the way I'm.

F

I'm sorry sue Harris in this hat speaking as IDR chair looking for help I.

C

Thought that kind of the agreement was if IDR, as the owner of the bgp protocol saw something come in that you know they were like hey that looks weird. You should kick them at the grow group and then grows. Hopefully, operational focus should say hey. This is something I actually need, or this is something it's actually solving the problem for me or hey. This is crazy sauce. Please kick it away. Yeah.

F

These are a whole class of documents, so.

C

F

Wanted to ask before I kicked it to the curb cuz. Maybe it would be like putting a trash dumpster in from someone's house and, seeing you know and seeing here are all these things we need help with, but I I see stuff coming in epitope applicability out of spring and it seems to have one focus and I would just like a second grow focus. My being too practical.

B

Will you a photo spherical I, somewhat suspect that spring drafts are attacking stuff? That is not really the ebgp global stuff that I. That, in my understanding, is the focus of this group.

F

Excuse me, unfortunately, they're putting more of that stuff into the global BGP fear stuff.

B

Okay, in that case, of course, the operational sanity check should be applied. Susan.

G

If you could be so kind to sent the specific drafts you want to be yes,.

F

Oh well, I just wanted to ask first and see whether you were perhaps.

G

We can do a three week comment periods and then you can draw conclusions from whatever and.

F

G

It to those three weeks, I'm.

F

Forwarding to the is g2 three or four drafts and service routing and I. Just like a comment on especially since were together they are going. You know these bt pls information on bgp.

G

Thank you. Alright, the floor is yours. Fellow yeah,.

H

So this is pollutant entity. I wanted to present my on this light about the status of the to draft about you know local rape and other about in BMP. I don't want to tire you again with you know the context that we have been explaining for the past few meetings, and so I wanted just to share with you. What is the status so, first of all, I'm coming with no edits or no updates to the draft in this meeting, which I don't know if it's a good or a bad thing, I think it's a good.

H

It could be a good thing since there was plenty of interest shown and we got feedback and the feedback was processed over time and as far as I know, there is no outstanding issues.

H

There was recently comment from so the only let's say outstanding thing is that there was a comment very extensive from Hank Smith from Nokia on the list and I. Think after some discussion it emerged that you know what Hank is proposing.

H

You know it's a part of a different work right, so it should not be really compounded with these two drafts, and it is you know like this morning. He was proposing a text for different draft, so I'm more than you know telling you with more words, I'm kind of looking for feedback from you guys.

H

What shall we do with these drafts? Hey.

E

Jeff has so juniper does have both of these being implemented right now, what I would suggest is hold on to it for a little bit, I suspect we'll have a couple of new minor comments coming out. You know the small things like further about when the pure goes down. Do you just simply no say we're down and don't have any trick, messes just trickle out and given that some implementations have you know some of these systems a little bit decoupled. The messages may still trickle out, even after you say, you're down, you know.

C

Small things like that.

E

Right but own that things are mostly looking good, fantastic, Thank, You.

I

Randy will show a circus I've got lost, okay,.

H

I

Are Iqra Montreal.

G

I

Not even sure I can enumerate the BMP drafts and things and how they relate to each other and there something stitching it all together or a state of the play or something might be helpful for idiots like I.

H

Yeah I wanted just to reply very briefly: I mean a to that I, I, agree and.

I

I'm not saying any of them feel bad yeah.

H

I just saying yes, yes, firehose! Yes, yes, so yeah I agree to that I think so far we just I think managed to not compound too many things together so that preventing you know a stole and I'm curious to see if the new BMP draft will be adopted and what they are about, and things like that.

J

John Scudder, unfortunately, Hank isn't here to respond to your your bullet there and I don't know what he would say my impression from reading the mailing list. Traffic was not I, didn't get the impression that it was sort of forward-looking and non competing in the sense of I. Would guess that maybe if Hank were here, he'd say actually, why don't you adopt the encoding that I proposed? In fact, I thought. That's what he started out saying was like proposing an alternative encoding, not alternative functionality.

J

Just alternative encoding and I will also say that I, you know meuk hope I was quite tardy in reading that whole message thread, but having done so, I will say that he, you know basically what Hank says in nicer language is, you know it was. It was a mistake to not tell the you know this particular message and speaking as the hack who hacked that hack I agree with them.

J

I, you know I, did it wrong in the first version and I think that his version is the right version, so you know it might not be a bad idea to think about adopting that encoding.

J

You know at least have the conversation with with will develop. Drafts like yours. A lot depends on you know: what's the state of play in terms of deployment, if right now, if people tell me oh there's already, deployment of this people are using it. That's a powerful argument, but I think we should have the conversation.

H

Where I thank you and to reply to that, let's say so: first of all, I'm I'm I've had a very quick read of the draft from Hank from this morning and I. You know, I like it right. Okay and I agree with you with the fact that there should be some TLV approach in the monitor message. So yeah, that's that's great I. What I would say is that, as far as I know, you said, as a No or Jeff said that us juniper, you are, you have implemented already.

H

These two drafts Cisco has done as well and from some conversation that I had with with Hank I understand that also Nokia's implemented this dis draft as well plus I, am you know, I have implemented, I mean PMS ECT as well, so we do have some implementations there, so my only proposal would be like.

