Internet Engineering Task Force 99, 18 Jul 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF99-6LO-20170718-1550

Description

6LO meeting session at IETF99
2017/07/18 1550

https://datatracker.ietf.org/meeting/99/proceedings/

A

Welcome everybody, hello, welcome everybody.

A

So which probably closed the doors and start it's already 16 v 3 53 52.

A

First of all, this is six load, just in case you're in the wrong place. Make sure this is what you want to be. We have a minute taker, Thank, You, Dominique, right there, yeah and I believe we have a jabber scribe as well. Thomas. Thank you, so I think we're all set, make sure you see the knot well, this is just to see if you're awake. Actually, you may have noticed. This is not the very latest and greatest, not well that doesn't matter.

A

You have to be aware of the very latest ingredients, not well. It was recently updated. I didn't actually manage to do it on time, but doesn't matter you have to be aware of the very latest and you can always get that on the printer agendas and on the ITF, webpage, etc. So make sure you understand this, something you don't agree with it then you should leave.

A

This is our agenda, for today it looks a little packed, but I think we should have at least in theory a little bit of time left at the end, which is good with because we have sort of a discussion plan for the very end. So if we have any extra time, I'm sure we'll be able to use it, but basically we're gonna have an update to 6lowpan & D, which is actually a collection of about three documents with Pascal and then younggren will use an update on the applicability and use cases for six.

A

Lo, we have Charlie on packet, expiration, transmission of ipv6, packets or PLC, and then we have a new proposal for body area networks, 15.6, I, Sajjad and then at the end we have Pascal again and with with the chairs, we can moderate session in total. There will be like 40 minutes, so Pascal start with some requirements, and then we can have more like an open discussion, hopefully with some folks from transport as well summarizing status. Since Chicago we have three new RFC's so that you le now became our CA t-105.

A

The Lobeck document now is 8163 and not that it was a working group document officially, but we definitely benefit from it. This is draft queuing given n on a to the 15. Ie became 80 137 and, as we had announced in Chicago, we actually went forward and carried through and dropped. The mesh link, establishment effort that we were undertaking, partly or quite in large part due to interest through from Jupiter mesh and Zig mean, and since they no longer needed it, and we didn't get the required update, then we dropped it and then we do.

A

Everybody was fine with that status of document. Since Chicago we have well back on router you'll, see an update on that. That's the long going and if see, we will see a an update by Singapore, so we don't have one right now, but by Singapore, incorporating more comments, etc. A P and B is also part of the quote: unquote 6075 update, which Pascal will tell us about a little bit more.

A

The other interesting one was Bluetooth le mesh and we could have had an update this time around, but the authors are very wisely. They say decided that it probably made sense to wait for implementation experience. That's always a very good reason to wait for a little bit more so implementation experience. We should have maybe something by Singapore, hopefully or and the applicability and use case we'll see an update now and, of course, I mean 6775 update itself. The main document there is what's updated with working with comments.

A

After that we actually issued a working, a blessed call which finish about a week ago, and there were no comments during that period. But I understand some people are still thinking of sending comments in so, but it looks, looks reasonable and that's that's pretty much it for status. So unless anybody has any agenda bashing, we can go forward. Pascal.

B

A

I'm not gonna, insert anything here. I.

A

Just saved next to no.

A

Yeah keep the laser okay.

C

So this is pascal chiba from cisco, so I will talk to you about our suite of drafts, which basically stands from the RSC 6775 half date and yen goes all the way to the backbone router and the address protection drafts next slide. Please. So we have extracted the RFC 67 75 draft from the backbone wattle draft. They were both in the same place initially, and so we have extracted what truly updates the RFC versus what is the new proxy operation.

C

So basically the RFC 6725 update can be seen as early as free association process where, as the backbone router is like a proxy operation, what does the router do to act as a layer? 3 proxy for neighbor discovery?

C

Next slide, please so status of the draft a refresher was published, late, June, so there's a number of cleanup activities which took place. There was this big review that Samhita published on the mailing list, and we acted on that.

C

We removed the text, which explained how you can avoid sending both ripple, dowels and Dadich, and the this being addition being made on the basis that this should be a separate document which really says focuses on savings. You can make when you get both on the same network, but this document is exclusively about ng, so the reference is to repower are gone, will have to write them somewhere else.

C

We might have been too efficient in doing this because well, ok well, point is: there is one important thing which is the operation of the sequence counter, and now it's rap, so it's initialized how it's reinitialized, etc. When you reboot, and all these text was actually a reference to the past sequence in RFC 6550, you do the same as this as we remove the references to 6550. Now we don't have any more operation of this sequence counter.

C

So we are missing that text and we have to find to decide how we reinsert it either I say reference to 65 50 or we copy the text. So our RFC self-contained. But it's it's a clear goal that it obeys the exact same rules, because you need to be able to use one or the other, but not having sent two messages in a ripoff already.

C

We we have added two versions ago. The recommendation cos tend to use 7400 to expose the capability to do this new draft.

C

Now we can live without it, but it's better and clearer to announce that through new bits in that draft that this capability spoiled there was also a need for the address protection draft to provide an additional status code, which means that the which basically returns that the proof that was given for the security token is not accepted, and so this data is provided sure, but the action and the status is in the a P and D Draft, and we had we separated the privacy discussion from the security consideration.

C

So we have to now those two separate it. The security consideration is the usual security saying and all the text which says oh by the way, we don't tell you how you build an address, we place no constraints and what you do. But if you do it, please make sure you further the privacy recommendations. This is now its own section.

C

By the way, there is at least one implementation, I've seen I, don't know if the second one is complete.

C

The document was ready for Lesko. It was actually last call I thought I'd fix that slide, so I'm confused, maybe I wrote it on the different slide for.

D

Each last call.

C

We do we finished last call. There are not so many con, some of them I received on directly so I asked the author to publish them on the mailing list. It was not done so. Please consider that is not completely complete and sure this commands are published and there we go for this side slide. Please I know it's not the right, slides.

C

Yes, this one is correct. That's a second draft, but I was supposed to be presenting 61st.

C

C

That's me mom when Senate asked me to send them separately, so the old deck with everything in it is difficult.

A

Yeah I think I got your three. Yes.

B

B

A

Is one of the reading? This is the second one of three, and this is the third a no two three with this.

C

E

C

Like and score it from the website, there's only one slide so.

A

What these are the three year.

A

These are the three years on, and the attention back, nor others. These are the new ones.

A

That's the one it's opening act. It doesn't do anything because it's already open and.

F

G

Hollering I wanna.

F

H

G

I can take the next person on I didn't know a little eager. That's all you wonder that.

B

G

Yep shut up reason: the slides to give whatever.

A

What do you want make sure they're there.

C

A

Yeah just burn up your work. Oh not this one! This do.

F

You have one yeah, okay,.

C

G

How it's been like.

I

So for the people remote, including.

D

I

Like Pascal is just switching slides.

D

C

Some little background about the whole work. Yes, we are at the crossroad of wireless around ipv6 and on both sides. There were expectations, requirements that was round about the fence on what basically a start group expected from the other.

