Internet Engineering Task Force Interim Meetings, 18 Oct 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF-LPWAN-20221018-1400

Description

LPWAN meeting session at IETF
2022/10/18 1400

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

A

B

A

Two one two can: can you hear me, though, yeah.

B

It is a much better much.

A

Better hello, Kirsten, you know Justin.

C

I think the custom is, is just a sleep user.

A

As you refuser, okay.

C

A

Oh yes, yes, yes, the chat here. Okay, hi Kirsten I mean children. You wonder at that time and uh I said you.

C

Said you is life.

A

So uh Maine, how did you do every every time I'm still to do it.

B

A

B

Let's see how you have you have sentiment, uh slides I've, seen with some some slides.

C

Right, yes, I did some ugly slides in Array, because I saw that we, you put the topic on the edge on that.

B

And I don't know if I need to join to be seen as uh in the minutes or if it's just. You are already there I'm here, because I'm a little bit yes, but uh will that remember me, you know for the Blue Shields foreign.

B

I expect you are.

B

Because I have not joined the.

A

B

So yes, I mean the chairs are in person.

A

That is a very, very special meeting, also preparation for the next meeting in London right.

B

Where the chairs will again be in person, it's new since the beginning of Korea is 16 three years three years already well, I guess: 2019 and 2022 yeah 2020.

A

B

Because I had 13 19 so 19 we had the real meeting. Yes.

A

And I was not there because of uh the birth of Lana in July. I was not there well.

B

In in the fall I guess we were both there, yes and then cooking it. Yes, that was early 20. So all year 20 there was no. No, you have 21. There was none and.

C

B

Must have went.

C

To the Westland yeah.

B

A

B

It's like this used in United States, so um with the we will be waiting two more minutes for the for the people to join. Do we have candidates for taking the minutes?

B

Thank you. So much.

A

Thank you. Thank you, Yvonne. Thank you.

B

I will be taking minutes as well. I see that is not with us, so I don't expect her to take minutes and Dominic will Joy played. He said it will be the second half hour, so we may retain the discussion on the component to the second and forward. Yes, yes, that's very good. A few more minutes.

B

So yes, today we we don't talk about drugs, I mean our drafts are doing good progress. The big problem we have is architecture.

B

We need to understand which ideas we need which operations we need for the rules and that will go to the architecture and then, if there are drafts that are needed, we'll write the jobs, but we need to figure out exactly what we need and how it would work. So it starts with architecture.

B

D

C

A lot of people from the group will be in London. Maybe it will be great to organize a side meeting to oh, yes,.

B

C

It will be before the meeting.

B

Yeah there are a lot of interesting side meetings at this idea, so we'll have to navigate that's my unique, also I, know sustainability. For instance. We are very interesting meetings on sustainability that, if you care about iot you care about energy. If you care about energy, you care about sustainability, so I guess the the guys at ATF who worked on on iot could help a lot. Other groups thinking low energy, because many Protocols are not designed follow energy, for instance, with protocols which was broadcast.

B

They are not thought to follow energy, because if you have to listen to broadcast all the time, you cannot see yes and she broadcast your consuming a lot of spectrum and a lot of energy and Wi-Fi. So so it's sort of thing that the iot people should, you know, explain to the rest of the ATF, maybe how to design the product low energy protocols through this. That's a start.

A

B

There is even the table two chairs and details.

A

So yes, with this, uh let's get started: I will go ahead with the with the slice, so hello, everyone. This is, and in an interim meeting of the LPN working group uh with uh chart by Pascal and me so both present and uh with the presence of our A.D Eric hello Eric. It's always a great pleasure to have you.

A

uh So this is an uh an official IHF meeting and as such, the ITF policies apply uh so by participating the ability for ITF processes and policies and all these policies apply to you. So this is just the the summary the note 12, but uh you should go and read all the BCPS that are uh presented here um and uh typically in case you are aware of any IPR that is that is mentioned during the call.

A

uh You know you should either abstain, presenting that IPR or should uh talking about it or you should go to the to the chairs after and uh uh and uh and and tell tell us about it. uh Of course, everything is written in a much more uh explicit way in all these BCPS.

A

So with this being said, uh so we're taking minutes. So thank you Ivan for uh for uh leading the minute taking efforts, but of course we'll all be contributing to to that. So you have the link here, uh uh I'm, not sure. If uh Pascal did you fix the the link because uh in the no no okay, um if you click on the link, it takes you to a different uh URL. So it takes you to a past I mean to type.

B

Yes and make sure it's 11.

A

Yes, you need to type 11., so I'm going to post in the chat room, the the correct menu taking uh a link goes here, uh and uh so the meeting is recorded uh and the presence is automatically captured by meter Echo.