H

So if we do this TL vid, so the other about and the local rib is a much wanted functionality, at least from me as an operator for example, so why life I will fully support what hanky is saying right, I think we should really let this draft go because you know wrap them up as they are. That would be my opinion or with the minor changes and things like that, so that at least we capitalize on what has already been built, and then we can make you know the next step.

H

We can start working on the new draft and get new implementations, and things like that, so that that would be. You know my my view. I, don't know if this agreement or.

I

Brandy, we shall retain or kiss again on two things, I think John Tiffany on head and I said on list. I think what thank you saying is some syntax, not semantics and.

I

You know they're going to tell the addicts in the room so before.

E

I say I've been found numbered among.

I

Them and but secondly, John's question of deployment. It is not widely deployed for two tragic reasons: one is an implementation that X sucks caterpillar is not, and the other is an implementation that doesn't scale. But my question I think that is the wrong question.

I

We have all of us applying with it, but have people built significant two layers on top of it? That's the people who will die if we change the format. The only one I know is the whatever they're calling it this week was BMP mon once upon a time, um snazzle or something I take T track, but but Cyril and Tim its GUI to death thingy. That's the only two layer, I know that seriously on top of it is anybody else have two layers. On top me.

B

P.M. Asus is T and.

H

Entities using data so.

K

G

Snyder's, if I may interject, what I understood from angst raft is that on the collector side you can seamlessly either receive the old format or his propose new format. So, on the collector side ingesting either one is doesn't require configuration on the sending site. You'd need to configure I want to sense BMP in the new GOP format.

G

I

This is like washing how you can convert there. I forgot how you can migrate the fact that my software tool said which doesn't exist. My software toolset does the old format he changed. The format I have work to do right. That's all I'm, not saying that is impossible. I'm. Just saying that is the question that should be considered yes, proposing changing the format. That's all I'm, not saying it's good or bad green or blue yeah.

J

John Scudder I realized after I sat down that I might have given the wrong impression that I was saying you know.

J

If we have implementations that are done, then we should just automatically you know, that's the trump card and we should automatically go ahead with with the draft as written, which you know, I just think it puts us somewhere in the middle of the discussion spectrum because of course, like part of the flip side is, if you know, let's say: we'd univer go ahead and ship our implementation of the draft is written, and then we, the working group, say yeah, but actually we kind of like the TLB format better. So we're going to move to that.

J

That leaves us supporting you know into the infinite future. Two different formats, two different code paths and I kind of would rather take the relatively short hit of updating to the standard version. That is what the group actually wants to end up with it's it's a conversation, there's not an obvious right or wrong answer.

E

So I am arguing John a little bit if we shift the low crib. If we ship the rib out no great I. Think the bigger question is: what do we do about know the point that Hanks raising, which no he's saying here's the thing we should go with some sub tlvs and then you know, got that open conversation.

L

E

Stuff that needs to be done, for the most part, is easy things you know shove around. You know a message overhead at an extra TLV in the middle. The things could be turned on in either mode. If you need this port, you know just standard rip out low crib use. No one option, get it out better coding. If you want the extended stuff, you can get the extended coding.

E

So I don't see a huge obstacle to finishing off the work that we have for low-grade rip out and then it's the next step, trying to figure out in the backwards compatible way. What the additional sub tlvs would look like the additional information that should go in there, because I think that, though Hanks given some examples like Ben, Flags I, think the idea is nice, I think there's a lot of implementation details that really need to be shaken out before we start writing code, but I, don't think I think this is a stopping thing. Yeah.

H

And I agree essentially with your position.

B

Renea for the question of deployed base I think one should one should actually look exactly at well: okay, how much how many consumers of the old-style are out there and check? Well, okay, how many consumers of any of the new drafts are actually they're kind of if one considers shifting and moving on to something like Hank is proposing kind of.

B

If there are only very few consumers for the for the new drafts, that's actually a different picture compared to something where, while okay, everybody would have to move yeah.

H

Okay, thank you very much. Thank.

G

G

M

Hello, everyone, this Romijn from Holly. This is my presentation, is the networker? Oh sorry, not this slice.

C

We're just going by the order. I had some a fault.

M

Madam welcome are you Korean cross, alistair.

C

That one at the end.

M

They did offers this knot. Is next I mean since it's not my presentation.

C

Oh yeah, that's hot sorry will switch up.

L

From Maui I'm going to operate in the Java of waiting and able monitor, monitoring using PNP in the use case, okay,.

L

Some some our continent, many general network using Angela's VPN, and they want to insert a song, because police either include Anita Walker querida too severely in Jamaica, but now they want to know which within label will be used by Tyler, Luther, server or controller. This is the use case along some of our customer.

L

We introduce a color of law to double collecting within labels. Yin is a figure okay. Now our cotton monitor, AG each bottle came not using p.m. here and lay is totally sure when we p.m. before session for long notice over 280 polluter and is a charity within labels. Long, we p.m. before losing.

L

Our customers say they later want to collect so many with him. We for loozer, they only one I think even for lose flung. We are, for instance,.