C

For instance, ipv6 has design service did multicast with these requirements that there would be a twenty power of 24 multicast groups to make it perfectly efficient, but on the other hand, those groups were never the capability to twelve. This number of multicast group was never present on their Tripoli side, meaning that it has always been broadcasted as opposed to multicast it that never prevented and it work, but that that kind of creates limit to the applicability and efendi to experimental e 10,000.

C

The order of 10,000 nodes right, which this limitation rate comes from the fact that multicast is broadcasted now on, whereas it's stringently different now in a general manner, net India requires a very reliable multicast or broadcast operation.

C

Even if it's a broadcast, it needs to be very reliable, because that's how we discover address duplication and as it goes, we we all know that, due to some laws of physics, certainly the wireless will never provide that cheap, reliable multicast is not something that radio can provide so suddenly, 49 48 62 cannot operate as decide on ratios, which boils down to the fact that if it appears to work it's mostly because the entropy of ipv6 addresses makes it so that the collision is very rare, not because ipv6 operates as designed.

C

Okay, so this is basically the expectation from the IDF and what yet Tripoli should have or should be doing, but that was never present on the Tripoli side now the same effect exists on the other side as well.

C

There are those expectations that the actually has an IDF which were never really there, in particular in that era, when you read that Wi-Fi recommends that a proxy operation is performed on the IP 4 and the up as it goes, we have never specified this operation, and so we have resorted to proprietary implementations to decide how exactly a proxy would operate in that particular context and the one thing which which certainly does not exist, that cannot be done without a standard is the equivalent of the Association, because what those implementations do is that they snoop nd and DHCP.

C

Eventually, they resolve the traffic, if nothing else works and based on that, they discover that the certain address is present on a certain IP and that's the way all these proxy operation is kicked off.

C

Basically, we now use RFC 6775 update as the layer 3 Association process, where has the backbone router is the proxy and the operation process. So we are finally can bridging the gap between a Triple, E and IETF and providing the missing link, which really enables a proper operation of ntp over radios exactly so. This I already presented next.

C

Thank this already said.

C

Ok and then this one has the correct dates in it. So like like I was saying we have address, we are there, go back yeah so now this is what I was trying to say on the previous slide, where, which is? This is the pointer to the last long review that we had on this document, and so we had the disparity discussion. We were a bit too efficient in removing the some text about the tid. We have removed the text that enables the DAO story for title to refresh the DAR state.

C

And we passed working with, let's go okay, it makes like these so like I said: I get, some private commands will address them, but as the auto computed earliest otherwise, and for like it just said, we probably need to point again on 6550 or to copy the text which described how the sequence counter is initiated, know the lollipop thing and all that sort of thing and dude question for the chairs. If do we have enough review to call it victory or do we need more, oh well, good time.

A

Daddy one thing you know on the: how to restore the information that was taken out. I would much rather I would vote for having a self-contained document, because if, if you have a pointer there for something that would end up being normative ripple and you are implementing something as don't do with ripple, because possibly this could be a more general facility, you don't want ripple to be normative great.

A

So anyone the document is to be self-contained, so they'll be mine.

C

So we lost the author of that text in repo if he can copy give me the text. Yes, okay, that's.

A

Dark I'll: do it.

J

Righto from Robbie, so the update allows the proxy to register to the six proxy base. Registration. 266 bbr is what I understand right so extended ero your option allows epoxy based registration, a proper.

C

Registration to the proxy on the.

J

Arm of the target, okay.

C

There were proxy appears twice first years. We read it for instance, why five devices could register to the AP acting as early as water for the to get the proxy and the operation so that the first occurrence of the word proxy? There is second occurrence that there were proxy I. Think you're talking about the second one, whereby the root of the repo dyrdek can proxy the registration to the proxy.

K

C

Behalf of all the nodes which it learns, repo and maintains for repo. So that's why it's so important that this t, ID operation is aligned to the operation in repo. Basically, in normal run time, you won't see the dark periodically. You will just see the tail periodically, but the t ID inside the DAO that the root receives will allow the root to maintain the prompt. The state on behalf of the registrar cannot know the target Norgren. Yes, so that's why he.

C

One of the big changes that we have in this document versus 6725 is that we registered the target as opposed to the source, because the root of the report do dag is the source of the packet, but what's being registered, is the target? Actually it's in my mind, I mean at least mine, it's kind of quite consistent with the usual use of a target, so I'm pretty happy with that judge. So.

J

You know the next question, for me from me, would be: is it even possible from 6bb are to be register on behalf of the target to what Budi register like we have the box.

C

Of registration.

J

C

Yeah sure, if you get a no path now, no no I, don't think the spec allows you to do add to do that. I'm sorry, but I have some text on that I. Just don't.

J

Remember what I wrote the reason why it gets complicated is because of the di di edition of the t ID. Now it has an Associated state information which it arrived.

C

J

So, which has a lifetime but.

C

We have to think it's true whether it would be a good idea that thou with a lifetime zero and no path now creates their registration or not, and and out of my mind right now, I, don't remember the arguments. There are some and they are in the draft. I think I discussed that case, but they don't remember. Can we take that yeah.

K

Sure sure yeah thank you.

I

So Pascal like using that resulting changes in the draft are I.

C

Don't know yet, can we integrate that I say what question this.

A

Is part of working new bless call so, okay. So then.

C

A

The fourth common comments, educated.

C

So Raul, let's have this discussion remaining. Let's just clarify, I I think there is text on that because I kind of remember writing it I. Just don't remember what the arguments was so I thought remote conclusion is: if you disagree with what the text that's in document, then please raise main on the mailing list and we sorted out.

D

This is Sarita Oscar.

A

D

I was referring to Pascal's question and whether we should explain the working group last fall. May I suggest that maybe we could, after he publishes the next revision. Maybe we could have a short extension for one week for people to review. Does that make sense? Maybe gaben Suresh? What do you think.

C

Would that be semi, terrifying question? Would that be? You know dependent on the amount of changes that are put there right because I was asked yet what my question was not spared correctly? My question was: do we think we have enough review during this? Well group last call to call it victory. It was not about commands relation. That's.

A

How I understood it and I mean I'm, ok with waiting a little bit more because we know there are comments for coming. It would be a shame to not allow them to come in I mean it's just a few more to show yeah. Let's just wait for those comments come in and.

D

A

And then we'll deal with whatever happens just like any.

D

A

Word, unless called there are too many changes will do another work. Let's go.

D

L

A

Is so you're going to the backbone.

C

A

Of this one, no all.

C

Right so, like I said, the the RC 67-72 update, extends the operation of the update from a bad operation to a registration operation for services such as proxy and e. In the future. We may decide to use it to register all our things, but at this moment its proxy indeed, the bad man router, is the proxy and the operation to which the update enables the registration to next time.

C

Okay, so actually I just published refresh, because the document was reaching time out. We have a. We have an implementation. It has been demonstrated several times at the ATF, including at the 60s black test, with with a client by tomorrow, I. The way, certainly, we could use more implementation.

C

Certainly I would love to see more Interop, in particular, would love to see a open source, backbone, router implementation to test the interpolation between two backbone routers on the backbone, because what really happens is when you've got a mobile device, Wi-Fi or 6lowpan that moves from one 6bb off to another. 6Bb also moves from ripple, dag triple deck, which and and they are different 6lv are now. The two 6ld are need to agree on where the devices and that's part of the spec does right.

