Internet Engineering Task Force Interim Meetings, 25 May 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Roll WG Interim Meeting, 2020-05-25

Description

Roll WG Interim Meeting, 2020-05-25

A

Yes, you're breathing in the microphone. Can you put your microphone sideways? Okay,.

B

And I started, recording so I think you can't see that recording is being activated right.

B

Now we should application.

C

You can see the slides right.

A

B

D

B

B

So basically stop doing yet, but I think we can.

B

Welcome everyone.

B

B

Those are the materials basic part is it? The word has blue.

B

Okay, this is the agenda, so.

B

E

B

About a new option about compatibility, we can discuss further compression if how it's ready to an extension. If you can talk about the document or the guitar, there general.

B

Suppose these are written observation topics that are still opened his new work.

B

Some comments about agenda.

B

So for the IDS once it away, we have to be like five thing every day. Is it five hours between 11:00 to 4:00 UTC every day, and this notes I want to be for 15 or 100 minutes.

C

With twenty can spread.

B

And there will to plant eight tracks and Madeline. So with the situation, what do you think the workgroup it's the best for us? It's like we meet student idea, consider it or we make up right after and like last time how much we will need, but you think how many is not. What do you think.

A

Happened during the IETF which, because it's more focused and easier for me, to explain my employer that at the IETF for that which not doing contingent.

B

Sorry for you, it's better to have their own meeting goodand idea once away.

A

During that rich, yet.

C

F

Would be in approximately eight weeks.

F

It's an approximately, we had a interim about a month ago, I think even sooner than that, so I do like having these interim meetings I think they're more productive.

F

We are missing Pascal, but I'm concerned that the eight tracks in parallel mean that we will have some conflict and we won't. We won't have all the people we need anyway, so not opposed to having a meeting but I don't think we need more than one slot and I think we should still plan on having another interim having other interim meetings.

B

Well, I think that, yes, that was to be very crowded for.

E

Everyone I think so.

B

For me, works like I, think it's going to be easier. If we meet like after idea within three meetings, we can have two hours.

B

So I think people are going to be more relaxed that attended back in parallel. Okay, so they have several works, but well it's a up to date. Community.

B

It would be nicely for everyone.

B

So additional comments, so maybe we can have these two, the mainly least.

D

B

Thank you, okay! Thank you, my so broadly, we.

G

B

To the many lists, to get the confirmation and see what adds it's better.

C

A

If it's not during the IGF week, I recommend, we don't do it after that is August is not a good time.

C

B

Okay, we all these comments on right there action! Thank you very much so well. If we maintain the idea well, throw.

C

H

C

B

Will be you kind of okay and then we kill him. You guys need it. Okay, take note. So then Pascal is not yet here right.

B

Here so we can go to.

B

Sorry, what do you say many.

F

Email but I can't find them on any chalk that I have.

F

Normal chalk, but I don't and I, don't have them. I mean this is go web. It steams. That is.

I

B

So we can turn to our.

J

Okay, you know if, by the end of this meeting, if Moscow doesn't come back, I would like to come back to his slides. Concave slang sense be right.

J

Okay, so this is about the back of compatibility issues that we have been having for all the new options that we have been, that most of the drafts are looking forward.

J

One of the major issue is that whenever we have a new extension, we have a full extension it's in the form of charcoal control option and naturally, some of the 6l ours or six six, seven. No, that will be dropping those. So we had the same issue with capabilities, so we thought that maybe in the future, if some note does not understand the capability, then they will drop it.

J

So we introduced some peeps and such as, if a note doesn't understand a capability, then it has to it has to still cooperate forward depending upon another belt. So the same mechanism has to be introduced for I will control options as what we think now we cannot be as flexible so so what I have here are two proposals.

J

One proposal is, it is in which we don't decrease any control overhead, but it may not be as flexible as we might want it to be. We might be able to just manage the copying mechanism, but in the last few trained Pascal mentioned that, apart from the copying mechanism, if we can have similar fit.

J

But that does super slides perfect on this slide. Basically, I have two examples. One is the enrollment priority, which essentially defines a new control option. I mean priority, which has to be yeah. I'm, sorry can. Can you please go back to the previous slide.

J

Okay, so an owner priority has mean priority. There is a control option and it has to be copied, and there is another graph which is a leading article option, which introduces a new aoo auction, which essentially meat which which, which says that ideally, a sick seller should not copy forward this. This option, if it doesn't understand so currently our queen RFC, this 6550 doesn't see what has to be done.

J

If a control option is not in the store, 6550 is very clear about how to handle if a code is not understood, that is, a message should be silently ignored if a code is not in this tool.

J

But if I control option is not in this loop, there is no clarity in the document, so, basically for lighting article option, a lighting article for what we need is a mechanism in which, since sellers should be dropping the option, if it doesn't understand it, will be catastrophic if an option which is not understood by a 6lr if an I/o option, which is not in this classic seller, is not stripped off but carry forward. So these are the two current problems that we have. What it looks like this.

J

So the expectations are very straightforward.

J

The first paragraph says that at least we need to have a mechanism in which the option should be carried forward, copied as it is, if it is not understood or option, should be dropped altogether. This has to be explicitly stated in the in the herbs in the document. There are, however, more possibilities. It just like capability flags. We can have a flag which says, drop or discard the message: if option is not understood or join as 6ln, the function is not understood. It's not very easy to handle the the subsequent flags.

J

I'll tell you why, in the subsequent slides next slide, please.

J

So the first proposition is a very simple one. We have the second high order, but now, during our last entering I mentioned that we can have a high order bit to indicate copy flag, but I missed out on one point that has a high order. Bit has already been made use of in 65-54 the secured options, so we cannot use that so we have. All we have left is the second high order port, but it means basically, the sea flag can be added there.

J

This would mean that there won't be any change in the in the existing implementations without won't be any increase in the control over at at all, but at the same time we cannot afford to have more beds such as join and discard make capabilities, because it would simply reduce the entropy of auction type. That's when I say see, it doesn't necessarily decrease or reduce the entropy of the option type, because that that option is that, can that option is still valid? What I'm, trying to say is if a see flag is set.

J

It means that that value is actually getting used. So all the others can be 0 and see if I see we see like a zeros, but if you have joined discard, then eventually we are reducing the entropy of option type and we may not be able to. If we have joiner discard something like 32, options will be remaining and we are already using 20. So that is definitely out of question. So this is what is one of proposition? The proposition, although is not the next slide, looks like this.

J

So here what I'm suggesting is we have a X flag, extended option flag and then, after promotion length, we have the 8 bit option field. This is definitely more flexible, it would still work. It would still be backward compatible with the nodes who don't understand X, but it would essentially mean that all the extended options have additional fight that has to be carried for.

J

So it's. If we have this, then we can. We can have newer flights for option types as well for different options, so these are the two propositions that are possible. In my opinion, if someone has any other idea of handling, this like to understand, looks like this I. Don't know whether there's an X right, yeah yeah. So one of the thing is we would like to clarify this in the mopix chart and not the capabilities drawn should cause essentially more per staff tis.

J

The draft which which basically takes us forward through our lv2 and we would like, is to be managed. It's not logically speaking, it's not something that capabilities have to handle its its own should be part of more things. In my opinion, but again this is something that is open for discussion.

