Internet Engineering Task Force Interim Meetings, 19 Sep 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF-SCHC-20230919-1400

Description

SCHC interim meeting session
2023/09/19 1400

https://datatracker.ietf.org/group/schc/meetings/

A

Hello, do you hear me I have no sound on my side.

B

Yes, I can hear you, though,.

C

B

Like yesterday, you have a big low in volume.

A

B

Let's join she don't mind.

B

So um I guess the five minutes are passed and uh we have to to start the meeting uh by the way you see the old lp1, Chic, logo and I. Think Anna is working on providing us with a and you are a better logo. So uh and now, if, if you have your slides at some point, um it would be very nice that that you, you share um the possible drawings that we've discussed.

B

So we we vote on on what can be the new look for the working room, and so this um aesthetic center is not with us. So uh not well. This is uh F meetings, so all the best practices apply uh in particular, or the entire restaurant procedures and and patterns and copyrights procedures. So if you shall not aware of those procedures, please look at the best practices that are linked in this slide.

B

If you feel that, you're being disrespected by participants, you may contact people from the ombuds team, which is the link that's given to you and if you're aware of any IPR in the discussions, please let us know and I mean and disclosed an IPR. Please let the share, know or say it during the call or we find from giving any information about um or refund from using the kpi in our discussions. Basically, uh so the address for today is is very limited.

B

I asked for subjects that did not get any so the the goal is to continue the discussion we started at ihf117 and that people from the architecture document have been continuing, so it has to do with the way we do. She Getters the way we encapsulate Chic in UDP or natively over IPv6, and um what that means in terms of other chains, when you have multiple layers of shake one after the others and uh I hope you have some slides, I mean I've, not seen them imported.

A

No I don't applied for I. Have some in my computer, I I can show some discussion. We have but I.

B

Don't know if I can you, you still have the capability to upload them on on the site. So unless there is a you know, the data tracker meeting entering page so I can share the link. If you okay, all right, let me go in the chat and just share the link on the chat.

B

So this is what the session is and- and you can upload your slides from there and I will import them in a meeting.

B

Okay, um so so that's the topic, I mean. Is there something that somebody wants to discuss beyond that? That agenda.

B

I'm hearing nothing so let's move to Edgar yeah yeah.

D

Yes, I think I can take it in the other business, but yeah a little bit information about what is going on in a 3pp with reference to the energy devices.

B

You know what I got I would love to hear it uh before we start diving into the architecture effectively. So if nobody Minds right after the opening uh I will give you the ball to for you to to tell us the news. Okay,.

D

B

Sure and since it is now so please go on so what are the news.

D

Okay, so now it has been started to be discussed. uh What is the scope of the Zero Energy devices, so what devices will be um be considered for a standardization so that they will connect to the networks? They have been basically three type of devices being described from very, very simple ones: uh basically, almost a sticker type of device. So they are these box scattering uh all the way to devices which might even have a battery or some kind of high capacity capacitor, so that they might have more possibilities.

D

So at the moment there is still a discussion uh which of these devices should should be started, the standardization- it might be what my thinking. What will happen is that, most probably we will start with a more complex devices, so the ones that are still Gathering energy from different sources. It can be solar, wind or even RF, but RF is not so um efficient at the moment, but well it could be also, and then they have this capacitor, so they they are a bit more capable devices.

D

So then uh that is at the moment been studying, and we are also thinking that chick will have a very good fitting for reducing the energy consumption needed for each transmission. So then, we have a plan to submit actually the first version of a draft explaining this kind of, um let's say characteristics of the serenity devices and and what could be um the needs and requirements for the cellular network and what characteristic of chick address?

D

What problems and what is the gap so, um basically that that was what I had. So the idea is to maybe start with the first draft, zero four plaque to have already the first version.

D

Any comments or.

C

Mini yes, hello, hello, thank you for the information. Can you mention uh what release uh this is targeting? Please.

D

Be release 19 I think I have heard that this, it might be the one where it comes.

C

Okay, very good, and so you said three categories from the very simple back scattering device to the full featured the.

D

C

You mentioned with Scavenging. Is that the second category of that already, the third one uh which one sorry can you repeat you mentioned Scavenging reverse.