L

Missing finger, they say antenna, okay, they don't need to monitor with him before those, but in culinary PNG is very occasion when we monitor a PDP session between P and a series, but the missus here is at the instance of here and land. They are known within label year such a BMP message.

L

So we we pull up on the 1's order to collect collector with in label. Is we want to unit and PNP our message used using some elevator to carry within label and when, when the position station between C and the pier going Kamaka, then not hear our message, you can carry all kinds of we can label to PMP server.

L

Compared to common method used, you know network, our solution can reduce amount of labor reporter per label apply in the face and animator policies. Our label chambers intifada, improve neutral process image in the current release,.

L

Using this method, the Pima Sun sender may be a follower to p2. Now he will copy one one Luther to neuter server, while with young will appear station.

L

E

Jeff as I don't think, BMP is probably the right tool for this. So the example you give the open message putting on VPN label information doesn't help you in the circumstances where this is per prefix labels, where the label is the same potentially for the entire vrf. It's helpful what that's not a guaranteed scenario.

E

Most people. We know that need to do this sort of thing will just simply do standard eg, P curing and just get the bgp evpn v4 routes. No, that way, BMP is mostly helpful when you have route hiding that happens because you're getting more than one path. The case of VPN v4, that's not usually a problem.

L

E

If you'd like more time to talk about it, I can talk to you after the session.

L

Okay, I will talk about you, Avila.

L

For openness and equivalent- maybe they are, there are many many versions.

E

L

The other our customers say they currently use use PNP to to manage the layer worker learn rate. They want to return the bmp-2 to enhance the layer use case.

M

Related with our this, the control client elementary, so that we tried it who utilized the PM, he took our module the cases. In fact, that is the user cases that we communicated with the chinese the transcend company yeah. So I think this is a typical use case.

M

They would like to use the PM heat was solve this issue, but we don't have to inertia, are not sure you could they see the comment of the cases or not yet so that's we are praising her here so who the lack, who okay, I, think the Jeff Jeff I, think that's a probably so the idea so so would like a more discussion about this one.

I

Excuse this Randy Bush excuse the slightly snarky humor. Why not put it in the DNS and my point isn't to really put it in the DNS? My point is to further John's point is: is this the appropriate place? To put it? Aren't we just oh, let's put it in bgp right, BMP,.

K

I

A child as a BGP, so you know there are a couple places that we just put it places and DNS and BGP family are two. The two big ones.

B

B

Would it be kind of possible and fees and realistic to essentially essentially ask the model of the BGP instance.

G

All right, thank you for your time,.

G

We'll continue a discussion on the mailing list.

L

I'm going to pretend Elledge out on monitor bgp Palomita using BMP. Currently BMP is a is popular to use the yen in many many work, many native worker and we encounter yep. Do you, okay,.

L

When our customer use the PNP laying in Canada sign in shoulder of many requirement a tool to our company de leasing, the one requirement they may ask for her. Okay.

L

Now miss chowder, okay material is chowder: okay, The, Guardian, PGP origin of kilometer in Canyon, colors reservation.

L

Png only monitor capability enabled on the device. Well when, when no enable the on the device, the PMP server cannot see the work work capabilities.

L

Even we can correct connector the enable capability and are no inability to LA BMP server. Then we can can see all PP revelation or you know parameter. Then we can do homemade Walker about imitation like a pee-pee teepee before the calamita as a marvel of conservative long of later, when in the Martin, when the environment, something for the comic coagulation or behavior. The own surveys are windows that Waker you need seen that scenario. We encounter sound Stanford habla en este Judah. We we have. We have a case, no figure, the Felice Kaiser.

L

Least other proposed to the attention of PNG image initiation message: Judah. Currently, the PMP initial messages only provides the information of when the software we're watching and citizen name of the looter means Jota defender to to noodle terrorist wise PPP overlay of elation of kilometers, and the other is the default. Behavior kilometers, the Guardian PGP optional, optional parameter Holloway.

L

The information for you is used to specify specify the order B&T of publishing no kilometers the data happy enable or no enable the monitor device and the formatter is, is a pullover for long I will see for Tuesday when you want The Guardian PPTP 84, the behavior kilometer, the information for your contains at least 44 Derby. Here we are calivita.

L

Sound sample, of course, the parameter this, the la spada.

L

B

A vehicle I think the question I asked last applies to this as well, actually, actually I, think I. Think everything that you want to transfer in BMP should be available in the BGP models. And yes, BMP is not is not, is not a thing to ask infrastructure about characteristics and status and parameters; it is for catching and transmitting flying data.

B

What you are looking here is stuff that you should find in the in the young models, and you should use the appropriate technique for querying the young model instead, instead of inventing inventing new PDUs and data structures and put push them into an app into a protocol, that is not really used for that and if you find, if you find any parameter, any parameter that you are looking for, that is not covered by the existing young models. Well, okay, that's valid input for improving the young model.

M

I think that this is a totally relate here may be my next presentation yeah. Maybe here we have Modi's kasha about this one. That's a you know. What's the relationship between the young under the byung-hee yeah Anna was the was the future? Maybe our the possible action.

G

Any other comments concerns questions.