F

So in our plv to would we just implicitly say the X flag was set and then we would get that no yeah.

J

So, let's assume that the extract, so there are two cases, let's say. For example, there is an r p lv one load and our kill. We do know in the same network RPL v. One node is not going to understand the X flag, but it would still be able to skip or it will be it will. It will be doing the same.

J

F

Let's assume that we haven't, we moved to an RP lv 2 with a new mop, which we still need the X flag, or could we we could just assume that all options from now on have that extra. The extra flags.

J

That so what I'm proposing is that the X flag only can be added for those options which need those extended flags.

J

So we can still have the same old style option types in place with our pill v2 as well, but if a certain option requires an extended flag that can be set and our PL 2 nodes will understand that flag is cool, you can use the control over it for those traditional style options.

H

Are V 1 nodes? They have to understand that X bit in pipe right, yeah.

J

So they won't understand, so that is that is naturally or till v1 is not going to understand those that that pitch. So the only thing that that that kind of node can do is keep that option because anyways, it doesn't understand adoption right. That option is up in the tree.

H

Until we market this option is optional, or this option isn't mandatory, depending on that they have to decide right now.

J

Let's, let's say there is a new option which requires the six flag. Now, let's say that this option goes to our PLP one load. All that our plv one node can do is skip this option. It can either strip it off or 4865 is not clear about it. So we can't do anything so via what we have said. What I'm saying is we have the we? We have a handling which is synonymous to the current handling or is same as current handling for Hart will be one for our bill.

J

V2 we have extended flag, which increases one byte control, overhead or options which require is extended, run.

H

Cannot keep it simple like in v2 like the like in a PVC, the extender here. Does this side is optional, header or mandatory amendatory you have to process is optional. You can either ski, but you can just forward so in v2 cannot do we keep some that fast to beat any first video you are saying it is secure bit secured or unsecured. Next we did can just say.

H

If it is set, then it is a mandatory or it is not set, is an optional so using this fast it only in a old owner rode, so new notes can decide whether they have to process the option or they have to skip the option.

J

So, essentially, what you are saying is the same thing: isn't it I mean our bill b1? If we, if you still run into a situation where are we won't, won't understand this and.

H

When somebody like it, the old legacy code, many who have already implemented what they would have been doing now, you directly checking the type and deciding they will not be doing a bitwise check. Okay, this beta initial check.

J

For security is going to check for concept arity, no one implement security. That is a different point, but yeah for securing things. One has to check the pit. So what is your point? I'm? Sorry, I'm, not sure. If I got you I.

H

Am saying, instead of King, that extra one bite you tolerate in the previous slide, we.

J

H

Length rate, yes, xbt, you will be set this option: flux right, because new options may have this option: flux, okay,.

D

H

Going to say, if we use this xB to just to say, okay, this is Amanda, it is a mandatory option or an optional option. All notice they get for them. It is may be all the new options can be considered as optional. If they don't understand, they can simply skip.

J

But isn't that? What I'm saying that? Isn't that exactly what I'm saying so it takes like I said only then the option flag so each other.

J

To have this, if you still think you know, maybe we can have this discussion on mailing list, but yeah yeah. Definitely, let's discuss their.

A

This proposal number two seems good for me because it has.

A

Coming by captain in the chapters want me to continue into tempting Mike.

K

Yes, I just wanted to propose a parent couple of talking about optional options and many options using which quickly, so in a coop of course, we ran through exactly the same kind of situation and we came up with the charms, critical and elective options, so an option is either critical or elective. Of course, repair would need to define what tha do when you get a critical option that you don't understand and in in co-op that's easy.

K

You can just about the transaction, but in the routing protocol, that's a bit more complicated and then we had another.

K

We have another called safe to forward, which means, if you don't understand it, you can still pass it on and- and here of course, you wouldn't use safe to forward- you would just use forward, even if not not understood, I, don't know if you have any other parameters of option handling in Corp. There are a few more like, like no cache key and and so on.

J

Thank you. Thank you. Thank you, a question actually that those are the flags so elective, critical and same to forward. These are the three flags that we are looking for and the X flag that is shown here just says that you know option flags will hold these values. So if we follow proposition two, we can have all the three flags, but if we follow proposition one, we may not be able to have all the three flags. Will you major critical elective elective and safety forward right.

K

Is one critical elective is one leg, so even it separates critical all right, it's not said it's elective and safe to forward is another flag and co-op. We need a few more because things are cached on a Kirsten and and so on.

J

So those are so critical and elective is one flag for us to, but apart from, and copy is same as safe to forward is what I believe the other thing. The other thing that our PL has is a node should join as a 6ln if a flag is not understood or if our option is not understood, so that is one more permitting, which is our bill specific, which may not be generalized, which cannot which other ways to that. So these are the three flags that we indeed we're so essential, excellent yeah.

K

Think that, essentially, you define a set of option flags that are set when the expert is not present and when the expert is present, then you can do it explicitly.

J

So so that's that's about it so like so just to just to confirm the last statement. So only when the X pit is set, we can have these three flags which can be used in the new in the proposition in case of proposition one. We cannot have three flags at all, but yeah I mean it's it's it's it's very. It's not flexible at all.

B

H

Is there or not right option for in our conversion to option plug? Will you always be there right.

J

It's not there. So if only if the X flag is set option, flags will be there if X flag is not set, then it's just the same old style option type.

H

A

The proposal is, if X is not said, the the behavior that we just described critical elective sixty-fourth, joiners leaf node those need to be implicit. It should be known from the option type.

J

K

F

That you're also hearing that option proposal. One is not that popular I thought.

H

F

Implicit in everyone's comments and then I think that you're also hearing that, yes, you should do this in the mop ex-da cumin and that we need more examples to discuss. So we understand the what's what's happening. That's.

J

Right, so so so in them, so we already have existing examples, so I'm, hoping that you know existing use cases at least we can satisfy. And but you know this discussion is almost similar to the collaborative efforts discussion. So we have pretty much. You know thought about that so I'm, hoping that the same experience will hold true here.

B

Maybe you can do some combination about these kind of three flags with these optional or mandatory capabilities, combinations.

J

B

You call specify examples like combining this kind of relax with the capabilities when they're, mandatory or elective optional.

J

Serving is actually, this has got nothing to do with capabilities. There is absolutely no coupling between capabilities and this adjustability proposed. We have the same same flag size in semantics, which can be applied here, but I mean there is no relation between this right.

C

J

Let's, let's get to the next one.

B

So next topic that we came in the last interim was about a compression of ripple control messages. There was a previous work in 2011 that was proposed by compression and other configuration option compression they are, but the document did not went forward then as well. We have the generic head compressions. So so we start discussing how we compression of the ripple control messages, so we kind of start like it brainstorming, but as well to move forward with the NC extension draft.

B

We want to confirm in this medium that we don't need additional features in that document. That is related with the compression mechanism for the IP addresses of the NZXT Shandra. So I think last meeting we already param that we want that for confirmations. If you think that we can proceed with the extension or the nizam of compression mechanism.

C

D

For a new work at we'll meet Janet.