D

The Box Cuttery you mean.

C

Right that was the simplest one, and then you.

D

C

That simple yeah.

D

C

From solar and wind, so is that the second or third category um it.

D

Can be, let's call it type 1, so type 0 would be the one the box cutter type one is it can be that one, but it has Maybe a very small battery or a super capacitor and and then there is a type 2 that could have a rechargeable batteries, so it can be either one or two.

C

Okay, very good. Thank you. Welcome.

D

Yes, sir, here you have a question.

E

Hello, can you hear me.

D

E

Yes, well I I would just like to to mention that that I will be interesting in looking this draft and helping in any in any of the on anything. That is, that is required.

D

ah That could be great sure, welcome.

E

Excellent, just let me know and we can meet or or anything.

D

Sure, if you can pass me, your contact information I will I can contact you later on. For that.

E

Excellent. Thank you very much.

D

Okay, that's it! Thank you, Pascal, for the time.

B

B

Fine, so somebody is editing thanks so much because they did not create a usual minutes page this time. Okay, um so yes, the architecture, uh it will be paraphrasing a little bit what we discussed last time uh at the ATF on 170 17.

B

um about the the encapsulation that the sheikheders, if you remember she can work at every layer and you can even have Shake compressing some a pdu from some layer and then compressing uh later another layer of between different pair of notes, for instance, 112, an application, a pdu and pdu being compressed by Shake, which means there will be a descriptor for Sheikh explaining how which session? That is between those those two applications and points, and then maybe at the transport level.

B

And then maybe you have our link you, you will have Chic as well, meaning uh Chic will operate several times inside the same packet and what Chic produced for uh the inner peace will be uh binary blood for for the encapsulation Chic piece, and so we have to describe how that works and it starts. It could start at Layer Two and the chico, where PPP draft is like that.

B

um Basically you have. We have to describe in the the PVP adult station we're talking about and we met about this Chic session to the PPP session, and um then you you can do it a clear three, meaning that we need something like an uh a header, an IPv6 extension header, to provide the Chic information or it could happen even inside the application, and you realize that the formats may change, but the information that shake will need um is always the same.

B

You need to and we'll see that in more details, but you need to identify the endpoints that are compressing and decompressing. So you need you need that identification. Then if there can be more than one six session in parallel, you need a session ID and depending at which layer on your operate. You might want to protect the content that she has produced by some form of Sears, and so that's that's. What led us to uh the the definition of the Shakira that will I will uh present you now.

B

The Chic session is something that we've we've started to discuss in the architecture account, and um we realized that the twin points. It's it's always a point-to-point thing. To start with, so the thresholding points, it's always two notes.

B

So now we don't have a multicast shake, it could happen, but for now it's it's always point to point not only point to point, but we also have this concept of a role whereby one node is the device and one node is the network and that three narrates from things like Laura or sigfox, where the device you know with the low power thing and the network was um really a system entity connecting to many many devices and part of the things that the architecture will have to describe is how you set up the session in the 12 points in the usual low power device.

B

It's hard coded, like I, mean there's only one session and it starts when the device builds and- and you just keep on like that. But when you have more capable IPL points which can have many sessions with different align points, then you will have to bootstrap the session to provide enough information on both sides and that's part of what the architectures describe, which is information that both sides needs to have. What is in that information? What is re-synchronized must be the same on both ions and what?

B

What is more, contacts that have to leave on both sides, like counting the frames or whatever um and and part of that, will be effectively what are the rules and where the data that are needed to bootstrap this session. So what kind of network management do we put in place?

B

um Is there some Network management that pushes all the rules to the endpoints or network management provides a URL for the rules and then the endpoint goes and fetch the urine? So all those things have to be discussed now, once we have set up the the session and the Managed IT Etc and, like I, said it's one, big thing that we have to discuss, then we have to look at the um the data stream. I mean how does Sheikh fly over the different layers and, in particular for this picture?

B

How does it fly over IPv6 like if you have an IPv6 header, um and you want to comprise UDP on above, for instance? How does that work? How do does endpoint a know that it stop the packet that it receives from endpoint? Two?