A

uh So uh just uh I'll take just one one minute uh uh before moving on I'd like to um ask for approval of the minutes that were taken during the last meeting. So does anyone have any corrections, additions remarks uh regarding the minutes that were taken last time and that were put on the data tracker.

B

I hear no objection, no.

A

Objection so they are approved so moving on, so uh for the agenda for today, pascara I'm, giving you the yes.

B

Before we, we took a little bit more time than usual for the administrative here, because we'll be talking a little bit about atf-15 and then we'll move on with the discussion that we started last time um and I guess Laura will take the lead of the discussion. But basically um we have isolated the number of different issues.

B

um Sometimes we need to isolate which device we are talking to. So we can apply the right set of rules now. Usually a set of rules is command to a good number of devices, which means we need a rule ID and then, which gives us a rule set which applies to a group of devices. But then at some point there are some values inside those rules which need to be instantiated for a given device.

B

Like is IP address, so we need to discuss what is really the device ID for what is the rule ID for indicates which set of rules and if the rules evolve to begin sequence number in them and then the rule instantiation and and that's the thing we started discussing and I know on our side we have a little bit of IPR on that. You know using some kind of dollar sign uh to replace stuff like the idea of the device. So if you have the device ID, you have this IP.

B

If you have his IP, you can replace the IP in the row like you know, IP or something and I I know that there might be some API in there, but it's it's kind of a necessary uh function. So we need to discuss exactly what is needed and how it would work.

B

And so, after that, it's going to be aob, so a sliding Shield. Yes,.

B

A

Yes, um so we have uh the Chicago Ziplocs working group last call so Anna is not with us here, because uh uh she is uh I mean she's, not here with us here for health reasons,.

B

It's not big, it's okay, don't worry too much.

A

Yes, so uh that's uh that's! That's so! That's for, for this part, I think that so I did not look it up. I mean I'm not going to speak more about this Anna is going to provide the the full uh um overview of what's happening with the Looking Glass call, but it seems to me that everything works fine on the compound act. So I was waiting for uh the IPR uh confirmations from the authors, so I have everyone except Leon that have answered.

A

Oh this one. Unless okay.

A

Please discourse all the.

B

A

B

It will take time because you have a lot, but yes,.

A

um But and that's uh that's basically, the the last part of uh of this.

B

uh It's actually two drops every time right. It's the two seat, Fox front. Yes,.

A

So uh yes, Sergio, can you hear me? Yes? Yes, just one note in the compound act in the young model, there is some sentences that doesn't fit the the page, so I think there are a little bit of the need that says that but I'm not exactly sure how to fix it. So I don't know if, if maybe Laurent can help us check that out to see if we can compress that line to fit or or do something similar to what they did in the young model.

B

Not always the expected.

A

C

Yes, I am now I can reduce the lines. I know all the tricks to do that.

A

C

A

What that was from the shepherd uh write-up that I wrote right, I did send it right, I do I said I sent it and that these are the needs. Okay, thank.

B

You that's local spent six months in Amazon, yeah reducing ads, and now we can make uh the young model fit in one page, foreign.

A

Module, like the name, is it's just. How can I say the way the the young module name is defined? It doesn't match the name within the module itself. I mean it's a small Niche, but yeah. You can also fix this one. If you're going to fix the length as.

B

Well, yes, and yes, we have to fix both and both to the void, commands. I mean I. Think all the isg members give you the same command is a waste of their time. So please don't worry about you know, uh including increasing the number and resubmitting it doesn't you. You have 99 variations versions. If you like to fix everything and republish to automobiles.

A

I I there's just one question that I asked about the the Sheikh the compound deck and it was not reviewed but uh by a young. So that there's a specific question about that, and it seems to me that the mod is not big, that it's quite straightforward and I compiled it. It works but like in the shepherd's really specific question like was it reviewed by a young doctor and you have to say no so I said no, but blah blah blah but I said no. But.

B

Now you became our biggest expert in the end. Have you looked at the Young in the component? Could you just have a quick? He learns at it. If it's you know, in line with the way you've done stuff in your your document.

B

Sorry, could you please have a look at the little piece of here?

B

Yes, just for the spirit you know, so it feels easier when you go next to your work.

B

So that's cool well done. Okay,.

A

B

So for the Chica versus it was concluded yesterday and I got only positive note, but I must say that it was only from the authors. um You know, my impression is to group at large. Doesn't care too much about sick rocks. We never got many reviews uh on the sequence document, so Eric, it's for you right I mean the it's it's kind of an internal effort because between uh the parties, which were interested in sick Fox, which is like a group within the group and they they seem happy with the work they produced.