D

C

J

One thing that we should do before so so VB- and this is something that I'm not clear about this. How so so we have aggregated targets. We have parents set from NS extension, we have parent address and we have our address vectors in case of our value, t VAP to be our appeal. So we have lot of addresses that are going through this control messages. I'm just wondering that how much of compression can GHz actually brings so it's GLC.

J

If we have, how much would we be optimizing if we will be if we will be putting in our new primitives? If, if that optimization is maybe five to ten percent I don't know, but but if that optimization is maybe 25 30 percent, maybe so what up.

J

I have not seen the phone she'd see RFC, but I, don't know how the numbers will look like, so it will be creative. Someone before you know George's. If we know we can have some sort of numbers, I.

D

Mean there are many different, they are extending somehow.

D

F

Got me very trivially got a 50 percent compression there's a very naive compressor.

F

That was like eight years ago that I did the test and it was like 30 40 lines of C code like it was the stupidest compressor possible and I got I, got very, very good numbers on on control messages, because there's lots of zeroes and stuff in there so try again with newer stuff, but it's been. The best is, if like. If we get a library of sample packets that are representative right.

J

In my opinion, actually 50% compression may not be good enough because because, if you see the 16 bytes, if we address is compressed to just eight by it's in I mean in some in log moved lot of cases, it may not be possible to compress to below 8 by X but yeah again. Maybe maybe we should maybe the the scenario that you used was male. Possibly q1 is 50 message compression best case, so I have sample sample packets, interested.

A

I think you had implemented six large type of compression packet, so what kind of compression ratio did you achieve with that? You answered Road interrogation,.

D

A lot essentially reduced.

D

D

Well then, whatever the thing here is not to me I, don't understand why we need to link the generic compression engineer compression. We are going to work on with a nice extension I mean we may have a standalone compression mechanism that can be applied to any other, including any extension. This is what this is. My main concern right.

A

Exactly I agree with you, but the reason why I think several people who are worried or interested in looking at figures is that in this case you have a list of addresses and, and so compression is really badly needed, and the the reason why I said I would want to have a vision for how the person is going to be achieved. Is that some other mechanisms, like the first draft by Mukul guru, who is listed, which is listed here, is an ad hoc compression.

A

It takes it each shield and figures how it can be represented on sure beds, and so, if we need such kind of things, then we need to the the ropes the flags, the reserved it somewhere in in NSA to do this adult compression, but as well had much rather go for generic mechanisms. Just what I wanted to send this everybody agrees with that, and it's ready to go in that direction and work amend, generate mechanism either for compressing addresses or for compressing any kind of header. Then I'm fine,.

D

F

Also, we need to negotiate compression whether it's going to be used, and so it fits into the whole capabilities dry document as we might want to negotiate, and it seems that we're creating a lot of interesting.

F

Extensions, and so it might be good if people could optimize last because they could assume compression.

F

Just saying the silence that I've texted in message, pascal through three or four different ways, I'm a bit worried now.

J

For the network we have currently three types of compression right, so socio impr, which is I, mean six p2p or Pilon, are clearly ODB is making money type of compression using one type of content, six lower edges and the GLC is another. So six leverage, if we discussed before that, maybe six lower edge, might become a norm, will be too and if it does I mean then then then we might. We might have something like that be used in all the context.

A

It seems I'm hearing everybody agrees that we will have some generic compression between them, something that works across ripple options, and so I have to do anything special for anything and, as I said, maybe we already came to the conclusion: must.

J

Respect this would mean. Would this mean updating, then? Is it document? Because because there is extension currently does not do any compression? If six floor adage has to be done, then maybe it has to be provision now, because six wattage is already there and we are also closing on by four. It will be too so so what I'm trying to say is for.

A

Its general question: okay, the original question: do we need to provision something to say this is the uncompressed version. This is a compressed version, but no with the conversion work above these options, irrespective of their encoding, I mean they're flags or whatever.

J

Believes that we should do this.

D

Before and then you don't care inside right, so it can be with or without the compression it should be.

J

Yeah so right now, NS extension doesn't have any flag at all to say that there is a company that is compression order. It is compressed or not.

D

We have to pull the phone, so you want inside no.

J

Yeah I'm not sure.

D

Outside then, cause nothing to do with the MC, which is which is the NSA.

D

Same with the siblings right, if a note will give the simply.

J

I'm not sure if I'm hearing you properly I'm sorry but I can hear a lot of breathing sound and.

A

J

You're breathing.

A

In into the night, thank you thank.

J

You in some so sorry judges. Can you please repeat what you said last I mean.

D

What I was trying to say is depends what we want to place a flag right if it's going to be.

D

Being outside to any other requires such compression.

D

A

Sorry, you stood over when each other I think we're saying we want compression for a lot of options and options. We already have four options with we're going to create in the future, and so we gamble that we're going to be able to do good compression over all options. So we don't need a whole thing. This one.

A

Any anybody opposed so that's conclusion.

F

Write it in the easier path.

A

A

B

Okay, thank you. Work.

C

Our kids, okay.

J

So I this, these slides are not going to be detailed, slides most of the problem statements about article observations present more than couple of times simply beside he lives in infill in fiscal sessions. So the aim for this discussion is that I have a list of all the points in our field. Observation draft and I've mentioned the current status and the possible next steps in solve the cases. I know what the next step should be.

J

Some of the work has already been initiated for some of the problem. Statements for this is for some of the problem. Stick with. There is absolutely no progress at all, and you know so. Let's go about point by point. So first is the handling TTS M increment problem in storing mode of operation.

J

Right now, the way I see it is that it's very difficult to get interoperability in place for between between between implementations, for example, rioter quantity, just because the way they in which they TTS sense, I increment. In between these interpretations and say factly for storing mode of operation, there is no. There is no clarity available from sixty-five fifty-one storing mode of operation. There is absolutely no problem, because a route is controlling everything lot of problems.

J

How this detail increment is handled results in lot of control over head. A lot of you know, especially the subdued a crowd update, completely depends upon this.

J

Previously we had a discussion when a Pascal mentioned that subdue dagger out of gate problem is a different problem. You know a 6lr when it switches the parent updates, subdued act rooted at that node to its preferred parent, the new preferred parent. So there are a lot of options there, not sure if we should handle it in a PCP.

J

So all these points are actually mentioned if the interrupt as to what are all the problems and what are all the deliberations that have to be clarified. If someone is putting it this best practice, but the way I see it. There is no change, at least as of now I, don't see any change that is required in the standards for that. But if someone can optimize actually, then then someone can think of some change so as of now I think it may be a PCP yeah.

J

The second one is about Dawa handling, so the problem was the end-to-end traffic has to be initiated by the node only once the end-to-end path has been established, so so that is one problem also, if a node switches, the parent, in some cases the the sick seller- might want to know if that parent, if it was able to establish a complete end-to-end path through this new parent. So this again, this work is already in progress. If there is a tour for this, so I want to wander the details.

J

So interpreting click on reset is the next point. This came up because there is a really resetting trickle back to the back to the min or resetting that pickle timer within the current time right. So there were, there were two interpretations. Maybe this should be clarified. Our observations already clarifies it, but since article observation is not going to be a published document, so I don't know why they, tomorrow draft or whether it can be used.