C

So Chris tomorrow in Google, it seems like a lot of the focus of this is I, have n number of vendors and I would like to deploy them interoperable and they pick different versions of default and I. Don't understand that so I need to get it out of BMP, where I would think you would test the platform first and say this: is the configuration I'm going to deploy and I know what the defaults are, because I have set them right? I, like deterministic behavior in my network, not crazy sauce, so I, don't know that I.

C

Would care put this in BMP really in the device properties. Gonna lie to me anyway, so I'll just test it and make sure it works. The way I expect and configure it exactly. The way I want to happen, and maybe I do that with the anger whatever but yeah.

C

E

I

Randi pollution, RJ, orcas I think it's what the defaults for a device are.

E

Across multiple.

I

E

I

My configurations are, what's been pushed what has been miss, pushed etcetera, etc, and what I, of course would like is some.

I

Presentation format which I can fetch, which will give that to me in as vendor independent manner as possible and I thought that's what him and to amplify Ruettiger statement. I thought! That's why I've been dealing with yang and but config for these last couple of years and putting my money over there.

I

So this seems like the same chicken in a different sauce and that sauce kind of got sold to me and it appeals to me so tell me why I'm wrong.

E

Jeff house and I'm here they tell you that you're, mostly right.

D

E

So, having been one of those guys who have written tools, know consuming routing data and having to cross correlated with things, so the point rudiger makes of No should I put this stuff in my yang model. What do I need to cross correlate to make things agree with each other? The comment I offer on this draft is there's probably too much in it, but the thing that I disagree with is this is not completely useless. No so much to Chris's point, or these at least his issue.

E

You want enough information in your BMP data stream to be able to get the most out of the data stream itself without needing to rely on know some ephemeral state that may be coming from Yang and having to go fetch it somewhere else.

E

You know try to keep the things that necessary to consume the BMP data in the BMP ecosystem, so, for example, the capabilities that were exchanged is a great example of a thing that you may actually want very similarly, you may want to know what capabilities you actually decided to use and I tell you what my capabilities are, but that doesn't mean that what I've used so my advice for this draft is probably too much now keep the things there that are necessary to interpret the BMP data don't put extra stuff in, but some stuff is still useful.

G

Okay, thank you.

M

Finally, finally,.

M

Okay, I love, you, okay, I know gay okay. This is my topic. This is a network monitoring protocol yeah. So this is in fact that this is the graph that you, the poster without it to define which working group used to be launched yeah because of a sink and use the related with the BMP. But here and also you should not leave me in here- the PGP monitoring, so that we know that you called as so. We not a determined is the working group so I yesterday, I present in you know: PSA WG, Saturday I.

M

Also here would like to collect more comments and feedback. Okay, okay, so they say you the purpose of our work. In fact, the forum last year I worked with a colleague, earth and beginning on the network at elementary for a walk, and here now there are so many at elementary work. User beam-down Patel is a the technologies and solutions may be differ in the form each other. So we tried to set up a whole picture with the Hubbard is a clear picture, so we divided it into three a plane.

M

So why is the manager manager mentally? So from our point of view that the past years, the GRDC Netcom improvement at the young model work wasting the belong to this category and second, why you, the data plane, so that related related with the forwarding plane, the most promising technology and solution using by the OEM, is to try to collect more data and the information from the forwarding Polly.

M

But now that we think this is a time to relate to the Weiser control II. So that means that we needed to collect this information for the network maintenance for the control belly, because we synchronously is also important category to be tough. So that's the you can see that we did some work related with the enhancement of the PMP and also we had this. The extended Anita to the network monitoring protocol for more control protocol, yeah.

E

M

They say user, our the purposes for this one.

M

Okay, so this is the motivation and introduction. So now we think the controlling monitoring, monitoring the Runyan status of a control protocols will enable evolution course, automated Network, oh yeah, yeah, so they say you the four primary purpose so based on this, the purpose network monitoring protocol is a Provost to collect the protocol, reinstate hers, data they see, including the protocol packets and the protocol state histy, and their to export. This data to the analytics station yeah.

E

M

We think this will facilitate the network troubleshooting there. So in this draft here we want in order to oppose the ocean, so adjusted, who probably I'm here for SS, and then you can go yes.

E

M

This is the use, a still walk. Okay, so here we propose the use cases for using the to justify the necessity of the nam-hee for the SS. In fact, we have this encounter many network of failures caused by the failure of a control protocols, and also we are based on our internal statistics. They are almost in year. 50% is a related with the failure of the control protocols. So here we propose a several important here. You, the cases first day you the ISS neighbor Tom, because they say the popular failures. There are different reasons.

M

This is the mismatch of the parameters and also maybe the failure of the circuit. So this can call the SS neighbor down, so this is using the first cases, a second a use. Our speed, synchronization failure- because this is also related with this- is a packet of the ITP- is a faladi, the in the network, but because of some is the error of the packet of all of the error of opacities, the control packet. So that's called the failure of the air speed, synchronization and the third one.

M

We propose SE the forwarding plane disconnection, so that means that we definitely we can use a pin where the forwarding claim you can either to pester this one. But some your the cases the pin doesn't walk but because of the MTU used in consistent is called the other traffic forwarder feeling. So we think this is another use cases. The first one is not proposed in the draft, but we think that this is a very popular that usual route flapping.