B

um Is for this particular session, and so we realized that at least abstractly we need. We need a container for that information for all the information that Chic needs, including decision ID. We need a form of container and that's exactly what IPv6 extension leaders are. So we started defining hey based on the data. We just discussed that we need to identify the session and stuff what would be in this header and for now we realize that it's well I formatted, like a user, use a classical usual IPv6 extensionator.

B

So you have an extender and a length, but then we need a checksum just like UDP would and we need a session ID and for now that's all we have, but we keep the capability to extend it with options. Now we need this header, it doesn't mean that it will fly in the packets and we'll see how that goes, but we needed at least abstractly to conceptualize what happens in the stack before you compress everything.

B

And so, um if we look at at the packet which is normally issued, like you know, libraries Etc, let's say if you don't have any extension leader, so the next setup field in the IPS Etc says protocol is yourp. Then you have a ulp, either think VDP TCP block and then after the ulp editor you've got the portable data unit. That goes with that ulp. Now, if we want to compress above IPv6, we basically need to Signal the compression session and this all decision information inside the packet to describe what goes after.

B

So that means that at least virtually somewhere in the stack you have to insert that Chic header that we just showed so that the bucket looks like the second line. I mean the next step. They are now in the lpd6 header, says, check and in the shake header, then you say you'll like standardize your p, and now you have a proper uh IPv6 chain.

B

Yes, we probably will have a hard time getting six men to give us a number for that check header if we say the other flies in the air, but actually it does not we'll see that as soon as we've done this. The next step is to effectively compress based on that.

B

So the first step in that compression would be to effectively compress the European, the URL pptu and say you don't have nested Shake here, just for the sake of simple simplification and then we'll talk later about this day. But basically, what seek will do based on the set of rules is comprise the URL together at ulp.

B

So all this now becomes a rule ID and say it's not fragmented, but it could be so rule ID explaining you know which role is used to compress. You know the black and the purple things and B for that you will have the Chic header indicating which session uh between the 20 points, the 20 point being the societies National IP in the ipadder um so which, which session contains the rule ID.

B

That is used to compress the green thing so when and that could fly uh in the air, but like I said IPv6, people will tell you that. No, no don't do it in your editor.

B

um The firewalls will not like it Etc. They would not like it, but but the next step is probably to compress that anyway, and so, when you do this second compression to get a native Green Sheet pdu, you you could, if you don't do any nesting, say I've comprised everything in one shot, including the sheikheda.

B

But that means that the thing that compressed the chicken or the rules that compress the chicken are lost are part of the description from the compression of the ulp and that's not architecturally. Correct I mean there should be that this is layers. You have a Chic layer and you have a urp. You should really have rules which describe how you compress the right thing and then rules which describe how you compress the um the black and purple things.

B

um At least one remark here is, since we have placed a checksum in the Chic header which protects the shakeda as part of the rest. This checksum also protects the black and purple things in the compressed form, meaning that you don't really need to transport the UDP checksum that you can save those two bytes, because you know that the the urp black and purple things are protected by the checksum of the red thing. So that's why you see in this picture?

B

If we have to do that would probably be light to check some in the udpps, because we have the checksum in the Chic piece. So the black checks are would be aligned because we have the right chipset, the red piece we would have to find in line at least the checksum in the session. Because if you have multiple sessions, then you need to identify which one but like I said it's a lot: cleaner, architecturally speaking to split the two Chic sessions, the Chic session that compresses the urp and the session ID.

B

For that would be indicated in the red thing and then the session that compressed the right thing. Then you realize that we have nowhere in the blue header to indicate which session chick is, which tells you that it has to be implicit. It has to be well known, so what we are getting at is if we compress the chic header, so the red thing, then the rules to compress it to decompress it I'm sorry at least all the way to get to the sixth session.

B

That piece of the room must be well known.

B

It has to be upcoded, because there is nothing before to Signal what it is once you've extracted that then, maybe in the Chicago you could have additional information to say: hey I've got chick options and yeah. There is a session to decode my shake options, but at least to decode what we have shown earlier. Let me go back.