C

Enabling movement I have not tested that with another implementation, certainly because I only have mine, so I would love to see another implementation and and play with it. But that's where we are I also asked I Tripoli in particular dirty lever and experts to review those two specs. So this is happening we'll see what pops up on the mailing list. My goal, being you know this was enough reviewed for the 15 for world.

C

I would like to have an assessment of the dirty Levin world just so, if, if they see you know, if you change those three bytes this all these two words, this would be applicable to my face that it's not like that. That would be bad, so so I've asked that, but otherwise for I'm concerned I mean seems that this is well already for well complies called in very stable.

C

In any reaction there, nobody pops up, yes, I'm doing an open source implementation. Now.

C

Yes, you, where I know I know I rehabbed almost is almost any good enough.

C

Okay! Well, we haven't formed, that's when I guess and then there is the address protection.

C

Okay and besides I see you're in the room, so please champion. If there's anything, you want to add to what I'm saying so. Certainly this is the third piece of our story. This is how we perform source address validation, which is savvy because we have this state in the 6lv. Are we can validate, including a crypto state?

C

We can validate that the user of an address is the one which registered it in the first place, we can validate that when a dirt was moved, the new place were the new device which claims the address at the new place is the same as the old device, and we can validate that with some cryptographic exchange proof of ownership, basically of a token, which was given the first registration, that's pretty much what it is, so we can have revamped it quite a bit to be honest, its editorial, but we have removed a lot of extraneous text in appendix so there is a lot less normative text which normally would help later in the process.

C

We have I. Did this section about updating your C 675 and we claim that were review already for a review by at least first review by situated direct right, for which the chairs basically called around a sterile cleanin? Who has been helping us a lot in many situation if it is, and it was an experiment, fifteen five, if you need for security, you know well, someone was that the best time, but we expect that only after summer will take will get the feedback from zero again.

C

A

We have to what we have to get moving because actually took a lot of time.

A

So next on the agenda, the young moon would be applicability, a news case.

A

Hopefully I have the right slide since time.

M

Okay hi, my name is Jung Geun hyung, so I pretend they're sick, throw up capability and use cases.

M

Okay, this is the history or you know, or we discuss from this issue from IP f-89 meeting three years ago, so we present our first individual thread and idea from 94 meeting and we, the working group adoption was done in the last year. So this is the second revision. So I would like to focus of the goal of this document or this document. The goal is to help sick, throw secret instead of the patient by layer to contain technology and help a newcomer understand how six Lawrence Tech can be applicable in practice so useful.

M

You odd adapter, oh I, okay, at idea. So in this revision we try to include some practical six row technology. So it is the shuhite mesh and ricin.

M

So to update based on tour last idea meeting, so we got a several comment. So the goal is to make this document or shortened and clean and the pocus on a ability. So we specify six row. Then click in create canal OD in the third. So we make the LT MP. She is a example of the potential six row l2 candidate and as I said, that we include.

M

Yeah, we include two practical, six wrote implemented scenario. It is right, silent, schibetta machine and we add a new new section of Qaeda line for six row. Six repent altercation and we moved to remove sixth row use case example to appendix because we are focusing on the applicability.

M

So the some text poll use kissed is moved to appendix. So we add some additional ad throw command and the stamina are helped us many immersed. So she is Eddie's. So one of the author of this threat.

M

So in this document we have seven six role in great canal: G, Teddy, z-wave, really and actually, and MSTP NFC and PLC and I prefer 815 that echo so in this earth ring create canal. Agito Piercy is under the individual director, so I hope. So this technology and and the trapped probe regarding this technology should be some.

M

The working group document in this working group and we unmarked the LT MTC- is the potential candidate because we sub a and we solicitation or visual state the practical use of the six role, but unfortunately for regarding LT MTC, there are no or practical or six role or deployment scenario, so the storage' meet or mark is the potential candidate yeah comparison across the six road in color technology.

M

So there are seven technology and we want to compare so they are you seized and their technology and seven air and mobility, Luke, Lyman and security requirement and problem you climb on a radiance cutes requirement. Iterate, and there are some. Ladies: are she were trapped? So, as you can see, there are some come on common point and there are some different to a point. So if you regret this table or you can analyze or you can emulate, some increased acknowledge a poor, your target services.

M

So that's the reason we are made this the comparison table yeah and we add the new section. That is the guideline for nothing ipv6 there. So he target candidate upon new constraint, air to technology that consider running modified six row, six weapons Tech. So we must consider that item, that is addressing model, mq consideration ratio, s3 loading address or simon header, compression security and encryption and additional processing. So until now we have the seven ring clear technology, but in future maybe new wrinkle six row link layer technology could be made.

M

So in this case we should consider there are technology for these items yeah. So we add some top practical to six row deployment scenario. That is why Sun so this text or was up or her fight papi, he is the working in the white Sun Alliance. So you can see the rising technology is a based on this technology and among them why son, Pierre area network FAL technology.

M

It covers something related on their topic analogy, so we found that my son uses of six Rose Stagg is that is the advanced metering, infrastructure, ami and distribution automation ta.

M

So it is some interesting point there. Six row technology is really used in the real world and another six row. Deployment scenario is japanimation, so Jupiter mesh sophistication is based on piece of technology and so to fight. Mesh in smart grid is using the six ronita layer, so thick text or was a provide, is my credibility's and beer, and so thanks to their support for this document,.

M

So and we are headed some design space dimension for sixth row use cases, for example, deployment deployment, put the threading, topology l3, 1, l, 2 and something blah blah. So originally we described and we are right, some thought specific or the densities dimension for each or sixth row technology. But there are lots of context. So we delete many part and we would like to focus on the design space in here and six real use case.

M

As I said there in the main part, only one or use case or purely there is the smartphone based interaction with the content devices and the other use cases for using perjury that really and MSTP and.

M

Nfc appear ci3 pureed other person that echo that are use cases are moved to appendix so we make a district or shorten and of work clear to how to understand. What's the goal of this threat so Rachel, so thanks.

N

Carolyn, first of all, thanks for pulling all this together, I think it's a very useful survey of the field. I guess my question to you is one of those things on the list is not like the others, which is MSTP, which is wired and everything else is Wireless. Have you thought about doing some sort of I? Don't know I'm just wondering if there's some sort of a figure of Merit that might be based on you know, good put per you know times distance times price or something like that. I mean.

N

How would how would somebody, you know, choose a wired data link over wireless based on reading. You know your your document.

M

It's the clue kitchen so actually or we don't consider the differences between wire and the wire is so except MSTP. Other Canaries are based on the wireless right. You can see so feel free to sorry PLC. We also Piercey Piercey is about not the working. The document yes so MSTP and Piercy is the wire and other technology is Wireless. So, okay, also, currently we don't have text that are reflect your question command, so I'm thinking and I tries to how-to or reflector. So you can help us. Okay,.

C

Thank you basically I'm just reacting to what Kerry said, and there is some feedback from the real world in in the measuring business, where they will tell you that filarsky metering. There is not one link assertion that fits all situations, for instance via building. If you get lighten outside, if you don't have liner size, your concrete but yeah, the return for implementation and deployment that I got is that having a layer, 3 routing technique which allows you to actually use multiple sort of link, and maybe you deploy both and you see what works?