M

In fact, the dutiful a team also has can be caused by many reasons, in fact, is a neighbor town is the one type of the reasons, but we use also has some other reasons, in fact here the route of lobbying. In my experience I experience sometimes that time it will become a disaster for the network because of that I used the IDP faladi characteristics once he is flunking, so all the rooters will have the usual so that called a service. Other service cannot walk away.

M

Oh okay, so now that we say that you got this team troubleshooting measured. So we know that now that once the failure happens, we try to manually login the login under different devices and it will check the root her one day were, and also that the even the one by one, because the huge information to be compared it, including the RSP, including the link estate, so that a user cause time consuming and they will consume me. So the.

M

That is also you know that the faladi, the very fast, if you cannot hear speed of a locator usual cannot catch up, reduce the carpet is a the speed of the failure. Changer, you become more difficult to locate of the failure you know so these are you know. Also the you calendar for the you got a scheme, troubleshooting measures so based on this one. We would like to some set up at this automated Network, a troubleshooting you can either, that is a, cannot depend on only the manpower.

M

We hope that most of the work can be done by the machine yeah. So that's the point.

M

Okay, so you so this is the usual. We propose two options why he is from our point for the control Polly. In fact, we don't want to sublet his work or distribute he the work to a lot of protocols. We think we should a converter unit. One protocol easy enough just like this: the management that he is at your PC or net cough, and also data for the controlling some days of the human, oh yeah, so further control pony. We think if we use necessary our the answer uses only one protocol.

M

Yes, so we think the possible option. The first option we sink. The PMP is a great you could uncle for this walk because, though we have some doubt about the PMP, but evidently PMP is a existing could an uncle for reference, so we think we can reuse the PMP and also that we can extended a BM here for this purpose, so that a user can reduce the effort of the I think we use the PMP and that will take advantage of the PMP for the incremental deployment and also reduce the implementation efforts.

M

So they say the first option, but I also mentioned that the PMP hadn't used the owns scope, super sci-fi, the curve, subversive act clearly in the cool working group. So we think maybe now that he used earlier to do like this one, so, firstly, temporarily with his option to to propose, as a new network monitoring protocol des is independent from existing PMP yeah.

M

Okay, so here they see the just given um how about this one. In fact, to be honest, your seminar lacquers IBM, he you is that his EP station based and also this is a message from my pardon wheel. The most important message is the PDU monitoring message: I mean so the Explorer. This is a PD you directly to the to the analyse analystics station for further analysis.

M

Okay, so this is the message format. I will not talk about that details. In fact, for me this is the most important to discuss why this walk okay. So this is a slice for this way first day and as I say sorry for because before this is the idea mean he are very busy with many affairs okay, but here the lassies say thanks to the holidays of the this experts to pay attention to this walk.

M

We collect many comments from the meaning list, so I think thanks to AC, ready and there Jeff and no sure Anna is Anna here.

M

Okay, sir, let her know about her I didn't know: Wilton Wilton here: okay, maybe you sorry, okay, I, also that that he has some family are mayor's, not here and also say thanks to correct Skinner, okay and here the I list. Here there are the comments for the draft, so here this is a summary of these. The comments, the mostly important here to comments user.

M

Why not at York is a young second, why use a well not a bit appealing state, so I will give proper that who slice to explain that it hills and so the why the history for reference I also believe the word forget the history tools to repeat it. Yes, in fact, we are also the part of the history in the past years and involved in much work over the GRP, see, net confer and a young but adjusted the base around the experience. I think you should not.

M

He said he for a enough you to trigger me and my colleague earth to seek complementary solutions. We try to combine these the solutions with the existing work to set up a better solution to facilitate network or EMI issues. Okay- and also this is the some days the suggestions to refine the draft okay, we will refine the issue draft accordingly for beta readily for better quality in the future Russia. Okay, now I will first hear you explain the most important comments, the first, why not at grcc and a young.

M

Okay, the first, the most important Walker I- think the for young work. I appreciated a young model work in fact, I also do a lot of work. I understand this is a very you got lost in network lost in network because we wanted to use it or you moi, to either kind. That is, for the management her plane, but a in fact for our IP features to many theaters. So there are too many models and these models has a relationship with each other yeah at at the same time, the Netcom.

M

Sometimes we need a changer, the mechanism when's. The mechanism is a change such as an MP a so almost all models should be refreshed and updated, so they say is a walk. User truly should be updated, but on the other hand, so this is a so they should take other Challenger for this walk so that he is a take. A long time cannot a chopper the existing application. So I talked in the RTG working group Wednesday. You can not catch up with this the application, so it called the negative circle. Yeah.

L

M

They use the private model when the primary sadhana here they would another like to change a user to the MU wa or the standard one. So I think this is. This is a challenger for this walk. So this is from this upon the wheel, so I think for some specific issues for some super Civic works. Maybe we use some days of the other solutions. User can be solder some days of the cases one by one, so you can easily for the incremental deployment from this point. Who will I?

M

Think as a for you, somebody is a control. Protocol may be suitable, so this is a first thing. The second thing I say, for example, user I I, said for the monitoring the control protocol. The most important work may be a monitoring the error of the control packet protocol packet, because some failure offer days of the packet of the control protocol is called the failure of the network.