J

All these things can be combined together in a single PCB point number six, the section six talks about handling resource and ability. What it means is, in case of storing a movie, a high level higher articles no to level up may not be able to signal its resource and ability to the note below, but now in some way mean priority handles it.

J

The main priority Phil will should be handling it, so I think I think we are good for that, but I try to check if there are any deliberations which which may not be handled by me, but I couldn't find any so I think that should take care of it. Point. Section 7 talks about handling aggregated targets.

J

Now, if a swoop subdue that during parents, which, if 6lr updates the complete subdue dad at its point to another preferred parent, then the Ruby loader sent as part of the Dow. It definitely makes sense to aggregate those targets and compress it if possible. Again, there has been some discussion about compressing. This address vectors and I'm I'm, hoping that.

J

Whatever existing mechanism of new mechanism that will talk off will take care of this for students, then there is transmitted transit information option. There are certain fields, such as path sequence, which are mandatory for operation, but RF citizen mandating so how whether they should they should be mandated, or you know how this should be handled is is another point.

J

9 section number nine is: is updates to our field. Now, when I wrote, the section I didn't think only from the control options, point of view, which were we are sort of handling now, but in general about all the other places, also where usually the backward compatibility or upgrades upgrades have a problem.

J

But in my opinion, maybe backward compatibility option handles it in the most most of the questions that I have raised in that context, in our observations are answered by backward compatibility, so I don't know if anyone else has some other point with regards to grades in general, for after.

J

Next next slide, please.

J

Next slide, please, in this.

B

Slice current practice, continuation that start with 10 bad control, bits handle.

J

Iii, don't see the new slide, I'm, not sure if others can see it, I'm still stuck on the same same same ste slightest before okay.

B

I changed it. Current study changed.

J

I can't see that okay now I can see the so.

G

Section tell it's about thank.

J

You thank you, the path control beds. So this is an interesting observation, because 6550 says that the path control bits are a mandatory fields and, interestingly, or surprisingly, none of the existing implementations actually follow that must class. So what should we do?

J

Actually the path control bit should come into picture only with multiple preferred parents, but RFC 6550 say is that even without multiple preferred parents path, control bits have to be set indicating at least one part is reachable, but none of the implementations currently do that, and none of the opening limitations currently do that so wondering how should this be handled?

J

Next is section 11 at justice approving with our pill, I think you're just draft sort of covers it. It not sort of it actually covers it properly and I guess it's for adoption. Then we have lighting. So how do we handle? So there is lot of static information which flows in DI all the time lighting our pillar info essentially takes care of it. It's it's it's in progress already.

J

Next movie number 13 is very interesting because I remember, we had discussed this and we had said that lollipop counters, the linear part, has to backed up into the persistent memory to avoid problems.

J

So we still are saying that we still need persist. A storage in the linear part of all the lollipop counters. This is a problem for lot of deployments. This is certainly a problem for that approval. I mean the deployment that I am working on, so so I'm wondering if we should handle this problem, we're in either we make the linear parts small or just making it small won't help. We may need some other.

J

So one of the interesting point is Pascal has handled this problem interest in a very interesting way for the new option that he has suggested the for the aoo option in the alighting article in full draft. So what he has done is he has used he so the our CSS, so the new options.

J

So what a lighting article info does is it avoids sending the static configuration all the time it does this by sending a counter in place of all that information, and if you see the same counter, you know that the same information is well. What Pascal has mentioned it. The draft is, if until an end and until and unless the lollipop counter cools into the circular part, the whole information has to be kept kept on broadcasting.

J

This eliminates the need for persistence to resolve together for that horrible counter, something similar if if it can be done for K th, then to diversion a lot of other now sequence, I guess don't have problem, but all other lollipop counters.

J

Section 14 comes about capabilities. The capability to draft is already in progress and section 15. There are a few points about our data specification where 6550 is not clear and I feel those are. Those are not a big deal, but but they might just need some clarification.

J

Actually, this part would clarify to me. They were not clear to me when I it's tough and then they were glad to back to me by the wacky coupe and then I move it to the specifications that.

J

Sort actually these these are the film.

J

So it looks like we have Pascal online as well.

G

I'm sorry I get interrupted the meetings for three hours from two hours from now so I, don't know. I'm, sorry, okay, first say I'm, which I understood as being Pacific Coast or something I. Don't know.

G

No ironic, she needs to know.

A

Right, we shifted your presentation, will a turn. The meeting still have a slot.

J

So so, basically through the slide, so what I'm trying to say is in some cases I, don't know what to do or maybe working group should decide what to do for those points or you know, have just a PCP and put all the points there or not sure what to do. A lot of whole lot of other points are already handled in one way or the other yeah the eliminating need for persistent storage. Also, we we had discussed a lot of times before and I feel this. This problem.

J

This has to be handled. One big big and it's not acts acceptable that a deployment has to depend on a persister storage just because it uses arm peel. You know, because if the application doesn't depend upon any person to storage, but our fuel and an ATM exponent, so.

G

I was already there when you started this kit, discussing the persistent storage problem and I think we could have discussed it, but maybe it was one to one in the past. So the two things that that you can do one is not to use a straight part of 16 right I mean it's the max, but but you could actually start at 254 or something and I. Guess that's what you're in mind when you mentioned my other draft on synchronizing the options yeah. So if you start at 250 something, then the straight part is very narrow.

G

You have control on that any given spec an implementation, I guess we could say you can configure it, for instance, why you stopped and then, if you have something which cannot is quite quickly, even if you start that 240 right or I don't know, even if you increment quickly say somebody has a level which is 252 and you start at 248 after for iteration you'll pass him. So if somebody is stuck somewhere in the straight line, there is a point where you will pass him for the I use.

G

I mean I when we discussed it, how you mentioned that, sometimes the di you don't progress very quickly and if they don't and we can they stuck for very long time, so it has to do with how you you, you advance in a straight line, so the new implementation could make sure it stays only for Southern duration inside the straight line. For instance, we don't have a spec for that when we do the Republic. What when we do 6550 beers just like you're a lotta question when you do 65 cents?

G

Yes, we could say: hey: there should be a parameter somewhere, which is how long you can stay in the straight line, and so after that you have to go into the rod space, and so you really pass we're very stuck in a sprinkler. So this is something that that we could do. There is another thing that we could do yet, which is each time the route reboots. It generates. Four bytes random, something like that and you place it in the yellows just to say: oh, it's, my random good blah.

G

You know like two bites or four bites random. Every chance that that you know it's a new something new. So if somebody has history about which it was never there, then it's an.

G

That's the sort of thing we could do now, how we specify this beyond what people can implement. I mean I believe that at some point, that will be a 65 C 250 days when we are ready, with ripple V, to look at what ipv6 has done. They the basically matured, a good number of specification around 24 16, and then they did dated 80 200 and a 200 192 and return. They did the family, so they have a consistent thing. So now you're working on capabilities and Rob backs, and we have the projection.

G

We have the use of referring for etcetera, so we are building all those specifications, and so that will be a 65 50 beers together, everything which will be starting at map both 8 of them, and so that's the that Timewell. Your draft would really really come handy because it's so early list of 420 to fix. That's that's how I see it.