B

um I guess people from the group have read it. They are. You know they are happy with it, but it's just a subset of the working group. Basically, so that was this question I guess in the shepherd which says. Is it like a cork team interesting? This is typically a golf team interested so I mean you know. When you promote this document Eric, you have to know that it was really the work of a specialized core team inside the world group.

B

Do I make sense.

A

It does make sense, indeed Pascal. uh Thank you like pretty much many documents coming out of this working group right.

B

B

No, but a sheet, for instance, was done.

A

The best document- yes that for sure that was much better, but everything which is linked to a specific. Let's.

B

A

B

But as it goes, I guess was raised more of generic interest and also it was the first one so so that I think this problem is exacerbated on the sixth notes. We are only readership thoughts. People have contributed.

A

Yeah exactly said in the shepherd's review and just be honest right right, I mean that's. There are six authors. There are a lot of people in involved in the world. Yes,.

B

So that's the question. Basically it says it's the work of system. It's not one right! So it's good! So it's a group within the yep, so that's pretty much it so basically, I did not get a new position, and so, if you look at the wedding list, I mean it should pass foreign.

B

So this is this: is the email that I sent so usually I, don't present it, but this time I will that's my usual email I just charge a thing here and there um as it was. um They have changed the agenda link. Now it's agenda without that HTML. If you try that HTML it doesn't work anymore. So it's just agenda, and uh so we have a November 11th and that's the very last meeting so from noon to 1, 30., so I hope it's! Okay.

B

With your travel arrangements and for the meeting it was a bit tight, but it fits because it's earlier than it was in the first round of the agenda. We have been moved to 12, so it's easier for Dominic to be with us.

B

um The usual suspect our race expect us to reserve the some room for them. Don't expect that please look at the dates inside this email. We need to know if you want a slot and we need to know before the agenda cutoff.

B

So the dates are there, uh the the engine that cutoff is on the 26th, so it would be nice of this month. So it would be nice that you respond by the 25, so we have one day to organize the agenda. We have one hour and a half considering the work that we have. It should be enough, uh but please tell us about the work of the discussions you want to go after in this video and as usual, we have the questions. Rough name present per Etc.

B

If you don't have a draft uh for things like the ID, it's gonna, the draft will be the architecture and also, as usual, we have uploaded a slide top plates. It's already there, so I get the link of the slide chocolate and we expect the slides of one one week in advance.

B

I know it's a wet dream, but we would like to have the the slides by November 4, if possible, so we can make the slides available to everybody on the agenda with the ATX, and so we were just talking that for the first time in pretty much three years, the two chairs will be normally present in London, which is yes because.

A

You could do.

B

A

B

You're still welcome to to sit near the table, so we actually thank a whole lot. For you know his help with some of the chairs were not there, but this time now you can enjoy your time with less stress, but still we, you might be needed again so stay in the starting point. Foreign.

B

The shake uh you can see it's in the RFC Victor queue. So congratulations uh to the authors.

B

Obviously, for this work, so now we are waiting for the Galaxy number, the well first, the editor to fix some things, but then we expect soon to be to get our new RFC so the whole congrats.

B

Now, if we look at Chicago India UT, it's a lot of green I mean I've. Rarely seen documents passing so well, and so is this through the ASG. Okay, so I'm impressed and uh congratulations to the authors as well congrats, except that the ASU recognizes it's a very good document and I expect that it will pass quite quickly considering the all the green darkness in here.

A

Which in Pascal, if I may interrupt you and to add some more information about the nviot, U.S, Center liaison statement to 3gpp and I? Don't think you you you? We already share this information with the working group already and we are waiting for a reply.

B

Do you think I mentioned it on the previous interview, but maybe.

A

B

A

Than done then, yeah yeah, but it's very good repeating the this. Is it's not bad? It's so.

B

It's one of those rare occasions when, when the first time in my life, I have to do a earlier statements, it was like wow. How do you do that? Thankfully Rick helped a lot so yeah that passed um so so yes, I mean she covered. The iot should be there.

B

So the the PCS which quote-unquote should be standardized by three gpp are just informational when the end to end is uh normative, so only the end-to-end pieces is presented as normative in this errors to come so with this leaves, you leaves us with the two Sig Fox draft to promote, and if you look at it after that, it's just the architecture and actually all the drugs that we have OEM Etc. They they are all like expired.

B

So for now the big document that we have in our plate is the architecture, and this is actually his introduction for what we're going to discuss. That.

A

Exactly so, with this being said, then: I'm switching to Lauren slice.

A

B

B

So now, how could you please enlighten us on what the problem is, which identifier or identifiers we need, and also this problem of investment Etc? That I kind of hinted from the body shops.