M

We, in fact we internally we simulate how to use the grcc to monitor in the packet, so I think you've used the traditional way. So first, we mustard who pass the packet of firstly, so that is so. Then we set how this the young model for this of the data through the GRDC to export it to the controller order, collector so I think there's a master be a when work when working in the device side. That means the past the theory of the control packet to the young model.

M

We think this is just a repeat: walk on the other hand, that this will again to add a too much work to the young model design. It will again to that user, make higher burden for the existing young model design. Okay, you see robots in Cisco, so one.

N

Quick comment: yang can code binary data, so you can carry binary data in yang, so there's no need to translate the PDU into equivalent yeah. You know data structure, you just carry it as binary data. If you want to do I, see.

M

How how you this one, how you a curate that and maybe you'll, have some the binary or the trpc2 ask for the data, but you must have a sound. Is the format so, what's the format, the the leaf.

N

The leaf carries binary date, you just finally, if you're saying it carries this PDU and you defined it as being it's a nice high speed, you I mean I supposed to find.

M

Okay, maybe I could I would like to talk about this. A question yeah so I think that you for this one, so I think this can be done. So I would like to take a day. So you the possible solution, yeah in fact that you, for we can convert this to the TR PC, so we will not adhere. Maybe we need an order to take this another another deci, you the protocol. So that's the purpose of C.

E

M

Fact for our progress in order to try to create a new protocols, you know we just the user, the purpose that you look at their great protocols- yeah, okay, so here daesil user- some, this the yeah based on- maybe maybe our the analicia Slee- use a nutty enough, but at least I think that these are defined a young model to some extent already, so they are to me and my colleague ours. Yes, so that's a use of wasting the BMP. We use that he are. We we sink.

M

A user may be that he, the for better incremental deployment. So here we listed is the released assembly so difference between these many dependent elementary and the tranche opening telemetry, yes, but a some days of the even for the traditional management of elementary are curious that they are also improving improving now, but but I. In fact. It to be only a statistic: I worry about it. It's a subida of the standardization to catch up with the existing application, yeah, okay! Next, why is the? Why not repeating state so I think that this is the same.

M

So from our point at the beginning, we think about it. This is the we think about it. This is the you the cases, but yet here the user KC is a probe host that you know our draft sure, even maybe it can be solved by the PD pinning this data, because the between in yesterday the Tax Court has some information for the IDP, but our the purpose EU not only for the IDP and maybe for the future. So the first thing, the second thing I think that is the beginning is data from our point.

M

Is the operation for the network of Iranian, for the service is another phone monitoring we should at the Kabala. We should add accompanying the monetary from the control protocol. This is the basic of principal. We cannot hear to that in order to a monitoring protocol in monitoring the protocol, rather yep.

O

M

O

We, when this was was uh suggested. It wasn't apparent that you were trying to get the actual PDUs I. Don't think anybody would suggest putting the P it all the PD use in BGP link-state.

O

Yeah, so that's what you just wanted: the link state itself, not not every PD you yeah.

M

How could they SWA so I think that's the so I think. That's a gossip.

O

Through the whole thing, that's right, yes,.

B

Else else we will demand your exact detailed argument why DNS cannot be used. Okay,.

M

Yeah so I agree this way, so I think that's. So that's a user because I trusted the needs to hear I listed here. Use the Chancellor mentioned that one each appeal in the state. Maybe you solve this, the user cases the prop house a unit raft, but in fact the user cannot resolve the purpose of our the walk. So that's all the description here.

I

M

I

Urge a may I suggest what's happening. Is that I think many people agree with you on a need for getting these data and getting these data in the useful, regular cross platform. Yada yada yada, the semantics are not yet well worked out and we've walked around this for some years, but arguing about the syntax in which you're going to express it. There is no proof of termination.

I

We can have this discussion for a long while, but you're chasing the wrong end of the animal right.

M

Okay, so I, they saw the way some days a summary about this walk. The first away, I could read the the first opinion that we think they are exist, difference between the control point elementary and the management Oakland elementary, but maybe we think they come Moria together. So we have also tried to solve the possibility yeah at least- and you know, we think this exists difference between the two types of the elementary walk yeah. We think Edessa.

M

Therefore, the control pony elementary, the major purpose of forage is it to monitoring a control protocol and also especially for monitoring the packet of the control protocol, and secondly, we also agree that the control plane protocol and the management plane elementary or sorry the control pennant elementary and the management elementary margins may overlap this Agosta. As the past of the years. You can also use an aircraft to do traffic. You can also use the PCE ubtv considering the traffic, so that's a that's the same reason for the CPP and the I'm hehe, the second opinion.

M

We think the UDC proposed in the draft to justify the necessity of the UNAM he yeah. Regarding this opinion, we have additional points. First, they do not exclude the possibility of other measures yeah, so the second one that used the code into our the in here who walk the different elementary walk, can be combined together for better solutions. You should another exclude each other, because there's a different in other cases made use a different solution or maybe use a combined of the solution.

M

Okay, last one there's this one, just like the debates between management and the control debase 4's demon over the past the years in the past years, as I in the edit EF at the beginning, they always a debate about what the management was control after several years. I think we learned some lesson, the first one. So do you see the combined solution may be the better choice.