C

It's a lot better than deploying only one and sound a technician back later, because that's not right. So so that's really where the power of writing and using repo in this kind of use cases is so visible because you will actually use both wires and wireless. Okay.

M

M

Any other questions comment, so you can see. This document includes some the barriers in cleared canal G, so so I'm not export of that area. So we need some other expert to review and give some practical information of the six role or in chronology uses in the real world. So I'm also asking and the carry says that he gave us and other person emulating your input of this document.

A

Thank You charlie.

H

Okay, well hello, I'm, John, Charlie Perkins and do I just say when it's time to change, slides or okay, yeah I guess this time. So just a quick overview. We've been working on this draft to specify the delivery deadline times for packets in a 6lowpan Network, and currently we have a packet delivery deadline time and then also you can put in the time at which the packet was first sent.

H

And with this information you can make decisions about whether to drop the packet, because it's not gonna make the deadline, and you can also make determinations about how long it's been in transit. So it's a pretty pretty simple idea, and it's also assumes basically that you have a time-synchronized network. So it's you can have a consistent idea of what time it is.

H

So that's sort of an underlying assumption, and given that you have different networks that are still basically synchronized in time, even if they have different ideas of what the current time is, you could still do a transmission between those networks on next slide, please so well. After the last ITF, there was discussion about including ASN as a time unit, and so that seemed like the right thing to do, and so we put that in there was awesome, just some discussion about how to handle the reserved field. We didn't do anything about that.

H

It said just to say that it's reserved, but other than that the draft has not been changed very much, but one sort of pervasive change was to the draft. Currently is the name of the draft is expiration time, but I think deadline is much more accurate description of what is provided so in the text of the draft. It basically usually says deadline or delivery deadline or something along those lines, but the name of the draft has not been changed. So, let me except to say it's go for so next slide.

H

Please, here's a quick look at the format for the for the extension and it's safe. Well, basically, there's uh oh means that the origination time is present. D says how you can handle the dropping the packet. If it's, for instance, you you may allow for this information to be transmitted. Even if the deadline is passed, if the D flag is set to be 0, Oh question as.

C

Best can see well charlie. There seems to be a very supportive of this work yeah. There seems to be this. This assumption that you know how fast it takes to get there under the routers in the middle know.

C

If this packet is light or not light or or do you just want to drop the packet when it's too light, but having no clue at all, if you can make it because it seems to me that basically, you should drop the packet as soon as you know, you cannot make it not go all the way till deadline. Yes,.

O

C

That seems to imply that you know how long it takes to make it and my commands pretty much the same. I did last time is constants. If the route schedules, the transmission, all the way using non non storing mode all the way, it would probably be able to compute also the various hops deadlines and so use that recursively or use that along all the hops, so that each hop would know how much time it has to follow this package.

C

Just to respect the schedule initial end up doing that, then you have a source, routing versus short version of TS n days ago. Well,.

H

You kind of do that with TTL right I mean if you want to say that this packet can't survive more than five hours and.

C

I'm not talking about hops, I'm, saying oh, if I, if this route schedules this path with this source, routing information uh-huh- and it knows the packet must make it by that time. It can also say Oh first up, it should make it by that time. Second, hop that time all expressed in ASL, so you already able to to to give the network. What is your reference folding expectation you.

H

Know I can somehow imagine a network where you'd like to have that close control of granularity of planning, but I don't know how to do it and I think that this is going to probably get a huge majority of the applications, and you know, as you say, let's say that you have an intermediate point where you're one microsecond away from the deadline. Well, you might want to drop it anyway, but this is simple enough and useful enough that I hesitate to volunteer for additional complication.

H

Okay, so you can go out there and get the implementation has been merged in with open WS n.

H

It was presented also, it's a six dish by aisle I show and review them as a natural customer for our for this sort of feature and the implementation has idea for earliest deadline. First policy, which I think is one of the most basic kind of deadline policies you might want to use a question.

N

Kerry Lynn I'm just wondering if you've got any information on whether the dominant delay is killing delay or transmission delay. Or is this very by technology, I'm? Thinking back to Pascal's, question or or a response, it seems to me that in a mesh network, for example, to be very difficult to predict that all of the hops are have the same cost, for example, because it's.

D

Really, as you get closer to the.

N

Edge those those nodes are doing more routing on behalf of other nodes below them, so that potentially there's more traffic going through those nodes. Yeah.

H

I think you know, does you know if you want to take a look at sort of measuring the distribution of delay times through the router and then slice it up, but chewing delay time and other effects? That's a pretty good paper and as far as I know, they're you defending on which network different effects are dominant. I mean some it's it's. Your network is congested boom, your that's it and I'm, not sure whether you want to count delay because of interference versus queuing delay on the router.

H

Those are two different kind of things, so I'm gonna sort of maintain them. My blissful ignorance here and just say: well, you just want to say what the deadline is and then let it go. Glut go at that.

P

Q

There wasn't that carryout I want to answer I think it depends on the technology from the obvious and stuff that we did in six ish networks. The delay time is mostly queuing. You wait for your slot.

N

Actually had a second question, and that is how closely do the clocks have to be synchronized? Do you is it? Is it just within some epsilon? In other words, can they can I sort of you know oscillate within some? You know some s along.

H

What if I told you that we measured experimentally that the clocks have to be within 1% of the deadline time? That would be a total lie, but you could you could make any sort I mean you can decide.

H

We could get into the question of how hard is the deadline, and that would go back to this D bit that we talked about whether you're allowed to you know pass it along or not, and this is a valid point of discussion, but on the other hand, maybe it's a better idea to just put it out there and see what's really needed. You know maybe.

N

We can just catch up for like two minutes tonight. If you're going to the social I want to tell you about a project that you know. Oh.

H

I'll be at the social. If you want to.

D

A quick question on the time synchronization so I understand that it works on 60. So, but do you have any recommendation of a time synchronization protocol or how time synchronized? What time synchronization mechanism should be used for this draft? No.

H

No, we don't have any additional implementation. I see now I shouldn't say that, because you know, I didn't actually do this implementation. It was done by Elijah and folks. So presumably I mean, if they're running on a normal platform, you can have a time synchronization protocol running, but I don't have any additional information about the specifics of that I.

D

Think good to have some sort of either boundaries that I think Carrie was mentioning or some sort of information of time synchronization because it seems like if this is a dependency for this for your draft. So some recommendation on time synchronization our reference good.

H

Okay, that's a good point. I bet if I compose a message to the mailing list and we have discussion and we can insert a description about that I'm. Looking at you.

Q

R

Q

So Thomas time, synchronization I I can answer exactly for this implementation in this implementation see the order of across the network around 100 microseconds in commercial application commercials around 10 15 microseconds across 100 node Network, but but this doesn't have anything to do with this draft right. This draft just says time stamp in time stamp out and whether you're well synchronized or not well, synchronized, not it doesn't affect that.

Q

Olders draft is draft, just timestamps, so I think I think you know how network is synchronized, doesn't really matter it's to the application to say: okay I just received a packet which has this timestamp. What do I make out of it? I know what's wrong in my network, I know how much I can trust it, but this draft will carry it regardless right. It's independent right.