C

Okay, thank you. So here is it's not solution is just questions and from implementation you have, we have, and also that comes from uh the young data model, so uh I have no solution. I will just put some ideas that I had and also that we come from from the group, so I don't think I can change the slides.

A

B

C

So here it's uh what we we have currently that we have several devices, so we have this start topology and uh that good I got the control, uh so yeah Vista topology, and we uh we have devices, that's around the box here and we have a single core that talk with all the devices and until you know what we said is that we have rules from all these devices that are stored in the rule, but we don't represent it here and also on on the core.

C

Okay, so that's the status we we have right now, at least in standardization. So when a device sends a message is send it by its full ID.

C

So the core are first to identify the device to find the appropriate rules, because a rule ID can be and view differently from every devices and when she has rotated the rule, he applies a decompression or the fragment the assembly and we'll send an icmp iacmpv6 message to an application, and this information, usually the F A, is stored on the rule and application can be something Dynamic but can also be stored on the.

C

So one question is, of course, how we are going to uh to identify this rule internally to find, which is a good word on the opposite. When the car receive an IPv6 packet, so you will look at the destination address or sorry it will get the IPv6 packet and, for example, in open Chic.

C

What we do is that we look at all the rules and we find a matching rule because for the destination address match with destination address in Ruby and getting that we are able to identify who is the owner of this or what is the device and we are able to uh to send a message to to that device.

C

So that's how it works in open, cheek, I, don't know for for our implementations, but it's a way to to make the communication in in both ways.

C

So what we have said also in during uh talk that we can imagine a more meshed environment here, is it's not full image? We are still core uh things that can be identifier, score and device that can be identifies like devices.

C

But, as careless said in a previous uh talk, then we can have something more complex because in fact where we are in mesh environment, we are not uh uh um we, we are not in um device and core, and so that's a problem and I know where now is my copy past, it's it was on on the side. uh So what we have to do.

A

Just a second because he had a question just a minute ago,.

C

B

Well, that's okay, because we didn't want to interrupt, but um so basically what what you said on the previous slide was like. You identify the rule get off because of this nation IP. So basically you have a business IP, yes, and- and you said basically, what you needed to find now is how to talk to device or something like it's Market drives, and you have a table somewhere which resolves IP into Mac like take advantage of an ND cash right, even if it's not used by ND but I'm.

B

Looking at the architecture, so I'm trying to figure out what else are they figure that there is one component which matches uh an ID in the IP packet like this National IP into an ID in the lower or lp1 network like a MAC address and and that table is not part of the rules? It's it's a social machine.

C

In open cheek, it's part of the world so.

A

C

B

I should not should not have to say this right because I don't see that implementation and you have collapsed, DND cash and that's why that's exactly what I wanted to talk about so I I, don't see that the rules should contain the MAC address of the device, because it's.

C

Not something the rules that contain the MAC address of the device, it's something above the rules.

A

B

You have the set of rules speaking it's a different table. It's like the NT cache.

C

We we can see, that's a that's a good point, it's something that is not specified in 8724.

C

um We must uh maybe not say how it has to be implemented, but we have to say that we we need this identifier and exactly.

D

For example, if.

C

Currently, one problem we we have is with the no compression, because when you have no cooperation means that you have no match, so you cannot find which device has to receive the information so well.

B

Then the cache, which is a separate table from your rules- you don't have this problem, so so it's it's kind of a a prime that you created by yourself. If you collapse to two tables, basically,.

C

For example, if you match a rule, if you put a mapping between an IPv6 address and in a device address and that your rules is not the correct one, then here you have trouble, so you have to check currency between the rule and this table.

B

uh I, don't see that it's an issue either right because I'm trying to look at which tables you have so you have the rule. We said we have something like the ND cash and now you're telling me.

B

You must also have another table which says for this device ID, it's this rule set and it's a third table which, because initially you said I need to find to have a device ID, because for this device, I will have a rule set set of rules yeah, so that you also have another table I'm trying to look at the architecture again right when I do the drawing I need to show which abstract databases you have one of what is the rule itself? uh You have a number of rules in the table of rules.

B

Okay, one is the ND cash and now you're telling me there must be a table somewhere which says for this device, maybe you're notified by Mac or something I have this set of rules.

B

Then you can validate when you get a packet coming in and you match a rule you can validate and and so so. Basically that gives you the rule. You you look at the IP, you get the device ID, you look from the device. Id you've got the whole set and you're very validate that it's the same. So it's okay, you're good.

C

Yeah but then we have our Direction and the other direction. What you you have is that if you receive a today be a new message to devby, so you select here uh who will be, but in your if you have a mapping somewhere, let's say that uh Dev B is uh Deva via.