M

The second one said the different. The solution may be used for the similar issues, but they have the different requirement: yeah, for example, for the traffic theory. At that time. At the beginning, Nobi Nobi a PDP as our policy, no peace IEP. We have to use a net half, but at that time they also agreed as a net convo performance you're, not a satisfactory yeah.

E

M

Now that the PDP, as our policy you'll available, they will transfer from the net convo based the network traffic theory to the PDP based the traffic theory. So we think this a seminar for the management and as a control pauline callum entry, Netta walk. Maybe there are different requirements different face. You choose that different their solution. Okay, a last away, we believe, with the requirements under the grunion code, to justify the necessity.

M

Okay. So now this is the next steps. First, we close the existing comments. So then we will solicit more comments, feedback and more user cases, so there's another unit idea of a community. We also use a community that, with many these operators and OTS today's about theirs operation communities to let collect their opinions. Yes, so in fact, based on my exchanging information with the IT community, I think they are very flexible and open for a lot of and a flexible tours.

M

Yes, so I think the desert, desert user call data for there and so there's some usual better for the maintenance for the computer resource and the story. The resource I think our the network should also take a summer right away and also to borrow some experience from this community.

M

Ok, so this is, and also these of the next one. We were detailed analysis of the different measures for the one here and disadvantage. Ok, ok! That's on my presentation. Okay,.

B

Who do you for Dodge Caliber? Can we go walk back? One slide, I believe I believe your stated starting point is kind of misleading you.

B

You are starting from creating a compartmentalization of information based on well, okay, some semantic concept and then and then discuss protocols for this actually actually I think you should recognize that there are, as I can tell three behavioral types of information that are of concern flowing data like BMP is and where BMP is used for tracking what happens in a BGP session in the form of the exchanged protocol PDUs state in the devices, as typically modeled in young models these days and events now, I think I think that seems to be a fairly complete classification of what is of concern dealing with these free, very different behavior.

B

No types of information is something that needs to be addressed by appropriate protocols.

B

For the tracking of flying protocol information one could use lawful intercept techniques without Wallachia you can. You can probably use every box, that's out there right now for doing that, and yes, you have to do something at the back end to process a stuff for the stuff. That's in that state of the boxes.

B

We have a young models and there may be some gaps which better be identified, and every every every transfer of information of state, the in this time and date is well advised to use the young models and appropriate code representation with the events I'm, actually a little bit unsure what I should say, but the more experience and deeper into implementation. Colleagues probably have answers for that and I think.

B

Your distinction between control and management plain does not really go for this behavioral type. Classification does not go much further, that well in the control in in the management plane kind of the state of the devices and the management of that state is of more concern, and there is fairly little of flying data that can be intercepted by the authorized Authority and while okay, of course, of course, in what you call the control plane. Yes, they are there.

B

There is the protocol exchange that has to be taken care of, but I think telemetry is streaming telemetry concepts and protocols and implementations are available for all.

B

Jeff may come in with questions about well, okay, how to correlate stuff but deciding deciding to go one way or the other for a protocol that is appropriate either for tracking protocol exchange or management of state information and force.

B

The other behavioral type into that protocol is just a bad idea now and while okay, if you want to go there, yes, you have lots of work to do and we arrive. I want to avoid that.

M

Okay, thank you. I want her yeah thanks, you're the comments yeah but hey to be honest here, I hate this one, this I wanted to at this one. This is always a sound working, the overland yeah, but you know that even we emphasize this is a young model but, on the other hand, I legacy even the PDP or some of the IDP. The the packet used a theory that Caesar also the model yeah.

M

So that's a usually in fact it so today, so you point already or relax yeah we saw that is a very difficult to clarify a clear pottery between this walk. This is my point, so that's either intuitively intuitively I. Think I, like the p.m. here at least I use the example.

M

Even we cannot clarify the pottery clearly, but into earlier. You know your model is the one thing p.m. he is another see yeah. Maybe it's not belong to the control plant elementary, but we think use a we can try or we can combine some days are together yeah. So that's that's my points. Okay. So now I am.

J

Relaying a comment from the jabber room from Sandy Murphy speaking for herself and she says: can I express some skepticism about quote requirements and running code and to justify the necessity? Finally unquote and that's the whole comment.

M

Okay, so that's what that's, how you the might be be: okay, Casa! That's a a.

O

Ceiling them again, I I, think God, I I, agree would occur, get a real good job of classifying those as I'm not saying these are the only types of management data for a protocol, but of classifying those three types is the intended implied configuration in the events and then the actual PDUs on the air on the underwire.

O

I'd also say that if you're, if you have an absolute requirement to get every one of the love, the last one, the PDUs I, don't think necessarily monitor doing it exactly like it was done for BMT for the I. Gps is a good idea. I think a better model might be something like how it was done for IP fix. You have a standard wreck and you collect didn't, pour you know, cuz it's coming because it because you can, you know what IG B's can have these and then you know dot.

O

You know these real and and there's nose tcp to slow them down, and so you saw something more like IG p, where you're saving it in a I'm not mean I, P fix and then pushing all those PD is for something to crank on. It might be a better model for it. If, if, if that is really required, yeah.