B

The Chic header, at least to decode all the way to the session ID, and this must be upgraded. It must be always the same because you need to decode the ship session now, if let me show you again the chicator. If we decide that we have options, then the session ID to decode the options should also be light.

B

So so we have to figure out how that works. If we have to, if we want to have options and if it's not just the first six bytes.

B

Which is interesting is this is an eight bite thing? That's the the first two lines: they they have to be eight bytes, because IPv6 aligns the headers which tells us that we, we must have embedding role in there and if the header relax is in line because it's not the default uh one, actually zero, I guess. But if it's not the default, meaning that there are options, then the other lines must be in line. And so we have to be able to pad all the way to the eight byte alignment.

B

If, after the options, which also tells us that probably we might look at at the rule that we don't yet have- which would be a padding or I, don't think we have a padding rule. But remember when we expand this this header, we must be aligned to it, so just a side concept, but okay back to to this description here.

B

What we said so far is ideally, we have two succession, one for the pdu2, which is the urp and one from the sheet video one which is Chic itself, and we said that the code should hard code enough to be able to at least know where the the pdu to starts and um what session that is for the pdu2. And then, if the checksum in black is lighted, then other checks are in red. That covers the red to Black and the top. But that's what we said so.

B

um Last but not least, if we leave a local domain, where this Chic format would be accepted by the firewalls and want to go across the internet, then we we probably need to encapsulate Chic into UDP. So even if the the apparently your pdu in black is like TCP just we still need to encapsulate the whole thing in UDP to pass the files. So if we have to do that, the purpose on the table is to say: okay, we got a UDP editor that goes before the Chic header.

B

Destination Port would be, software would be the sixth session, and so this this serves as um the indication of both the session and the port, meaning that we don't really need uh the the red header at all. It's it's mostly compressed because the two information that it contain, where the Chic session and the checks up. So if there is no option again and we don't have any game of option for which there should be a session for Video One, um then we're all set.

B

Basically, the pu1 could be only the information of the Europe, and if, if in this game, there is a single ulp, then the first coin block, so the red block basically could be comprised to zero. Ulp is well known. Session is in the is the source Port of UDP and checksum is the one of the encapsulated so so shikado now goes all the way to there.

B

If you have more than one urp, then you still need to have some residue for the first green thing uh saying what are your opinions and that's pretty much all the discussion we had at the time of uh ietf 117. Just um we also discussed the the Chic context like what both nodes both endpoints have to to know and part of that context is what the curve c8724 really called context, but not just a set of walls.

B

Then, if we do some management that could be amendments and and Laura started to discuss uh how to change an original set, then there is the rule. Data like, for instance, is there is a variable in the expression of the rule, which is your IP address. We must have a set of informations binary value of the IP address before you. You really instantiate the rules So based on the original one said the Amendments and the data that you apply on it.

B

You really get the instantiating rules that that operate between the Imports, and so, for instance, you have your IP address and the IP address of the other endpoints, which are popular all data, and when you start the session, you you you apply them to do the rule set.

B

That gives you the instantiated rules, which is loaded in memory, but from there the rule starts operating, you start sending packet, so you must have, for instance, fragmentation, timers uh and some other runtime state, which tells you what you served and why you are in your exchange. There is a lot of State in the fragmentation piece, so all these, in my view, constitutes the context that chick needs to operate in an endpoint, and then this context can be managed by some Network management that Lahore started to discuss as well.

B

We need to to work on in this check, work and well, that's that's pretty much it. We started the discussion about how we have the rules, uh how we write the rules. You realize that some rules maybe share the same beginning of the packet so that they have the same beginning and then Fox if the continuation of the packets can be different and then again again, which tells you that the rules are really organized as a trade.

B

So it might be something interesting to know and represent when we write data model folds like I'll, say this is the beginning of all those roles and then the system acts, and then this is the next and this therefore, that's just another side consideration.

B

Okay, with this I've spoken enough, um do we have questions from what makes sense? What does not make sense? It's supposed to be uh just a reminder: ATF 117, as the base for the discussion, and then we will have everything that Lahore wants to tell us today.

B

So questions synchronization.

B

A

Foreign I lost you.