H

So what I understand is pretty just to make a dumb example. Let's say you send out deadline, saying here and say: I want to have a deadline of 1 microsecond from now and you're just you're not going to really get very far, so we could insert some language in the draft that says. Well, you know the precision of your deadline. Time has to be affected by the precision of your time synchronization, but that won't change the format of the packet yeah agreed yeah.

D

But you know if the time synchronization persists. Precision is not very accurate name time out a time in time out may not be very correct. So if you specify some, you know range of time, synchronization precision that would be good enough, ideal, I believe, but I think you can probably discuss it on the mailing list. Well,.

H

If I I know, if you see something on the mailing list, please comment all right: yeah, Oh, Gabriel.

A

So I had a question trying to understand the semantics of this thing. So looking at the D bit, you turn it on then you're supposed to throw away the packet if it's beyond the deadline correct, but the operative word there was should so do applications know if they have to rely on this. If they can maybe line it is it meant to be just to get rid of extraneous traffic on a best effort basis, I mean what kind of semantics our application expecting or are there's gonna be?

A

Are they gonna be issues if there are duplicates for account, because somebody did not obey that should as they're you know open to do? If it's a should recommend that we changed, should immerse no I'm asking because I don't know how it's used so I don't know what applications expect. It was a question about that application suspect very clear, Chris semantics, then I guess a must is probably a better choice, but no other application to with these things. Okay, it's meant to just get rid of extraneous traffic. That is fine.

H

Okay, I don't have know how to answer that, but I mean we could ask six folks. They might know sue.

Q

Thomas I'm not answering six I'm, just answering thinking out light about applications. If you send a packet with one of these headers in there- and you said the D baked for drop after deadline, I mean you expect it to drop right. So.

H

You should you should make it a muster set of issues.

Q

Well, if it's a should, then I said the bit and what happens so either I said I, don't set it and it just carries on or I said it and must be dropped. So I would say that must otherwise I don't see. I'm! Okay, with must.

H

Okay next slide so well. We would like to request working group and Roush and adoption of this draft I think it obviously has a purpose. Sql. Do you have a question amber.

C

A little bit this is a I've, never I forget the name, but this is a header that you can ignore right.

C

So how comes you can expect the most operations and something you can ignore? I'm.

H

I

So I I don't really see how that can be a must like or a serious must write it because it's optional the process right. So if you put it in, you cannot expect it to be enforced. So this is like ha by hop options in in v6 right.

I

You can say you can say like if you don't understand this drop the packet right, but like there's another RFC or other CD which says that, oh, if you don't process how, by off options, don't do them right so like you're in a very, very similar boat like because you can say must if you want right, that's okay with me, but the sender still cannot expect. If you sit the D bit that the packet will actually get dropped because somebody, you won't understand, dislike deadlines, option right so.

H

Okay, uh can you say a little bit more about why some would not would not understand deadline option so.

I

Let's say somebody doesn't implement this draft on the pad, does not implement it. No, okay, not at all right so, and you said that II bet do you expect it to be dropped or not right, like I would say no.

C

I

Cannot write you cannot so like you can put the must in or a shouldn't, but the thinking needs to be like you cannot depend on this happening. This is still best effort because you don't know who's on the path who can actually recognize this right, because what are you saying is like, like I, think, Kerry said it best right. If you understand this, like you must drop it, that's that's different from expecting it to be dropped. I think that is Gabe's question right like how do you expect it to be honored right.

C

Well, let's Carrigan, you know Charlie in in family. There was this discount eligible bit. If, for those who were born that day, there's there was a de bit in the dlci and what that would say is if you have to you: have those five packets in front of you and you've got time to transmit three? Well, if two of them are the D bit, then the choice is clear: it transmit these three others, so so, basically, this looks like a debit with a time bomb on it.

C

Let's read what it is: I mean before that time treated as a normal packet after that time. If you really need to discuss something, then discount this one, preferably but that's a behavior that that can be enforced by those who, but the behavior that you cannot get. Yes, let me make an application that is pro by design is certain that this packet will not be delivered behind that to beyond that time. Unless you get this thing implemented throughout, then this application will be broken. Okay,.

H

So this is, you know, you're discussing the end-to-end, behavior and part of the value of this is in the intermediate nodes in UTS right, so if they have to, for instance, do some preemption or anything like that they might preferentially that's.

C

What is all about in frame right, pre discount this one prefer, because the application knows that it will lose a lot of value or it does a lot less value that these are the packet. So if you add those two, this one has deviate. This one does not discount the discount eligible bit or.

D

C

Thing after that time, this one becomes preferably you know you inform, read that it should not be surrender after all, right I.

I

Think Shelly, the point that, like you know, I was making at least I think like most of us are making, is that the must doesn't give you much more than it should, but that's it like everything else. What you said is right. You.

H

Know there have been entire hours spent on mas versus, should.

H

And one argument is that people don't care about, should and I don't believe that argument, but anyway, please Peter.

E

Founder stock, if I want and deadlines be a little.

S

Mike Peter.

E

How much talk? Okay, sorry, if I make a network or make your own real time to be executed, I make sure that this draft is there in all the in all the notes. Yes and then I expect the must okay.

H

E

H

Can I can I say that we should discuss this in the mailing list and then they will just do whatever people believe? Okay, so.

Q

So, as John described, somebody is pointing out it we're 20 minutes 800.

H

K so please add our draft.

A

Yeah so not sure I'm we're ready yet to go to go for call for adoptions. But let's let me continue the discussions about how applications work with. Is that sort of thing a little bit more and maybe go for call for adoption after that, depending on how the interactions go great.

D

At this point, it.

A

Seems a little bit like we need to discuss, but.

A

He also a networks.

O

Okay, hello: this is Jang ho and today my topic is transmission of ipv6 packets of replacing networks. So this work was done by me together with young guns charging and the current leads version Wow, and today we would like to record our she okay next page. So what is PLC PLC refers to power line communication. This technology uses the electric power lines for indoor and outdoor communications and Jesse has been widely applied to support advanced metering infrastructure.

O

So the figure below is a simple example showing that the smart meters transmit data to a concentrator directly through the neck power lines. Okay, next page, so I presented this route last time in Chicago and have received comments from some interest and theory and based on their comments.

O

I updated this chart, including that and the following pass has are being updated, including the addressing modes, the unicast address, mapping enable discovery command frame, header and connectivity and topology okay, next page so comments after IGF 98, so here I would like to stack Sumitra and Gabriel for building the connection with itu-t studio. So and finally, we. Finally, we got the support from Stefano and the theory. They are tube chassis experts from itu-t and they have reviewed our drafts and, following on your comments, so here I will not go deeper into these comments.

O

But I would like measuring that Stefano in the chair of SEIU 15, which is a peer system which is POC study group in itu-t, and he suggested that they sent the informal feedback at first. They exchanged iesson schematic of issue. Ok next page, so following are the updates of this route. First is a first is the addressing modes. There are two kinds of address using classic devices.