D

C

Address this B is a device a then here you have a trouble because you send to the wrong device. So you need this information. I, don't know if it has to be on the table. If it is, the table is implicit packet.

B

Oh, so it's the same direction as before. It's coming from the internet yeah, it doesn't IP address in it right. So sorry, okay, so the IP app device B, you look at the the the DND cache you find the MAC address of device, a that's. What you're telling me.

C

Something Dynamic that can be, uh but what I say that I mean it has to be coherent.

B

Yeah yeah, we can check it, but if we have the three tables we have the rules, we've got VIP to Mark and we've got the Mac to rule set. We can effectively check find validate that the rule that we apply on the incoming packet is actually in the rules set that matches the magnet class we can validate.

B

C

Good, so it's something that yes, I, think it must be in the uh in.

D

Both directions, the.

C

Rule cannot generate several if it's a same device normally can generate only but to uh yeah to the same IP address. But it's maybe it's not true all the time. So, but that's something that we.

A

Yes, this is what, from from what from what they understand. Yes, there is this so I I like very much at the idea about the ND cash.

A

Exactly it's a mapping table then there is, there needs to be this coherence between the the compression rules in that table right yeah, but it's.

B

You can validate it, so we can. We can show this flow that the logo has been talking about that this moment of times. We we so the rules we understand, then we have the young model the mapping table between if the device ID is the Mac. If an 8K Mac it's a generalization of the ND cache, so we understand it, and so the the real question we are discussing is: um do we need a device ID to index the table uh to know which rules apply to this device?

B

So so that's basically because otherwise.

C

And another Point, that's what I was hesitating is that a device can be a router, so it means that it can have different elements behind them. So uh we have so we have to take care, maybe it works, but we we have to take care of.

A

That, yes, so so the point you are saying is and why I understood this was, if you apply the, if you apply the rules, maybe you can say okay. This will match us so so, based just on on that, you.

B

Say: okay, that's that's what open G does, but it's my my way of looking at it would be I'm getting a packet before I match any rule. I go to my ND cash I find the device ID. If I don't have a device. Id I you're lost finished. Yes, if I have a device, ID I know this device. I should have rules for him. So I take this table device ID to rules, I have a set of rules and I only try to match those rules.

B

They match a little, but I would I would not match. First I would match after I've found from IP to device device to all set and look at those rules. That's why I would look at it. So you can. You can never have a rule which matches for the wrong device, because you will only look at the rules which fit the device. That's.

A

Right and if you want to pass through where you just have I mean you, you say: okay, well, anything that I don't know I sent to the next router, which can have many.

B

Many neighbors, we don't have rules, you don't destroy right.

A

Then I mean then you need a protocol from the other router to tell this. Router, hey I, have all these Neighbors.

B

B

um We were talking, for instance, on the rebuild case with careless and with ripple you always have a tunnel to the root. So say you have. You have the lower device. You have the first XLR which is attached to the Roja device, and so the raw device itself is not triple aware. It's a ripple and aware leaf. It can talk to the first XLR.

B

The XLR will take everything he gets turn on to the root, so the route will terminate Sheikh if first it decapsulates the APD IP inside it's a Chic packet so from the tunnel it close to 6lr but not the device. So that's when we did the device ID, so the root can effectively find which device you know is on the other side of that tunnel, because it knows the 6lr, but not one of the many devices.

B

So the repo the repo domain would be exactly like you haven't spoke, you get the root and through the tunnels every lp1 device appears, like you know, one half away because the tunnel was just Ripple itself is not visible. So we we need a device ID. So we know which rules are used to compress the root will select.

B

But when it looks exactly what law is description.

B

But the device ID needs to be in the packet. That's.

A

My point exactly and then there is one one other question: it is uh so related to Laura. Again your previous slide.

A

Sorry I know you want to move further, but so here you put that the IP address is the identifier, but uh is there a possibility that no, but then it becomes like a nut, I mean if you look at purely from Chic perspective, you can have the if you apply shift rules, I mean the way uh opencheck works today. Theoretically, I think that you can have two devices that have the same IP address and that they only differ in the UDP port.

A

B

Say yeah, so can you if it's like having a PC which has the IP address and it has different USB ports and the port points to the device? Is.

C

Not so obvious.

A

As a which which which you can do, if you apply the the ship, the ship compression but in us in a neighborhood in ND uh neighbors table basically.

B

Tables should have not a media AP, but also the portraits index, and so you can find the right device. Id, I guess.

C

I think a device can have several address, but an address yeah can, yes, it can be shared between several devices. Yes,.

B

C

B

That's the usual meaning of the poor tissue. Usually you get serial on a PC and it would indicate which is that on a certain Port, you get on the router on the portable router, to which football demos yeah.