J

Pascal, so this is what I gather from so that there two approaches, as you mentioned, one is reducing the window. Size sequence window size- that is one possible option possible option to dimension, was tending to yeah that the straight part, the the window sequence, the sequence window- should be reduced. That's the state part, so the other option that you mentioned was sending a random information from a random four bytes from the root for.

G

The first thing means there was sufficient to to twelve time bound as well right say you can't see more than two hours straight line. So after two hours you move, you have to increment and move out of the straight line.

J

Pascal the observation drafts, you know one more thing: you know what we have observed is that it may not be necessary for all the sequence counters to be handled in a generic way. For example, TTS n might need a specialized handle.

J

The part sequence of DT SN are the most critical and tau sequence may not have problem, you know. So what I'm trying to say is maybe a common sequence loading, the sequence we do for across all the sequence counters.

J

It may not be needed. You know one more thing that animation mean to do.

G

That, basically section seven of people gives you a generic mechanism and then, as you say, for each counter and I have one, as you mentioned, so the synchronization of options. We may say, oh within that generic framework, which which has a straight line up to sixteen, we'll just use two of screen, or we put a time long or whatever Oh Willa. You know that's so so I'm not saying do that for everyone I'm saying you have the option, but but yes, I agree, it could be for each individual.

G

J

And one more thing that I wanted to just confirm with you and Pascal: this is about the draft and the handling that your suggested for aoo auction in the lighting our billing, for so it also has a new lollipop counter. That is added, but you clearly state in that case that unless and until the lollipop counter goes into the circular state, you should keep on broadcasting all the information. This is an explicit statement that is made in the lighting. Artists is great because it solves the problem in a nice context-specific way.

J

You know I'm just wondering if something says on something like this can be, can can be thought about. At least you know, it may not be possible to handle it in something like this. This way for all the scholarly book counter, but if it gets handed in this way, something like this way, you know that you just say.

G

Makes sense, but you know for anything repo, even if you write an errata or something it's not even an errata, never is really correct and for what exists.

A

G

Not for adding something new, but the implementations are already there. So that's why I'm more interested in focusing on v2 while we can reset you know the starting point you can expect to have this these days this. So, let's make sure we have a clear list of what we want to make sure we have anybody to and that's where your document is so critically important.

J

So basically, what we're saying is we'll put something specific to our v2 and if we can think about how to handle the sequence window, if we can make use of some context, specific information in context to individual lollipop counters, we'll see if they're right.

G

What's the sorry for being late guys, but did you already discussed the capabilities? Drought, I.

J

Think I don't think we are discussing a beverages carefully in this syndrome. Ok.

G

Because ok so I reviewed it, but maybe we can work as quarters, so I think with your honor.

J

That's a that's it from this set of slides, I. Guess next slide. Please let me just check if there's another one.

B

It was another one, it was not acknowledged.

J

B

Now you can see the past cult, slides, Makka, pakka, Thank,.

G

You eNOS: did you give me the ball I'm a q1 working will not ask me, but that's because I use it just as you do and I guess. You will appear working group when you log in some way. Yes just a memory inside WebEx. So please give me the ball.

G

B

E

G

You give me the control, but.

E

If you want to do this, I make one working group: okay, I, stop sharing now.

B

You should have.

G

Right yeah, you still shown us us having the ball, so you should click on everyone working group and say make presenter.

B

Okay, there are two presented: okay, no, oh now you are I am thank you.

C

Okay, but they don't see.

G

The slice so did you upload there? No did you share them? How did you I won't share.

E

G

Enos offline, let's chat, I need to show you if you think that who looks.

B

Okay, now you should be able to see me.

G

Silence: okay, yeah yeah, because I expected that you had applauded them, but you did not.

G

Okay, we don't.

A

G

Webex we will talk after.

E

G

Talk about when you want to give the ball, you want to upload the slides to WebEx first and then they stay on the screen. You don't need to share your screen anymore. They'll just have their, but but I need to show you that so we'll do that offline. So please next slide.

G

So this is a draft that I wrote like sometimes like six month ago. That's because we had an interim and we agreed that for repeal veto.

G

We wanted you to have a certain amount of functions of new functionality and whether those functions was the capability to make sure that all the configuration was up-to-date and, as you know, with ripple as it stands, the root decides when it sends too configuration option and after that it may not sell it for a long time and the device such as boots or reboots may not get the configuration, and there is no way to know if you have the latest version of the configuration and as it goes, we are now adding map packs capability.

G

I mean a number of new options, new options and new messages, and- and we are not clear- has a new node joining that you have the latest of everything. So we said hey. We need to have something to synchronize the database floating work and that's that's when I started this drive, so the draft has not evolved because we had other priorities and with working on user free-pour realities all those drawers. But at some point we still need to move on with this one right. Now it's a personal submission.

G

It's just what came out of my mind and I did not get a good chance to talk to anybody, so we can reach an JIT. I mean it's just a first draft, but since I got no feedback, it's still there and it was like that at IETF 106. So it's been six mostly.

G

So next slide, please.

G

So the goal the goal is to make sure that in run time, when nothing changes, you can have small di o--'s and small doubt both so the small D IO would elide all the new options that we are talking about configuration and capabilities, etc, and the new dao could alight a number of rotten trees and just say same as before, so we in the dark, as it would take us nearer what you get with any state protocol such as was careful yes, a yes when you, when you sing the database and as long as you know that the last number that you saw has not changed, then you know that your database is up to two today with those the way we plan was designed will repeat over and over and over, because we we don't know if the guy we said to is there and we may use messages.

G

So we believe that by overwrites will end up working, but there is a cost associated to that. The fact that you repeat your Dow periodically so having a way to say Oh send hours before and if you add it, then you still have it that's kind of cool as well. So these drafters both compressed data options and compressed down option next line.

G

So we basically, we could have said a. We have a sequence, counter per option that we protect, but that would be a lot of sequence counters. So the proposal in this draft and really really that's- that's- meant to be discussed. Yes, um basically, we have a sequence counter for a stable State from the perspective of the mood and that stable state includes all the possible options that we want to protect them in the current ones and if we options in the future- and we also bcrypt protected by that single sequence number.

G

So what you really want to know it's for each option that you care about. What is the last configuration state sequence, this LCR cells, our sequence, number at which it was challenged so by just exchanging you know with your parent here is the the LCSs to which our sync, then you can detect with your parent. Now what you missed and just synchronize that let's go go now. You will find in the gory details that, because we are using section, 7 of repo I mean the same. The lollipop thing that Howell was talking about earlier.

G

We have this comparison window of 16 well by, even if we have a sequence, control of 256. Two numbers are only comparable if they are. If there are sixteen abort, even if there is no challenge every something like 16, you will need to visit.

G

Just sixteen could be here hours or it could be weeks just depend on how fast you change anything, but every 16 global change, even if they are in option B, you will have to rethink every option because you, otherwise you cannot compare this one's number. You have in memory and the sequence number that the giu is telling you they can only be compared if they are within our 16 window.

G

That's part of the way report on doors is the sequence number to make sure that we don't have this any less than B less than C less than a that? We also SP everyone.

G

So so that is pretty much it. We have to update the base objects. We change the Diu because we need to say a it's yeah. You go with a CSS brown. Then again you only increment the our CSS. When there is a configuration channel one of the protected options, so it's not that every gr you will have press one it's only when you change an option somewhere like the configuration. So you can live a long long time with the same assistance same thing: we we have an LCS.