M

I think that's the thanks for your suggestion, yeah. In fact, the in fact the IV fix has ever been our one option. We haven't do something work on this one yeah.

F

M

The way we will let her we'll have more analysis, yeah Chris.

C

I'm not entirely sure this is I, think maybe we somebody asked this yesterday in ops, 8bg sort of where are you trying to go with this? It seems like you presented here as a more of an informative thing: hey we're trying to work on this thing. What do you think is that about right? For the first part, okay, it doesn't seem like there's a lot of global bgp preparations. Work in this so I'm not sure that it's particularly well fit for grow.

C

I. Think the ops area, probably is the right place after work opposite eight of the G, but I struggle with a lot of the same points have been brought up. It seems in some ways like a solution looking for the problem, and maybe the least their existing tools. That kind of do some of this already I.

C

Having run a couple of large networks for a couple of years, I haven't really run into this sort of problem that you are outlining here. I can mostly get what I need that and I, maybe I'm, just not as creative but I, don't know so I mean I would say. It certainly seems like you're you're working hard on something.

C

O

I know I know a cylinder, I know at least four always PF. There are people doing passive monitoring where they will bring up a matter of fact. One Operator actually had a mechanism where they bring up a neighbor and they didn't want it ever to out so they kept it in the exchange state was never going full.

O

It was kind of up cause lots of problems, yes, so they could get, but but they weren't, but the thing about that because at least between a flooding domain, ostensibly everybody has the same copy of the database, so they would only do they, wouldn't they wouldn't peer with every router. In an area like like your picture, you had complete fan out for the IGP to every one of them and everybody's, sending every PDU they every I.

O

Don't know that that is, there may be certain classes of problems that will help you, but I, don't know if there's enough of them over just monitoring a few. You know you know one in each flooding scope or two in each flooding scope, just in case the net. It gets partitioned that I'm getting down to details, but I, don't think BM BM t the BMP P approach is the right approach for a GPS.

M

Okay, so this a I call it this one, so this from our continent wheel. So that's a you know. Also, that's the one purpose of this draft with the example we'd, not a user, directly directly to define that this is the protocol for the export, the export, the all the video or like this, the package of the IDP out. In fact, we first to define this possible. You know cases so that maybe that's a you know, for example the the neighbor town user related with the hello P, do so you can export it.

M

We don't know one who faladi the this is the protocols that you know that irony you will be defined for this protocol yeah, maybe maybe later you know, you know all the usual once on them. Maybe at the last that's the user order. P do be export, but until now that is a note, our purpose. If that's, why is very simple? We need in order to identify any of the cases just like Indies to all the PDO yeah. So that's that's. The purpose.

I

Wendy I oj had only I am a researcher I have virus is monitors in the mesh in one global network, I have OSPF monitored in the mesh in one.

I

They do not always that the difference between two monitoring points is interesting, but probably only to a researcher. I would also note that packet design has products in this area that are rather tasty, but.

I

Operators don't seem to be interested, it's getting it's hard enough to get them to look at their damn BGP. What, along there IG? Peas? Sorry.

C

From the peanut gallery, Chris again like I agree, there are Peckham's 9 net dmg lots of folks who build or you can custom build your own, but some people do if the crazy Isis are OSPF monitors and if you're running a large network you're going to need information about your I GB, but that exists today, right I, don't need I, don't need something I'm, not looking for something new here right, I have the thing, I! Think there's why.

I

Are you aiming at me I think, that's why I just say no.

C

No I agree right what, but I would say that there are a lot. There are lots of large networks to do this. Monitoring today, right I mean I, know in which every night in Belgium I interact with, has a form of that monitor so and they're. Clearly they have it so they're not looking for something particularly new.

C

That's always trying to say sorry.

M

Okay, they say a.

O

Cylinder so that slowly I think what we're saying is: do we is there any standardization needed for this type of information other than the standardization other in the IEP? Is it already exists? Do you need to wrap it up in something to deliver it to a collector and I think the answer is more to the negative than the positive that we really need. This.

G

To answer that, I think from a charter perspective. Well, let's start with BMP. First, the B&B MP stands for BGP, so I think it will not be appropriate. To include is, is information in BMP feats.

G

And from agro charter perspective, I'm not confident at this point that we should be working on link States monitoring protocols.

I

Randy Bush Raj I, wouldn't know. There's the relevant ad is even left.

G

Okay, that's close to light after Jones Khan's relay all.

J

Right no I'm, just carrying my laptop around to keep my hands warm.

J

John Scudder I at the risk of continuing this conversation even longer sorry I.

J

Please do not take this as support for the proposal, but I do want to say that if it were going to proceed forward, I would rather see one protocol. Even if you have to change the B just aim. For you know, bag instead of BGP I would rather see you know one protocol rather than taking the protocol and doing a cut and paste, and you know name change because it has the wrong name on it.

G

J

M

Thank you, okay. Okay! Thank you. All. Okay,.

G

I think this I think this was the last presentation.

C

Quick everyone goes someplace where it's warm hurry up. It's a war elsewhere go see.

G

Anything. Thank you for your time. We'll see you in Bangkok, that's ITF, 100, free.