B

I was muted, I was asking for questions and then.

A

B

It's not me, what do you mean? That's not me. I'm.

A

Here I know: I am delegated. Can we mute you but I didn't.

B

Know I was saying: I, muted myself, because okay.

B

So any remarks: what did you want to take over from there I mean? Did you upload your slides by the way.

A

A

B

A

Otherwise they would share my screen if.

B

You know it's there, I uploaded the the slides, so you might share them. I see that another question go ahead and.

F

So the question is the Chic rules that we are defining for the next header on the session. Are they Global or not,.

B

What do you mean Global? They? They should be well known, meaning every note without the saying, but they would operate only above IP right. They would comprise the IPv6 extension header, not so many of two things, and not some so many of three things, but suddenly have four things, because you well that's what we discussed so far. Now, if we want to start uh I, think something which is the generic rules for Shake in general at any layer.

B

We could also do that, but it could be more complex because, for instance, you could have to describe the ulps. If you are, um if you don't, if you don't want to inline the urp value, you would have to to to start listing them, but then that list would only work at the layer three and that layer four in an application layer you would have no use of it. So the way I was seeing it is the rules would be.

B

That would be a layer, 3 rule for the shake header that layer, 3 rule would be well known. But for anything beyond that like using shake for cozy I, don't know, then you would have to have another set of forms, meaning it's a different session and anyway the social ID applies, but a different layer. So so it doesn't identifies a different set of walls. Etc.

F

Yeah but that's the set of rules of compression, but the Chicago rules will be known for everybody, so you you are using the same format. The next header and decision ID in all the on all the layers. So the chicken is.

B

The same in the in the other layer you would basically have to because you have a chain of chic data and Chic and data, so you would always have to be able to say if there is another Shake inside you, then here is the Chic session and the chicken points. So maybe that's something that we would need. Let me work so yeah yeah. So so, if this is just, if you don't Nest any Shake, but if you have a nested shake at.

C

B

Somebody has to signal what 10 points are and what the session ID probably checksum, is not needed, but the session ideas and the end points out um functions which application talks to which application.

B

So for that you will need a sheikhedder and you will have to indicate what session that is. So so you can decrypt that you can decompress it and then you have disappeared, but I don't think it's the same chicken or it does not contain exactly the same.

B

In the case of IPv6, the endpoints are really taken from the apd6 header, so they are implicit in the shake header. But if you wanted to write a very abstract shiked out, which has everything then depending on where you are, it contains a lot of stuff. So I don't believe that there's a single sick either at every layer. It says it's a different one which depends on what you're doing.

B

This is the IPv6 chicken.

F

B

So, to find you know your rules Etc, we need to indeed identify.

B

uh uh At the frontier at the end of the the bits which represent uh some layer, if you compress a linear layer, you have to have another secreter and you have to be able to to find the endpoints and the session for that compression and it has to be in line there in, and there must be some some readable information that tells you that I don't know exactly how this is done. Right I mean the solution. We just that's. What law is going to present to us? It.

B

It's just that for me: it's it's, not the rules that that we'll use for V6 headers, it's it's mostly the same abstract, Conte, content right and the abstract content is what we have shared such an ID and points which roll the endpoint, the receiver and the center are. But but um the the rules and the compression will will be different.

B

That could be option. There could be a lot of stuff which depend on the layer.

B

That's just the beginning of the discussion right we are nowhere. I mean we. We have to figure out that we don't want to do that.

B

That you could present yes.

A

I think if I can share the slide, yes yeah.

B

A

uh I think is this one? Yes, okay, so this is really a primary work, so this slider from this morning- and there were more used for for description about the architecture than uh so. There is a lot of open question and maybe a lot of mistake in it. So it's good to to have this this description.

A

So the idea is to to look at some complex architecture. So we took the one from Oscar draft. It means that we have three layer of cheeks, so we have the Oscar layer, which means that here we compress Oscar packet and then we encrypt them. Then we have the co-op layer where we put Co-op and then we we compress it and we have the UDP AP layer.