G

We have to modify DD to Dowe, it's not yes, ESS, but you have to modify that out to also indicate it's the same as thou number blah, and then we we are to change the disk and that's when you know this Dominique came into the discussion because of changes, because we use the disk to query the parent in order to resync the options. So the answer is always the options come folks come from the root and they go down. So your parent is always more knowledgeable than you are. Actually.

G

If you have multiple parents, you will see that in the draft we discussed finding the most up-to-date parent and synchronized only one parent, that you must be a more up-to-date than you are. So you basically sync with your parents. You use the DS to say: hey. There is my LCSs, and here is what I want to see and the parent comes in. You say: oh this one has the challenge.

G

This one is chip, and the last thing that we added is this new, a oh, oh hello option and the auction is basically an option that contains an option. So when you see the format Street, no nao tells you I'm the IAO for option blood-like configuration option, so you have the disk command value.

G

This type, which says how many you, then you have the subtype cotton coat, which size and the IAO for the configuration option, and then you find the CSS so in the subsequence aoes instead of placing option, you might just place the IO IO, which will say for this option. The CSS of the last challenge was that for this adoption you have CSS of let's change once that's, so that's what what da allows you to do just give for any given protected option, what the our CSS was.

G

So if you have a child- and it knows the configured option for a CSS 15, then even if the a CSS has moved to 17, there will be an AO o, which will say for the configuration option. 16 is too good. 16 is less I'm. Sorry 15 in still has change, so the guy was 15. He knows that it doesn't need to synchronize the configuration option, but since the CSS has moved, that must be another option that cares about.

G

So that's how you can place all those options in this compressed form in a single year. Nothing changes then just put the abbreviated option to say. Yes. Yes, s applies changelessness.

G

Okay, next slide, please.

G

Okay, so the answer is so far: we protect all deductions, so we have the ma packs. We get the capability that is well Oh. The slide is obsolete. Now we have more packs and capabilities dropped. I have dated some things, but they fail to like that. So so four and five are the new drafts that that hole is done. Then we protect the rot information. It's not really used, but it's actually how a given route Kansai can reach that particular operations. Where has another route, may not be as good to reach the bottom prefix.

G

So it's it's. It's like the routing option in your right, so they do that configuration option. Is the one wanted to protect most, but then the prefix information option, if you don't change the prefect, why all over and over the prefix volatility, so it's protected as well. Next slide, like I said: if, in the future we have more options, the same mechanism can apply, we won't need to define a new audiences, so I introduced the a or earlier. So here is what it is.

G

Does motion type of option like that sighs, hey guys, I'm on a roll, then you've kept the abbreviated option so the option that you are preheating, which is one of the five that we just saw, and then you have the FCSS where that protect option was changed for the last time. That's how the child can see. Oh, it was not changed between there now and and class. The new so I have the latest.

G

So this was chair to place two things and not 100% clear that we one thing was a set of flags to indicate which option the child wants to get, and the other is the LCSs that the child was synchronized to. So if the our CSS is now 17 and the child was lasting, it so was 14. Then the child will say 14 now, if the judge says 14, then the parent should be able to say.

G

Oh here is what he needs, in which case the parent doesn't really need to five bills, but they place them as well. Now, if we decide that we don't need that than the five bits in the left, maybe we could avoid them. Those five bills is one per option. Basically, the child could say: hey I want to get option. Blah just sets the flag. Now.

G

It's it's it's two to two ways of asking something from the child perspective. I thought it was better to be able to ask with a fine-grain whatever you wanted, but maybe somebody will tell me it's useless anyway. You need one of the two you need to be able to say here is my latest state.

G

Please give me the latest for every bit next line.

G

And now I will ask you guys to read. This is basically a summary of the rules on how the LCSs works yeah. She thought I really wanted to discuss that on this car, because I wanted you to have a chance to look at that text. So it'sit's the important piece inside the draft.

A

Looks like a note 1, it looks like a note 1. Please read.

G

This time, I will give you time.

A

That's why you missed the note 1.

G

So how are we sitting Esther so chill hag and if you upload the slides as I indicated how you don't have this lag? It's one of the benefits, so a nurse will stay. If you look, man will stay at the end of the corner. I'll show you.

J

So Pascal I have a question I'm wondering right now, so so the graph changes the base objects right. So is that what you are trying to say that you know this? This thing should be made use of only from our field v2, and you know. Yes, that's the point.

G

Well, I would like to make sure that if you have not written me to implementation, it has this thing, but you cannot have to repel anyone to understand.

G

Okay, it's supposed to be backward compatible in that the p1 will ignore it, but really you will not get the service that you want, so it's kind of she's less. So this thing really starts with ripple me too.

J

Yeah, so it's so if, if there are any women knows basically they want, they won't be able to join the network network network at all, because because I mean once the our CSS starts, getting sent and I'll kill. V1 no tries to attach. It won't know how to query and all those things.

G

I mean the goal is: when you start playing with this you'll be a you will be having a mapable signal, so the node will journey for some.

G

There is a lot of noise, so I could not share your voices. There was some additional noise.

G

But yeah I guess we agree, I mean the idea of the capability of this etcetera is to to agree on the baseline that we want to have they want, for. We probably to make sure that every repo, v2 node will understand those things, so that will make a huge increment that will justify having a meeting.

G

H

Everybody read.

G

Those five lines.

H

Pascal I have one question like in in do do you have separate assess our CSS counter for each of the protected options.

G

That's a design point that we want to discuss. Maybe but I mean having to sound too many. Our CSS and keeping memory to Vinnie a CSS looked like another scale because most of the tiger changes so having a single Assessor that covers every option and changes. Every time in your channel is updated, looked like better.

G

So when you increment their CSS, you may change all options if you want to- or you may change only one option, but each time you change any of the option. The our CSS must be incremented.

H

Change, we have to always change the global, our CSS option of Dao and in a deviation of Sun right, whatever new mez new option, you have introduced there for the each of the options we have to send our CSS counter. Also it and all they will have the same value right exactly.

G

That that's why you have the AO, because fine sharp-shinned, you will have the latest LCSs at which deception was changed. So it's a way of doing it.

G

Counter only one a CSS it's to the CSS of the giu, not a CSS of the option.

G

But you don't increment GSS if nothing changes right, you just incremented. If something changes, I.

G

Mean if you find there's a problem with that go ahead, because it's it's just a proposal on the table.

J

Yeah I think this. This analysis currently takes care of options which rarely changes. That is the that is right. We never ever you know. Pio is something which I rarely changes.

J

Configuration option is another thing which is a repeater readily changes. So these are the things we should do. Yeah.

G

That's part of yet: yes, if we, if you things, don't change much having a single single counter for them all now, that's like I said you should have one option the way it's designed right now, if you have one option that changes a lot after sixteen changes, you will have to resound all the other options info, because otherwise the sequence counters are no more comparable or we need to change the window in the sequence counter 16 to 32, or something like that.

G

Remember in repo we we want to avoid the traditional sequence counter issue, whereby a would be less than B, which would be less than C, which would be less than D, which would be less than a because after going around the circle and the way we avoid that is, we can only say that a is less than B if there are less than 16 apart. So you see with a contour of 256 how a less than B be less than C C.