A

Let's switch it to us now, but then it can be that the table has duplicate entries with the same IP address and the same and different devices, and- and basically that says well, hey can match to any of these. So you need to apply ship later to discover which one you actually are going to go later on.

B

Okay, I: don't like the idea to use the rules as the way to go. Resolution I think you should resolve which set of rules work for this guy and then find one rule in that set. So with the indications take this initial IPS, this is one device for IP. You know business type, key best, ID device, ID, Pro, Set, now I, but choosing the rule set as the discovery mechanism.

B

I actually have a pattern. That does something like that that I I would not use it there, but advisor know for sure which set of phones apply first and then, for instance, if you say the port is significant, then you're on your table, instead of having just ID you'd, have a deeper spot, I mean.

A

B

But still at this table for me, when I was doing shake, it was more yeah. Actually, I was I, have a pattern where I use Chic to route yeah, because, basically, if you it's more of a security thing, if you match some sheet patterns, then you go to the firewall. You know that sort of thing so.

C

Okay, uh so let's so, we have to maybe write something to understand to discuss about what is a good proposal, and what I was saying is that when we we have a mesh Network. So here we don't have any more core on the device.

C

We have no more up and down ascale's States so, but we can continue new to have this notion of uh up and down, but it will be very limited in to some to some traffic, and so that's why I try to represent here is that we need in to identify the rule by an element that will be, let's say, the core and another element that will be the device and if we have communication that has symmetric. So here we need to to have to identifier.

C

One for one is occur and the other the reverse position and, of course, to avoid any confusion. It means that we cannot have the same rule ID in both entry, X, Y and Y X, so it was what was uh described in Ivan's draft.

C

So on this way we can I suppose apply a Chic to a mesh environment.

B

So just two things in the architecture: we try to say that there is always a form of Transport between A and B, so it could be just PPP.

B

uh It could be a repo uh ip90 tunnel, but there's always something to transport between a and Y and is that something is created by a or it's created by y, and we said basically, the the application is always the guy who starts. Who creates that thing. So.

C

But we have to identify we, we play the role of the car, which will be viewed as uh what's.

B

The application.

C

If you look at X and y's is here now quite symmetrically, they Exchange.

B

C

Terms of three or three, and if you have a and if you have a graph I think Alice can complement this. But if you have a graph uh it was a white egg to be B, take to a and a take to Y. So you are both core and device so.

B

The the way we we have done it so far was to say that Sheikh is transported into something. Can BPP can be a tunnel can be anything yeah and then your TCP session? It could be whatever you like, but someone would start that session. So it's not pure IP over a multi-hop.

B

It's always like a point-to-point tunnel, yes and or transport, and somebody starts that thing.

C

You have several elements you, you can have a loop in. In fact,.

C

B

Matter, it's it's you all of this point-to-point connections. They live their life, so so you it doesn't matter. If you're the core of X and Y is Yoko, doesn't matter, there's no chord. There is an application inside and and the device side and for each point-to-point connection the device will be the one that starts the connectivity that stops.

C

That I don't think it works so well because, for example, we don't have source and destination address, we have the they have a Dev address and application address. Yes,.

A

So countless wants to say something.

D

Yes, hello, so yeah I I had like a similar questions like the ones that uh Lauren is now making and uh somehow I I thought like well.

D

um The the way to address all of these is that the the end points that will communicate need to know in advance whether they will be acting as Dev or up um in some cases. uh Perhaps there are like different uh it relationships here like different interactions like node B, uh maybe using one role when talking to a but on a different row when talking to Y.

D

So perhaps one option is that the nodes need to know in advance somehow, which will be their role on talking to each one of the other potential endpoints and also perhaps consistently with this. The the rules should be written accordingly, so that the notes should know in advance, which are the rows when talking to each specific endpoint.

B

Yeah, that's if you'd want to do Chic flak flat over the mesh, but if you, if you establish a transport layer like a I, don't know UDP between a and Y. If, uh if, for instance, the overall I have a demon waiting for some connection like this or it could be PPP, it could be anything as soon as you wrap into something, and somebody has to start with something. Then then you know who started it and it's the application. It's the device.

C

No, no, no I I disagree because it's not the applic. The position is not who starts.

B

C

Don't you just receive a packet, but you have to know: do the other guy act as a car, or do they have a guy act as a device.

B

That's why I'm saying you don't do over IP, because you would have that problem. You do Chicago point-to-point session between devices and you do it inside the session and when you start the session that breaks you the device but.

C

We, for example, if we do it over a sixth row or via B, we don't have this notion of station.