O

The first is the 64-bit. They globally unique knowledge is, and the other is a 16-bit shot address. So in a chassis network, each device joins the network by using the long address and then communicates with other device by using the short radius, okay, next page and for regarding the unicast address mapping. We use a 16-bit address only and for 100 1.2, since there is one line referring to on the FC phone fo4. So we could taste simple format for the shuttle, yes here and for the itu-t co3.

O

We put a penny in front of the shuttle. Yes, okay, next page and in this rushing we updated and use and subsection, namely the neighbor discovery, for they are Shibui 100 1.2. There are two kinds of routing protocols: why is people for layer 3 and the other is low energy full for they'll arm for Nia? So if they on ripple is used, then they the discovery process show refers to day six on pnd and if I, just AP basics is used to assign addresses, then they duplicate address.

O

Detection should not be required, and if and if they want to uses the coexistence mode with Nara or three nano Nene low-energy protocol is used. Then the address registration key finding the section attendee should not be used because there is already a defined pampu shocking protocol, but those mechanisms in six of handy for managing the ipv6, traffic's and corresponding header compression contacts information can be used and again in the at UT Nana or three since they load energy protocol is used and there is a predefined campus job in protocol.

O

So we don't use a address registration defined in Section plan D. What smashing sure that the arrow 3 supposed a section pen context talking, so we specify that those mechanisms for managing the ipv6, prefix and corresponding header compression and context information must be used and, moreover, for sending the ISR a the chassis devices must follow the corresponding sections of a six-pack, Andy. Okay, next page for regarding the command frame header, the scarer was defined in 9 or 3, mainly for the bamboo shopping protocol and for the load ng protocol.

O

We are so you know we specified order in the header format. It should be put in the last position if more than one carriers present and it should be put before the 60hz harder and based on Sumner's comment. We updated our example young our draft we put the in between the fragmentation, header and the objective header we use, and yes city.

O

We use an ESA dispatch, tried to indicate there's another despite a follows, and this one-bite is used for the confirm whether your stored command, ID and then unfollow stickyman payload, okay, next page about connectivity anthropologist.

O

Here we have starchy and mesh. But in this version we clarify that they are on logical connectivity, not physical links, because there are some confusing that if we consider the day on physical connectivity, the PRC devices are connected in a trunk link. But if we consider the transmission just rains, then they are connected and logically induced are tree or mesh topology and with dis, some name changing this section so that it can be inconsistent with nano tree which had they sealed previously for coordinator to encode, an inter and the node Japan device.

O

Okay, next page and last on, I put I put a crashing that if we, if it is possible to include more Casa standards and now the answer is yes. So currently we have 100 102 and nano or 3 and then Mochis a standard night color and panel for prime and 100 1.1.

O

Can we considered and we only 900, chuji any and has been ruled out, because currently there is no deployment and about our next step will be like to call for the working group adoption so that we can have a further discussion with at UT, and I should officially and again your comments and feedback are always we account. Thank you.

C

Well, thanks a lot for this work, I mean we had an early draft by daniel papa wrong long ago and that's the activity stopped, and that was a shame. So it's very good that you're taking this overall, I thank you for this I'm curious I, initially did the first work. I was well off, was for I to play in 1901 dot, yes, which is the classical stack that we care for I, know: I, Triple, E, IGF, usual stuff. Now I thought that the itu-t was designing its own stack and and what kills me?

C

What bothers me, what I don't ten years if we do anything normative, which we feel is normative at the IDF for the stack of our itu-t G 99.3? Is it like? Is it a norm to be the standard to be used by a ITU people? Who would want this specification, knowing that the ITU does its own I mean? Is it a competition to the the piece that you're defining of or the ITU Network? Is it a competition to the stack that baby out I mean? How do you position your work versus their well?

C

Did they ask you or did say? Oh, we I mean I, mean I, don't want to put the ATF in a weird position while we are defending a stack over or something replacement that I already have right, because they already have six lap on the belt in 1903. That's why we gave them the Esk bite et cetera. So why should we be doing here?

C

For that I mean competition.

O

We are here, I would like to put because currently there are many plastic devices and for guarding the Nara ban. Piace I also have let's say the 101 thought to the GCP area, prime, and what I thought they are 101, but Wow, so I currently for the vendors, just it's very difficult to find a really good, pure standards for their use. So I would like to combine them together in a single drop, and if there is some thing in common, we can put it together and that's.

C

But imagine you you say you must use ng RI is to advertise this right in your document. Then the ITT already finds a stack and they say how they do things. Another is a collision there or something right. How do we resolve that and are we are we mon dated for doing this work I mean? Is it something that that the ITU, the chair of the itu-t 9-3, has asked you to do? Thinks it's nice that the HF does this I'm just concerned that we may have an issue trying to define something for them?

C

That's all! It's just a question of the answer. Okay,.

O

So um so they did not ask me to do anything, but when I send this a job to them and ask them for the review and if you're very happy not sure they can find someone who are still promoting their standard and they're still maintaining their nanny Austrian crime. But currently their focus is only on their one day or two, and so no they have a some guy at you and focus on that. They adaptation there. It would be better yeah.

A

But to be clear now it was not them that asked us anything at all. I was more. This was brought here, we're wondering whether there's other people in the world doing PLC. What do they think about this? Just like we asked of an FC just so grass of any other link layer we want make sure bluetooth et cetera, want to make sure that we're not rotten.

A

You know conflicting with them, so we want to make sure there was no conflict. We got him involved. The ITT looked at this they're actually offer comments, but you know we might want to clarify a bit further, how they see it because they, the comments, have been fine, saying: hey, okay, this none of them has said: oh, please don't do it or we'll ignore it. Just been helpful, constructive comments, but your point is about clarifying what what is the function of this, which is whatever it may.

C

Be I mean because maybe.

A

C

Know I can read the document from somewhere else, which is kind of informational and describing what we do. On the other hand, if somebody wants to know that no, it is something we are doing here. I would not be so happy, so it's just do they understand that we are doing a standard track document which, with the norm of how things should work on their network or do they think that we are just describing what they do? You know as an informational that that's a very different beast.

C

So just yes, well sure we clarified us make sure there is no later yeah.

A

Yeah sure yeah yeah, and we we definitely will do that. You did also mention some, but liaison liaison has a very specific meaning in in the ITF and that might have come from them. They wanted an official liaison, there's some liaisons already from certain groups in itu-t I. Don't think we want to go there for this, because the way that it's going so far seems to be okay, the best way to collaborate, it's actually not through establishing only big liaison zestful, that's for other types of interactions, but the best way is actually for the people.

A

Do what you're doing exchanging information comments going back and forth and so I think we should push forward with that, and hopefully we don't have to do a formal liaison which would have to involve IAB into because I'd be handle to your son. So that's another process altogether: okay, okay, yeah.

D

I actually I agree with Gabriel on the on the collaboration part with itu-t Pascal. We we, we actually got positive comments from the PLC as the working group FG in chairs, so so yeah. We we, our collaboration and I, think if there is anything new that are different from what they are working on over there, so that piece there should be communicated and a greater whole. Otherwise it is, as you know, this is similar to any other documents we have like NXT I mean.

C

Samhita position again, this is really where my confusion is. You know if this is an informational document saying what they do. Then it's it's a separate document, just informational if it's something that specifies new operation in an O, Matic fashion, yeah.