G

I mean it would take a large number of to go around the circle, so that that's that's the trick that we have in repo to make sure that we always know that we can compare the numbers at 16 apart with rip. They cannot be compared and that's how you realize that one of the two members is stuck. There is a new york solar.

G

And they become the profile. So can we go to the next slide? Please, like this.

G

G

The child can can set the bits for the options that you want, so if he gets an ell CSS, which is blah and yes, the a o which say for each of these options here is the last a CSS at which they were change, how all those which were changed, since this child synchronized knows need to resynchronize them. So it will set one of the several of the bits say on deception, deception, deception.

G

Now the child will also say at which a CSS, well, a synchronized, so like I said earlier, can be duplicated forward, because if he says I was less incra NIH's that this a CSS, the parent could figure Oh. Then he missed this. There's this so there's no point one child to ask. So as we design we'll figure out, we need the bits synchronize our CSS or both want that right now have proposed book.

G

Also, you can send more than one yeah this Jo message. If you can't fit all the options in one, so you have to send them in a short window of time, but actually most of the options are replaced by a in which case the message is that big?

G

If the child says our options.

G

But you you sent a CSS, which means that some of the options are actually still good. Then you can just send the AOL Hana so just giving an example. This child has persistent memory. It has synchronized to a CSS 200. Now time passes the child reboots, and he sends this this and by default he thinks oh I rebooted, so I don't know. So let me set all the five bits.

G

I want to see everything and my last access was 200, oh, that we are at 205, so he will place in for all the options that changed between 205 and 204, the others for which the bits that he will just place- the AO. Oh, don't worry! This one was chanson 199, so you don't get.

A

G

A

Sentence that the Jerry messages with this and in which sequence or even shorter this tank, it worries me a bit: it's always page for implementation problems and that might make a difference if she is one way of querying, which is the earth yes or the other way, which is exact because it's English a note says: I want a DP mo and yet Rd P. Then it figures.

A

Well, my parents heard me and since I requested Emma, no, it probably come back soon with Emma knew well six years for country at a CSS number of this and I get VP then might decide. Okay, then Emma no probably haven't changed and the soon needs to be sooner. In the latter case,.

G

Scratching my head, so did you say a I will ask him first and I will ask Oh later so I'm sure that everything fits in the di you.

A

Got multiple options and there was not enough message in the message: it will be right over multiple messages, as you say enough, that's.

G

How I would yeah right.

A

And, and so if you query by Frank and as soon as you get the first two responses, you know that the your parent told you and it's going to provide likely all the options you requested. Whereas if you query for our theater number, you never.

G

A

G

A

Get a few ones and figure: okay, then the other ones haven't changed.

G

To have the flags cuz I wish you asked for the five? Yes, you don't get the five even in the form of a euro, then you know you totally kill for it yeah. True, very true, so it's a good reason to get dislikes.

A

G

Reason would be to pull them one by one instance, if you're operating on the on the thigh, like 15 for G, you can set on the flags now, if you're operating on your 15 4, and you know that you can't have a pai euro and the configuration option and blah blah at the same time, then you could do to this one for the play. Oh and then one for the configuration option.

G

Once you grab the play, oh you found a dress deal stuff, and then you ask for the configuration option and then you join us ripple is a router, so you could do things in two steps now joining us a host for which we need to pee. Are you joining us, a router for which you need the configuration option.

G

Phone- it's like that, so I might be that it works, but we don't have all the details on the hope so intelligently. The device can can use that next slide. Please I think we're close to the end yeah. So so that's where we are mostly where I am because I did not get the feedback. So I really hope that during this go, I would get at initial feedback. Saying yeah, that's the sort of thing we wallet and if he tears, then we can discuss on the mailing list.

G

You know on how we progress in this, but at least is it what you guys expected.

A

It sounds good to me.

B

Yeah, we were right, though, season two, the mailing list and now trouble questions. There's no question doubt more for work. We have.

J

Yeah I think I think every room is. We should start.

J

Yeah I break the door before, but I think a lot of things have changed since then, I mean I read the darkling special it was anyways coming back to so the this is about attack and the last shooting was calling it rotate. Now I was touring more acts, something like that's decided to call it loot and for the lack of better name, I guess well.

J

So the primary reason why we have this is that we have some way of and to advance establishment indication for the node that no Trinity they have traffic, but as it was noted in the last train trip, this can be made use of other purposes as well and the other purpose when a further purpose is sedating or querying molds capabilities. So this is the only message in storing mode of operation which can directly grow from the root node to the target. No next like this.

J

Okay, okay, so we have a gay flag because that this gay flag, essentially is other, is set by the target node in the tli Oh, based on which the root AK is generated by the root. So one of the things that we discussed in the bastion term is that we don't need this. We don't need the root to send this root AG for all the towels. We have a flat for that. But then the problem is the intermediates Excel. Ours, don't simply forward the towel message when they get one. So there is a time difference.

J

There is a delay tau timer, based on which the tau is.

J

Generated again from the intermediates Excel ours, so what this means is that the K flag has to be maintained by the intermediates Excel ours.

J

This would be very much similar to how the 'flag is is handled correctly by the oldest Excel ours. The only difference is at any point of time. If the 6lr sees that it has received the TI o for the same target with key blank unset, you can just unset the K flag for the corresponding time it can just unset for the key point for the corresponding target.

J

Essentially, what it means is that it has. The 6lr has received a key flag as unset, because the target has already received root and what we are the other, so that should be. That should be good enough handling. In my opinion,.

G

Hole, don't you think that's the way to specify this.

G

That will be many whack coming Hudak, actually coming from the root cuz, you know, I sent it down with a flag for myself now for some reason, my parents, since these down several times and if I don't you know sound a lot of doubt with a K and set that could be a good number of of copies of my dough somewhere above me, and even if now, I sound down with a k flag and set by the time that the case like populates the root I mean the root will have gotten the dowel several times.

G

So if it's a root AK when there should be a sequence counter or something so that you get anyone so.

J

That is a sequence counter Pascal, so the path sequence. So if you see you know what what we are trying to do here, so the path sequence make sure that the target node knows the route daca generated, published.

G

The route gets, Badal are given pass sequence. If there is a K, you will Sonic and the route and then until the pass sequence is added, there will not be another route AK. So if the child doesn't get the route AK, he needs to increment the path secrets, even if he doesn't well, you just do so. You will increment the passage once and then before we set the key flag. Okay, so sorry.

J

Yeah, so this is one of the primary design point that had to be taken care of you know. So what happens if the route AK is sinful?

J

It should not matter, frankly speaking for the problem statement, that we are that I initially imagined for the into it traffic establishment. It does not matter, but it would it is. It is good to know that the route AK is initiated for which so yes.

J

G

Was one of the discussion that.

J

One discussion I had offline with Robbie, and then we had had some feedback from Pascal as well online, so there is RPL unaware relief and it's making us a far beyond network as a transit network, and in this in this case, even the rules needs to know when to initiate the application traffic, and there is NSA and na ero, with the arrows that are sent in an aware, leaves charmed. So Scalia suggested that any ero should be generated only on getting through tag back from the root. This I think is this.