B

If you, if so two cases first, if your network has one, then you don't have a problem right repo. There is a root, so you always know which one is. uh When you have a real mesh, basically, two devices you talk, they need to know they have to talk, and basically it's not just a random X talking to random y.

B

That needs to be a reason why X and Y took, and they need to kind of know that they will do together of some form, and so my suggestion for the case, where you don't effectively have something like repo to say who's. The root is that you expect that there is a session when you you're saying you don't have a session, they say if you don't have a scheduled, that's why we need to have a session.

A

Has something to.

D

Say yeah, sorry, so if I may um I think that perhaps in some cases it's possible to know in advance who is going to start communication but perhaps not always.

B

No I'm not saying you to know, but somebody will start it and once somebody has started it, whether it's a and why I don't care. But.

C

B

Game starts, it becomes, plays the role of the device.

D

The road to be written beforehand right so uh once you yeah you, you know, which will be the role for each device, um so I'm not sure that this would match. Like always,.

B

Sorry but yes, the rule needs to be written in advance, but maybe the IP address of a and the IP address of Y should not be written at the time. You write those rules, it should be saying dollar device and Dollar application, and once you decide to play those rules between a and Y, then you would reply and, and the session is starting by a meaning- is the device.

B

Then in every rule you should replace dollar device by a and dollar up by y, but it can only be done once you have decided who plays the role of device and who plays the role of the application because effectively otherwise we're in the loop I agree with you. If, if you write the rules such that a is the device, then a has to be the device you should know in advance and you force a to start the session. Well, all good.

B

But if you don't know we'll start the session, then you can only instantiate the rules once you have started the session. So that's that's why we add in the title we have this concept of instantiation. So basically the rule set should be the exact same. All those devices are equivalent right, they're, all the same same, so the rules for a to talk to Y or a to talk to X all right to talk to W should be the same.

B

But for one thing, the the the the name of a and the name of w inside, like the IP address inside those rules, but otherwise the rules should be the same. So we should not have 1000 set of rules flying in the network. We should have a single set of rules where a and Y and W Etc, also application and device, are.

B

um Empty things that you have to feel basically like a phone.

C

B

A

For me, that's an.

C

Optimization, it's just.

A

One just one just one point, because I I see that I'm running and actually I remember that in the past we discussed quite a lot. The question of.

B

A

Different problems that come from this mapping of like a plane up and down to open, Dev I, don't seem to recall what are what were all the issues with because they were like. There were issues. If you do that in a in a very naive way, let's say uh and but I think that it's not an unsurmountable problem right, so I think that we need to have a session specifically on that have like a small uh I mean here.

A

We have this, this General presentation and- and it's like I- think that we're we can we are, we can get lost, so maybe at the end in London at the ITF we can we can it and and just say, okay. Well, what was the issue with that? So what were.

B

The points we need the white gold right, we need the white, yes,.

A

B

Could be subject for society exactly to.

C

Okay, I agree can I have two seconds for the next two topic and I.

A

Will not explain just.

C

So we have a topic was to then, if you want to provision a car with rules, so it's to find a way to identify it and I. Think Pascal made a very nice comment last time, because here we don't have to identify the device, but we have to identify the rule and we have to take care of sort of privacy in in that part, and the last point is about security and I.

C

Think it's a funny thing, because if you look at the Young data model, in fact, you can give a readily or writerly to some leaves in uh in a model. So, for example, here I say that all the leaves concerning the field. Id are writable and we over I don't know. We can specify your sofa then, but in fact, when we will have a a Chic rule, it's not what we want.

C

For example, here I would say: I don't want a device to change the destination address on the core, for example, it could be a way to do a Dos attack because you take a control of plenty of devices and you ask the car to Target a specific host, and then you flood the network with that.

C

So what we have to say in our world is this part here is only or this part is a read write and the device can manage it, but it's not by leaves it's really by a type of fields that we we have to do it. So it means that we will have to extend what the control done by the young data model.

B

Because what you just said, if you understand well, is the the young data model, if you YOLO to change destination address, uh you will allow it for almost not just for this particular. Yes,.

C

Because it works in column, and here we have to walk in row.

C

This means that we have to, for example, to protect this specific line where the feed ID is 99 and 98., 59 and 50.. So.

A

That's actually that's uh I would I mean that's actually an excellent point. I'd say that maybe we need to ask to the young guys to the young guys to because I would be very surprised that we were the first ones uh having this kind of uh of problematic and to ask them hey. Have you seen that? So if they have seen it, you know, how did you solve it.

B

Yeah and it's not only to say I mean avocado, which say if it's really or not I mean you should have a way to enforce it. For instance, say we distribute those rules, so some fields are rewrite some Fields, so they are mutable. Basically, and some things are not mutable and the rule set should be signed so that the non-miccable fields can be validated because you will distribute the rule right. If you have this big mess so.