O

There is something new right.

C

And and if we do that, then we need to make sure that they expect us to do something, no matter, if not just like, for instance, in LP. Well, we are. We have informational document about other technologies, but just descriptive. Okay,.

D

I

And a point like, if are they doing like El, toast or l3 stuff, like I questioning, they are doing l3 stuff or not?.

I

I

If you can say that my Kia.

C

My starting point is: they came to us some time ago and asked for this escape field, meaning that they have defined the way they operate. So if we ever, if we, if this is a document which describe what they do, it's a different path, it's not just on a track should not merge with this other track piece for 1902. It would be a separate document which would just be informational.

C

Now if we want to standardize IP operation about their network, knowing that they've already done some of that I don't know to which extent, but they've done it. Then we may be in competition, so the if they are very clear that we are doing stellar track documents here, a map here, you know if they think we are just doing informational, describing what they do and all of a sudden they describe that. No, no!

C

We are prescribing operation which are competitive with what they were they yet then I'm sure there will be a lot less happy, but I mean maybe it's all done. I'm just was just asking well.

O

Here in this route, there is some content that refers to the annex part of 11.2, but also meant something. This is my own and hopefully, I will contact Li and they actually sign and extra and that they have no conflicts. So.

A

Just in the interest of time, I think we should continue on the mailing list definitely want to understand a bit better. Your your concern, but I think we will resolve that offline, not not in them in the next minutes or hours or whatever so I think we can definitely move to afterwards. We saw these at this point.

A

If it's resolved favorably, we could think about a call for call for adoption and then in the working group, but we would like to clarify that point first. Thank you very much.

A

M

A

Body area networks, Sajjad.

P

Hello good afternoon, the first slide. Sorry, so you can see. This is the first draft which is which belongs to over less body area, sensor networks, and we are looking to it with respect to ipv6 how we can make it the part of coyotes.

P

So the current status is this: is it for mislead off you just uploaded in the zone and we're looking for? We will think about the standard track, efforts and sedation, and straight back out requested yes next time. So there are different motivation. I will just describe the few ones like one is like population is increasing day by day. So this is the one of the survey like 761 millions in 2025 people. They need health care, they need remote health care.

P

Another motivation is regarding GDP like be on health care. Expenses are increasing, so on a report. You can see the different countries, they have different duties, but the learning stations like it's getting very high, especially with like with the USA. It is surveyed like it. It will approach up to 20 percent of GDP in 2022, so these are different indicators like which we are looking which are motivating for this document.

P

The next I will just introduce the wireless mode area since the network with respect to I I Triple E 8, u 2, dot 15.6 standard, which is active standard. No, and it is standardized in 2012. It is as our lotion now it is it based on mac and phy layer. The main the main thing is like it incorporates the better penetration through the human tissue without creating any damage to the human tissue, and it follows the approved mix band, which is medical implant in kitchen service penned by USA.

P

So it's uh it's following the specific absorption rate guidelines which is useful for human tissues. Next slides, so here you can see the different application, different use cases for wireless. What is a sensor networks? A patient can be equipped with different sensors with this standard is specific, with them. Countable sensors like when you and wanted the sensors inside the body- and there are different use cases you can use it for patient care. You can use it for elder care center, you can use for the fitness and fitness for the supports is the next slide.

P

The scope like it defines the scope. It is basically a short and low-power standard, like ranges mentioned from two to five meters and it wants to work on value low-power. It is using I use some bands and there are different bands very low back low frequencies band I mentioned I will show you in the next slide it compliance yeah complies with strick non-interference guidelines. It is taking different guidelines with respect to medical applications as well. Is next slide.

P

So why would you need to standardize this sort of application with the IDF like? There are a lot of propriety solutions which are available which are doing this this stuff, so we just want to make it the part of standards, so it can be used by everybody. There should be a standard like which they should be a standard body screen solution like current solution or optimized for different type of applications. Beside medical, like bluetooth, is optimized for voice links. Zigbee is optimized for the sensor industrial sensor.

P

Applications, Wi-Fi is optimized for data networks, so these are the general general focus points which they are doing is the next slide.

P

This is the just a comparison slide between different the existing standard and, if you to dot 15.6, you can see if you want to work on very low power like one milliwatt or less than 1 milliwatt, and you also goes it also. You can also get the high throughput like you can use ultra wideband in this standard, but still they are following the human tissues guidelines. So you can see it is low-power and indebted. You can achieve different again.

P

The purpose of the high data rate it like this diversity of the different medical sensor like if you are using ECD, so it's contained different different sort of cables. So you need high throughput. So that's why it the throughput that data it's it's varies in this standard. The next type. These are the different frequency bands. It's mentioned. You can see the human body communication there for different frequency band.

P

It has different, it says: t type of file layer like this narrowband ultra ultra wideband and human body communication. This. This is the speciality of this standard makes yeah. This is channel access mechanism, it's a little bit different like other. It's it work on. The same pattern like 15.4 is working on the super frame that channel.

D

P

The super frame, but for the medical application they have mentioned, like different phases like EAP, like emergency access phase, you can see the different EAP in a single in a single super film, then the random access period, and you can also see the content, access contention.

D

P

Like cap, which is also available in 15.4, so you it's little bit different from 15.4.

P

Yes, next sideways, so you can, they are obvious, officially benefit for if we use it with ipv6 like we have the addressing flexibilities like, and we have auto configuration the solutions available, simple like if, if you are connected with ipv6, which means you are part of in 10, like you're part of you can be a part of IO T's as well. So that's why we are looking for ipv6 there next slide, so there are different issues like, but especially we are meant to like the packet size like frame size.

P

It's in the standard at the frame size is 256 which is different from 15.4, which, like it's 127, 127 bytes. So this is a few few issues. We need to look if we want to make it a standard track like how we will deal like. Are we going to adopt six low 6lowpan or we are going to look to the compression mechanisms? How ipv6 deal with it? So is the next slide? Yeah they're different industrial?

P

You can see different industries which are adopting and making the chips for 15 15.6 XS, instrumental instrumental all devices, so the work started so.

P

Oh, you are the next yes, so this is it the brief introduction for document so, first document, so it's still informational.

I

So subscription no I hat I read the draft and there's really nothing in there about transmission of sick. I pv six frames right. Usually we need to ipv6 or fool. It's like a set of things. You specify. You know the unique I said this mapping multicast address mapping this one only has informational text about body area networks and nothing else, so is that the inter, after or like whatever intending.

P

You know like because, because it's just uploaded in the zone and just want to clarify the need like that's, why we mentioned about well, is body. Races are in Redbook as a focus point, but you are right like we should also mention about the ipv6 sort of things how it can be incorporated, like we have discussed in the group. But would you mention in the dock? Okay,.

I

No, but all it says, is like Oh: 6lowpan does not work on this right. It says the frame format is different. 6Lowpan does not work right and that's it it so because, like that, the title is kind of misleading. It says like transmission off dadada, or something has to specify all this right, and that's that's why? If you're talking about the need for using doing 6lowpan for this, then it's different than saying it's transmission over right. So, like.

P

Our like our focus is on like to use for this standard for the medical application, so yeah.