A

That is only useful on the lp1 network and when we leave the lp1 network, we have to reconstruct an IP UDP to go, or at least a IP, but ipudp can be useful too to to cross the internet to reach the application running, also Co-op and Oscar. So that's the idea of this this architecture.

A

um So what you you have also in in dotted line is where we can have the chicken. So the good news from Chicago is that we were using it from the beginning of chic, but we didn't. We never notice it. It means that if we don't Define an issue header, so we introduce nothing on the packet. If we have a sheikhedder, that is all the fields are well known. We have only one rule, so one rule, we don't need rule ID and if everything is not sent, then we have to send nothing.

A

So we have something that is also compatible at that place with the existing Shake, but we can introduce something where we can put some information that will be useful to to manage a network. So on this nice drawing I I tried to explain what we can have if we have shikider at all between all the the layer. So from so the first one will be at the lp1 level, then we have the co-op and then the real score.

A

So here we start at the lp1 networks. We can have a sheikhader and in some cases as Pascal said, the shikider will contain a session identifier, which unfortunately made also seed, which can be confusing with what we do in uh in kokom, but has nothing to do with that. It's just, for example, a session ID and we carry it on V session. Id will be able to identify different Chic instance or session running on on the same device, for example, or on several devices.

A

So with the shikader and the seed ID, we go to a set of rules that are the one for ipudp compression.

A

So inside it we have a rule ID that defines the compress the rule we need for to compress a UDP and IP, and then we will have a second layer of chicader that will be introduced. So we have a new ID 6 Prime ID that will be used and it will point to a set of rule that will be used for compression and fragmentation of Co-op.

A

Then we have a rule ID that is used to to do this and inside we have another uh chicader that will you. We may have, of course, another seed. Id that will define an instance for for this, and finally, we arrive to the compression of Oscar and not Co-op, as shown on on the slide. So we have this kind of onion uh encapsulation with older year of chic and we introduce shikeder that is not mandatory. That can be uh totally transparent, but for the exercise we we try to put it everywhere.

A

Okay, so in the next slide, we we try to see the organization of that. So, of course, is maybe Edgar.

D

Yeah sorry I didn't get the part of how do you know that you have another uh chick header inside the?

D

um So how? How is this signal? It's.

A

Part of the context, so we have to find a way to say in the context we have a chicken.

D

uh So in the rule you specify that you will have another yeah another header and is then you can actually decode it from the rule uh that that maps to that.

A

Yes, exactly so, we don't know yet because it's not easy to uh to see in the best way, so I have to look and open cheek and see how to define these rules. So during the ATF we made some attempt on where we had uh compression rule fragmentation, Rule and chicken rules, but it doesn't work well because normally the rule ID for shikeder are independent from the rule ID for compression decompression.

A

So it should so the way we represent. It is not completely clear right now, because, for example, here you have seed one that point uh to a device, and here you have a center number of compression rule. Then you have seed I, for example, in the middle but point to another set of rule, but you have an in common your chicader throughout. So the way to represent it easy way in an easy way is not defining right. Now.

D

And they Supply also to fragments, as I can't see there. So it means that you might have rules for fragments and rules for you for the full packet as well.

A

Yes, that's a good point, uh it's still not clear right now, but what I? For example, if you have a session ID, then you need to carry it all the time to tell uh to which a session ID or session you. You have to send the information, so we Supply two fragments- and this applies also to a compressed packet now, when you have next up so next Up Normally will only uh be used for compression and not to fragmentation, but maybe it's better to carry it in all the packets on all the cities.

D

A

E

Hi Lauren, so it would be like another CDA, but without a field, ID I mean.

A

So fill ID, all these things are in the compression fragmentation. So what is in green, uh blue and purple is not changed.

E

Yes, but you append, like in the blue screen inside the wall like just saying that there is another there next, something like that: yeah.

A

So maybe it's more clear in this, for example, in I, don't know if I can point out something using the Tool, uh but for example, if you lose look at the last last line, you have IPv6 UDP. uh So no, maybe not the last line, but uh let's say the line just before the vertical Arrow. So you have a chicken before the the rule ID in green and the residue of the root ID. Then you have a chicader pointing out to uh uh I.