A

Now that you so okay, you have it by type, you can say, that's read only, but you cannot say it as long as the proof.

B

So, like you know, I'm in the mesh, so basically what I want is of a single rule set because I will flood the mesh with it. Otherwise, I need to download.

C

Just just one point: this is an optimization. It's a way to.

A

Simplify the fact.

C

That each device has is onwards.

D

C

As its own Rule- and we have some ways to make it the same on changes you use Roofing, but the.

B

Basic model is the dollar IP being replaced by your IP address, but the fact that in a given rule set some fields are not mutable. Do we have a way to prove it by crypto that when you get the rule set, they are the original.

C

B

C

Alex approach, because when you write a young data model, you say this leave is readable and readable or writable. You have a model that can allow you to control that, but maybe there is a way to say when we can allow real writing only if the value of this Leaf is this this and this, but I am not so.

A

They're, just one, there is one more point also here: it's identity, so authentication who has the right because I mean you can say that yeah this is writable, but maybe then you need to say who has the right to write? It I mean we can say well, it's everyone and know like it's either it's a constant or everyone can can write it. So, like that's a simplification, yeah, so.

C

Normally, what I, typically the security issue is that only the device can modify its work and cannot modify the other ones, but but even.

D

If a device can.

C

Modify everything in a rule, it can be a vector of attack, and so we we have to to take care of that.

B

So when you don't know the rule, my question remains right: when you don't know the rule, there are material fields and non-metal Fields now I will say which one I mean about which one I'm not, but it will not prove that it was not hacked. So if everybody obeys the rules, yes, you will only mutate the mutatable fields, but my question from the beginning is: how do I know that nobody has done? Got it, because it's not because it's written with right that it is retry I mean somebody sends the bits you want right. So.

A

Yes, actually you have two things. One is the values that are there and this is like you get from somewhere. Oh, are you sure that that's really the thing? So that's the first thing and then once you have it installed, then becomes alive. Let's say the the open Chic someone sends a message: hey. Well change that you have the. Then you have the access control. Then you have that you know well hey, but who are you you need to do the authentication part and okay?

A

Okay, then I can apply this access control so that I can I can actually control. Is it oh? No, we cannot. You cannot touch that sorry, the data mode, but so the specific profile says you should you.

B

A

You don't have the right to change that or you do have, but that's a separate thing. That's then the point that you're saying is: how do I trust the thing that I got yeah.

B

Just like, when you don't know the image from open cells, you have a shark.

C

But that the beauty of uh seabor and uh is that when you do Ash I, think it's it's similar, you don't have a white space or a notification that makes that if you do a lash, you have the same reason form, but it repeats your present in different ways. Here we have something that is uh easy to sign. After so if we have a way to define how we uh we say what is writable and what is readable, and we trust the guy that sign I, think we we solve the problem.

A

Okay, so that's another topic that sounds very interesting, so folks.

B

I mean the problem, the question the way I was asking it Laura is: do we sign only the non-mutable fields or do we signed every field? Mutable and not mutable I was because maybe something I think what makes sense.

C

What you can find here, you have a c bar structure and you signed with the broad structure.

B

That means that you also signed the mutable Fields, so as soon as they are mutated, you cannot validate them. No.

C

That's different yeah.

B

D

C

From a repository, and then you apply it, so what you have to identical authenticate is this is the rule I have to apply and inside the role. You have a now change this by an IPS.

A

C

That's another stuff.

B

That piece works.

C

A

To answer of your question, you can actually solve the part where you can have the mutable Fields. You replace them with zero.

A

Your signature and then you see that okay well.

B

That's that's what IP does I mean? Yes, for instance, when you do uh IP site, you know, and we have a rig with us, but basically the Elijah is you, you zero out the mutable fields and that's that's what you compute. For instance, the flow label is mutable, and even if the IP standard size or those drugs, they say uh you write this or write that or don't touch it. There is no way to delete it if it was not touched off at all, because it's never signed right, so they just we're.

A

At the end of the Hour Super interesting discussions and uh we're very promising for London I hope to see you all there, because it's going to be a good place to be I, guess yeah.

B

So we will, um we need to look at which rooms and I don't know when when they are made available, you know to book basically to side rule these side roads, but we need to book one and it depends on everybody's travels. So all the guys who are interested in that discussion, uh please let us know about your travel environments, because we we have this constraint to book that room pretty much near the meeting, because we still won't be there for the whole week. Yes, so we need to look at Thursday or Friday I.

B

Guess that's all those days! Yes, okay,.

A

Okay, so thank you very much everyone. Thank you. Thank you.

B

Very much bye-bye all right. Thank you.