J

This is a completely stateless operation, so it should not be I mean from implementation point of view. They should be okay to handle, however I'm just wondering so if, if I simply make changed in this draft, will it be okay, because if you a Canadian ever leave graph, basically sensor innocent gets any ero as quickly? So what happens?

J

I mean rule does not have to change, but is there any I'm just trying to think? Is there any dependency on the another leave draw for this.

H

One point: let's say if if we send the arrow ACK based on the into endow work in that case, maybe if there is a negative acknowledgement form from the route, maybe we have to send a new status in the in the arrow in the.

J

Hanok village, sending.

H

The arrow we have to added new status information.

J

New status, as we here already has the status right so you're saying that status has to be updated based on the rotor. Let's.

H

Say dog felt because of something it will get a negative acknowledgement from the root. Okay. In that case, we have to send a negative mean City, as it is a negative arrow, we have to send right right we're sending a status back to the AR ùl in the ero option. We have to send the proper status why this registration felt. Maybe there we need a actual reason between maybe that status field. We are proud there.

J

Is a status already in here that.

H

Value of whatever is defined in RFC that this is not handled this part well,.

G

Between the messaging by which the see in this picture gets the prom and what has to tell the rule what he has to tell the real is one of the statuses which shall specified in the Paris, see that defines the inertia. No, it's not free Porter defines that it's a by four five, so missing something they try for five, then we'll have to update it or something, but it's not because it comes from a root AK versus D Co that it makes a big difference right.

G

Normally, the root AK is just to tell you a your rot is enabled it's mostly like. If you receive it, it means okay, even if there is nothing in it, Wow it just means. Okay, the root can reach you now, if, if there is a problem, normally the root which would send Anderson condos DCO, so that's that's how I said the root AK is more like confirming the Tigerettes.

G

The is a unicast. If there is a prime from the root fun, sensitive state in B would be maintained by the root at. Would that clean be? It would just an see if you wanted to clean C, but if you really want to clean B, then the only thing the root can do is under this year.

J

H

Respective of root are fair, is success or failure. This registration is going to success for re well right.

G

Yeah I mean which match is it's just something to hold home? So basically, I see is ready to send DNA euro, but it has set the K flag. So now it needs.

A

G

Hold sending ten euro till he gets the root AK when he gets to attack it's just the unlocks sending DNA euro, but the root access just confirms that the connectivity has been established. It's not supposed to mean anything for now, because, as a separate discussion, we can talk about capabilities, but for now is as a specified who just mean. Okay, the dollar was calculated Allah with root and that's it so the root can achieve.

G

An old in DNA, but DNA could have been prepared already do is not to give a bad status at that.

J

Because we would not know that's right, so if you actually, the loup-garou has to be cleared up for some reason that you see has to do.

J

One more thing that time.

J

One more thing that has to be clarified has passed out. Part of the slide is that currently in the draft, the route ad carries only the transit information option when now, with external targets, the root ak nudes to send back the target option to that to this to the 6lr external target, the reason been other to know for which external target was this back interpret.

J

Yeah, that is I think this should not.

J

That is something that has to be explicitly additive next, like this.

J

Okay and then we have this capability, query I mean I just wanted to depict a picture saying that now there is a way in which route has to route can say: ascend and a synchronous route act. Now. This is where you know this. Nomenclature of route back does not make sense, because in this case it's no more an act. It's a capability query. So this is where I wanted to see if we can come up with some other name, but basically here we are a synchronous, Rohtak and then we sent it out with the capabilities.

J

So this this this, this point can be handled like this.

H

This capability capability see, you would have shared with it, be already right. Yes, this.

A

H

Is further, like you use in your Dow projection kind of thing, right, yeah, so.

J

So, let's say, for example, route wants to know what she has, what see suppose so, if, if, if root wants to know, if we let's say, for example, Road wants to install a route PDI project at Dow, then group needs to get some information from C. So this is one way to get that information.

J

This is the way to get that information, because this is the message that can be directly sent to go to send to the targets all right.

G

Yeah last time we discussed I how I thought that this exchange would be a new message.

G

J

It's it's a new message, but I'm not sure if that aim is correct. So if you're talking, so it's it's a new message right. Oh you made it, but then the dow is not.

G

And for me the new message would be hood: sending a query and the see sonic battle response to.

G

The attack is only needed if there is nothing else like if, if phone sensor on the situation, where each time there is a doubt for which the route doesn't have the capability, you are sure that the route will come and query the capability or if there is anything else, then, if, if C is 6, long and East does that back exchange for node he will get the the the DAC back, meaning that is reachable. So so, if you remember the way, neighbor discovery is written. There is something that says that nod is only needed.

G

If you don't have any parallel information telling you that the connectivity is still going on same thing. This is who tack would be only needed if there is no other message. That's going around that shows you that you've got connect.

G

So essentially in this query, then you don't even set the cave yet because because the core is coming- and that means your original.

J

That's right yeah, so we want a separate message for this than there so yeah.

G

J

Okay, so we can completely a couple. This Easter points that yeah, so this can be part of the capability Draft rather.

G

This trust would say a if if this or anything else is going on, that makes sure that C is reachable, then she doesn't need to suffocate it.

H

Okay, yeah I. You scenario this row track is required. One we win. The node joins the network first time or did you did some parents which, in those scenario, maybe the route AK will be used? What I think.

J

For the parents, which, what is the reason to Ruby I mean our previous intimidation. We use that, but I still think maybe for the parents, which move.

H

Because I have switched a path and I have to know that in that bath it is reachable light. We.

H

See I am now connected to B because of something I'm more switched to see if switch to a but Maya a I might not be reachable because of some issue.

J

Actually, this is what we did in our implementation.

J

That's that's all like I, think I think this is the last leg all right.

B

Thank you very much question further questions.

B

Now you mentioned that when the basket joined you will like to back some to summer slides if I understood correctly.

J

So what is what is? It was if Pascal is not bad, then I can present. Some of the points with regards to alighting are killing for, but he has already done that so.

B

Okay, we will collect at some point for this meeting and we will send with the mailing list with a link to the recording. So do you would like question you would like to meet in June.

G

We have two on working groups. We have to request the meeting for virtual right. Yes,.

B

There was a discussion beginning. A working group would like to do like metering I 1 through 8 or later, because.

F

We are going to have.

B

Two slots I mean we discussed it with one slot would be now. If we meet with an IT f10 I have.

E

An engineer two.

B

A

Split opinions on whether having between during that week or having a separate information, what's your opinion like? Would you like to have a meeting with in the IDF which, but can, or rather not.

G

Necessarily I really love the way we did last time so one more so we could find the best time for logical.

G

Indeed, everything is packed is because we travel otherwise doing what you just did for finding this part across lot is very perfect, because we can really fight where we get most attendance.

A

G

I don't mind it's that we go another week: okay and.

E

Two hours is fine right. Two hours I mean no.

G

Yes, just many Fotis just do what you did for 107. It was required.

C

B

You very much we're will anyway request. Comparing with the many list I'll collect all the information.

B

Okay, there is no further questions.

G

Please come back to me on the alighting thing, as I still didn't get a good sense that the group wants it or not and well I invested time on it, so so I would have to see if we want it or not.

B

Okay, we'll stop recording if there is no further questions now.