A

Think it's Co-op here, it's a mistake uh pointing to co-op here then you have the residue for Co-Op. Then you may have another chicader and then you have the Oscar compression.

A

And what I think is important is that all the space where you are numbering your rule, ID so for ckder for and blue, green and pop uh Bluegreen and purple uh are totally independent.

A

So Pascal says that we have something hierarchical, I'm, not sure to understand what it meant.

B

That you have just rules which, which can compress IP UDP, grab everything the whole packet with optimistic, and you have different flavors of the iPad like like possibly for this initially to say that so you have. You have four rules which which describe that piece um for values well, one rule which, basically, in the name of one four or something, but maybe you have other variations and you have a few rules, but then for each of those few rules you might have different compression of the UDP and then different compression of what comes next.

B

So what I was getting at is you could represent that as as many lines where you you, you show, this rule says ipt station is this then UDP perk is that then the second rule was the same iPad as the first, but the UDP area is different and and uh it has another port and then whatever- and you realize that um you could have said all those rules of the same beginning like they always they all say this initial AP is this, but they stopped variating at the UDP, so so they could be.

B

You could represent that as a tree where.

B

They solve them.

A

So I think this is an optimization of she can not something related to shikider.

B

No, it's not related to Chicago. I was talking about the the rule structure really, um and that's so it's the last slide in the presentation that just made it was a hint that it's because we discussed it's the hf-17, but that, obviously, if we do things in layers, we have a lot less of what I just said because effectively it's really shake within Shake within Shake which cause that that tree that they had in mind. But otherwise you would.

B

Basically it was more about the model of how we program and express the rules if we express them as a tree rather than let's say um it's, not even the tree is probably a graph by the way, um if you represent it as a graph, rather than as a number of lines which repeats the other lines for large piece.

B

Because you can have different designation, IP and and for each of them you might have to send you DB Port, they don't know so so it's like you say it's an optimization, but not an optimization of the rules themselves, but on their representation for the measurement.

A

Okay, uh so manage.

G

uh Hello Professor uh just wanted to ask on your slide number two.

A

G

uh There is uh yeah, so the middle block said prime one and said Prime two: are these Delta encoded sets.

A

No, it's absolutely no seeds as a corkov he's just such an ID and unfortunately, the same name.

G

ah Okay, okay,.

A

Okay, nothing to to do with.

G

Okay got it. Thank you.

A

So uh here it's straight to work on on that, because it looks quite complex and I think that when we will and delete a little bit better, it will be. It will not be that complex and so here it is, it shows or try to show the different steps we have when we do compression on the Node, so at the beginning, so forget about UDP ipv62dp Co-op. At the beginning, we just have Oscar a message, and so what we do in the second line is that we do the compression.

A

So it means that we have a rule ID saying that how to uh to compress or decompressor score. We have the residues that are related to that rule ID, and then we have the payload so the next step. Normally what we do is to to encrypt the information. That's why now it apply in dark in in the slide, and next step is to go to co-op. So we add Co-op and uh uh just before sorry we have to put the Chic header regarding uh giving information.

A

Maybe a checksum I, don't know about how we carry this uh this already, then we do go to co-op. So to co-op we do also a compression and what is important to notice on this by what pascara says is that at the beginning, we had the.

A

Udp power should be something that point to co-op and after the compression we need a UDP Port that Port that point to check and etc, etc. So, after that, we we have to send it. So we add also the chicader regarding to this layer, and then we do the compression of ipudp the same way.

A

We are the Chicago and the vertical Arrow shows that we are saying sending it on the lp1 and it arrives at the other side and at the other side we decompress it and we reconstruct the UDP ipv UDP carrying the Chic, the props or the next leader, the pathway, the port will be sick and then we send it on the internet.

A

So it means that we are in the green box in the middle, and here we create the cicader that exists or not, because it can be fully compressed on the context on both sides or uh carry some bits. The sending this this information.

A

So that's some.

A

Details are on that so, as I say it's not totally clear. We have to see our we represent it, and the last point that Pascal talked about is how to put the management in that. So in fact, the management. Normally we use netconf cocoa for uh rest conf, so we will have to uh to Define some rules that will do that.

A

So if we go to a score, maybe we can have a a special IP address that tells that it's for management on the device, so one possibility, for example, could be to uh to use some anycast address or, and even I, don't know if it's possible, some link local anycast addresses to say this is but the two elements for the for the management and this way we may have some specific rules that under this, uh but if we are not dealing with IP on the compression, how we can do it, that's still an open question and the other question is: can we manage the chicader rules and for me, maybe it's too risky to to do that, especially the first one, because we have no way to really check with sending the information.

A

So it's things also, we we have to to put in the architecture, but my last word Pascal is that you see that even if we have only one device, we may have different chicken stents and it means that we have to identify it internally and external.

B

For you, you will talking about um the the way we identify the management and for one thing that could be a management session. So if we can identify multiple session, if you have a session a Nike packet coming in your device from an IP address blah to you and your decision, ID I, don't know Fox sucks, um then it's the well-known session ID for management. So it's expected to be a management session may not know.

A

Yeah that's the possibility. For example, we have session id1 for regular things and session id2 for management, but in that case it means that we lose one bit on each packet. So.

B

Because we discussed this one bit that we lose because we can't comprise the session anymore.

B

um There was also the the basically the point that they have in mind is: if you use in any case something you have to place, the MAC address uh to the right guy, and so that only works on one hub. But if you.

B

If you want to manage a device over IP from a distance, you probably need uh UDP as well. Remember the the conversation and, if you do, then the alternates share your DP address with your your punishments. So it's a different format. That's when you do it from remote to iudb, but if you do it from remote over UDP, when we ask for a UDP Port, which says check, we could also have a second UDP Port, which does Shake management. I, don't know it's just something at the table as well.

A

B

B

And by the way you lose the beat on, if you do management right, because the default you could have a rule which does not have oh, no, if all the rules must, you must always have a right if you have more than one so you would always pay for the rule ID, even if everything else is implicit as soon as you have more than one.

A

Yes, if you have only one rule, you don't pay for the rule ID. If you have several.

B

Rights but but here, if you have a rule which does not have a session and a rule that has a session in it, then it's two rules. You have to pay for the rule ID, so maybe it is and pay for the session.

A

Yes, we or we can have a rule but say we have a in your uh in the opportunity part. You can say this is a management message, uh and so this way we because that's a good question you ask: do we allow everyone to manage a device, a Remote device, or do we just allow the other entities to manage the device.

B

It's always a packet that you receive anyway. The bottom line is probably, you will add, near low, something which is cozy encrypted, and so it is basically sent by somebody. You trust, because you have this public key, so you only accept management from a public key that you could decrypt from a public with a public key that tune you, you accept.

B

And so another thought comes into play yeah, so we have reached the end of the hour. I mean that that was more. uh You know, that's the the beginning of the season, and that's also explaining the kind of problems that we'll be looking at during this season. So thanks so much Lohan for spending all this.

B

um For that we we're just exploring right we're just laying out in front of you guys what kind of problems we will be uh considering for this season, but we don't have the answers we we are looking at.

B

You know how we want to structure this and express it what's very clear for us is we need this ipv Etc, um even if it doesn't really fly in the air it could, but only in a very small domain, then you will need to to ask the IPv6 six months for that header value and we know it's going to be a fight. At the same time, we also about just sent me an email saying it could not.

B

It could not be with us today, but we also um considering asking for an ether type and and the next letter to say. Well, it was not initially an extend or it was a protocol type, but what what we just discussed now, it's more than the protocol type, it's real next header, meaning effectively there can be a an IPv6 header.

B

It's not just it's not considered as a European. It's an extender, the initially in the discussion. People were not clear on that, but I guess now. We are getting very, very clear that it's it's an extension leader because after that you have any Erp that you compress that's pretty much it for me. So do we have any question or famous last words before we are drawn.

B

Okay, I see that we're all set, so uh thank you all for for the participation and thanks for the minute takers, I guess that was enough for the most peace I do in three look at who was taking them, but yes, big thanks to the minute takers and I will upload all this too. To this I know.

B

You guys take care bye-bye and Dominic. You wanted to say something: yeah, okay, bye-bye.