I

P

In the second step, yes, you all right yeah, we need to look in detail or we will manage it with ipv6 and those sort of thing and we're working on it. But we thought, like after meeting we will add, we will update the talk to comment just to.

A

Share point of view: there's nothing in here for the working group to do so.

A

This would not ever become a working of document because there's something for us to do so. It was a good informational presentation, but.

P

Yeah for the timing is informational document, but we will look for the response and the feedback if we get so. We think for this. Oh yeah.

A

So for next time, think of there's something for us to do then yeah. We should work. You know think about that. But if there isn't, then that's fine that.

C

Pascale that pretty much it goes with my questions. Why I mean do you expect? What do you want to build on that I mean? Do you want to be able to image? Do you want I see that there is need for reliability? Would you like us to put 60 on that's why, as opposed to the classical 15 for, but that it would that fit the kind of application that you have in mind? We need to know the applications. I mean I.

C

I really can figure out a mesh of that on my body, I yeah I think the idea is quite cool. I, don't know what I would put on a sick.

P

Secret dish can be a good option like exactly.

C

If you were to describe this and and and mostly you know, I want to do a mesh sixties mesh on it here is, what's missing from six stage to get this mesh on my body, then we could act on it right.

P

Again many like the purpose of this initial document, is you I, like the importance of this standard like which is, which is not the part of this? So let me highlight this innate I know we are looking for those solutions.

T

A

K

A

Up mic on this one: okay,.

K

Yeah, thank you. So this is Alexander Petrescu I work at CIA in France I, looked at a draft, a very nice draft I.

D

K

A frame format section there are about 300 bytes in this frame. Is this the maximum transmission unit, 256.

D

K

So the MTU on wireless ban is 256. Yes, I think this is a good starting.

P

Point yeah: that's why I mentioned in the last night that, like the issue like we have need to think about this? Okay, we need to feed back. So we can give space to this standard, which is important for health care application and else care system.

P

Thank you thank.

A

You thank you. Now we get to the final part of the meeting, which is going to be on fragments, Australia.

C

Okay, so this is, this is a a discussion that we spent a bit of time. Last ITF I exposed the number of drawbacks of the current fragmentation as disclosed in RC 49:44, which are mostly problems in the route of the world.

C

Unless we get clear signals from the mesh underworld, maybe the mesh renderer is perfectly happy with 49:44, but that creates number of situation in righto. Well, that could be improved so I.

C

Won't we do the slides I did last time, I left them in the slide where in appendix so, if you don't know the slides from those from the website, you'll find that after the apparent end of it, you have all the slides of my style of last time, but I will basically use your knowledge of what you said last time to basically look at the prices are and what we need to solve right.

C

Please next line so then again created that last time and yes that there is, there is another aspect which needs to come for is that for the LP one work? We have also started working on fragmentation and we have been thinking of a number of solution, space things which could more or less apply to to the work here so like windowing and vigil retrial fragments. So obviously the use case is quite different.

C

Decision will be different, but we may take inspiration from the world even though they're a bit of concept for fragmentation, so initially in the first is a very very long ago. First 6lo network, there was some TCP and the reason why there were some TCP is just because of this problem with fragmentation, so people would rely on TCP because they would get MSS Mac segment size.

C

Now, the drawback of TCP, obviously mostly issues and very small chunks of data, is you have to pay the price of anak donation back every time and that's way too costly. So the on the industry moved to UDP coop etcetera, which gives the application the capability to request an ACK really when they want it and usually they provide. The AK is not is an uplevel act with some semantics which come back? Not just hey, I got you, but I got you. So here is what you should do with it.

C

So we on a space where the acknowledgement is also mostly loaded with application space information and on the side of your brain. Remember that we talked a lot of streamlining last time, explaining that the packet could actually be reassemble, that the other and much faster if we can streamline fragments, are supposed to recompose at every hop, and actually this gives us storing. Well that we can use for each wise.

C

Just rule of thumb is the application is not chance if we can live within the time that we gain by streamlining following our or which one's it next slide, please.

C

So the summary of the plan that we discussed last time, prime one, unless we do something which is not documented in any standard- we need to recompose the packet at every heart because we need to build the IP base. That gets us to write it and then pass it to the next stuff. So we need to fragment parcel next. Our precompose fragment pass the next Opry compost and that's what create? Is this very huge latency and buffer bloat? In the etomidate note?

C

So we basically want to be able to forward all the way there is no control at all. Nothing which tells you how to pace your fragment so to slow down when the fragments can stay in the air so implementations and to push on the fragments as fast as they can, and there is no sort of interference and before bloat, which can be created inside a network again.

C

Not only that, but if you have multiple flows which cross in the same row tonight we'll have to store buffers for all these fragments at the same time, and we end up with the classical congestion loss that we see in switch Roberts but know with the number of buffers that we are there. That gives us last Iraq Utley and using a fragment as the side effect that we know that the end point will be Lear waiting forever to get the fragment it's waiting for, which is my last point.

C

If you have some loss, then the recovery SAP that come beside will have to time out and actually that's experiencing this, which led this RFC 73 88 about which, which now I obsess debug Network. That's because they did. This is an MP thing that they found that the prom they had was fragment loss, question David. Moving to next slide, I suspect.

R

My about to reveal and I know next to nothing about 6lo. What exactly.

D

David black.

R

Tears vwg working chair trying to understand what's going on here, so what.

C

R

You're trying to do with that first bullet: okay,.

C

So I mean that's when I'm skipping, what I said the previous meeting every surface, but you get this mesh of 6lowpan devices and you're doing repo rot rotting over it and now you're the the 6lowpan, the classical well to 2.50, not for network justify, will have 128 bytes of MTU, which means that if you want to provide the expected service of 1284 ipv6, you will have to fragment if I want anything which is in the order of 100 bytes and above.

R

All ok, yeah yeah, some.

L

R

Repeating or something or audience, l2 fragment. That's that that's really important.

D

So what's happening.

R

Is is the transmitter has done an l2 fragment and now it says at every l3 hop. So what do you want to do with the router.

C

Shim so so you've got this. Ip you've got IP which sees 1280. Then you've got 6lowpan, which cuts it into number of 6lowpan fragments right.

D

C

This thing will do multi hop over other the ripple Network. Now, if you wish.

R

Or Dell is launcher. L2 nothing's gone wrong because I go I, go hop by hop over of over l2 I hit. The next I hit I hit the far end of the link according to according to IP, and at that point, I need not.

C

Because we are that's what we call right over right: every radio hop is a router, oh boy,.

C

R

That's what I'm afraid you're doing? Ok, what I'm right? What let me, let you finish your presentation, but oh boy, right.

A

There out over and there's mesh under when this was written, it was for mesh under in which you recompose at the very far end for route over there's all these problems. You.

C

Put the packet at every up all the way to IP layer and you've need the full packet there to rot it to the next hop. Yet.

R

The right and the problem, the reason said, oh boy, is for what gave of the subscribers were out as we're out over your your effectively asking a router to route and forward no three packet, an IP packet that it hasn't actually received. It's got the header and and exactly.

C

R

U

T

U

Brought in Reigate load broken record mode. What has what what's risen.

O

The first hello.

U

What's written there in the first village has been possible for over 10 years now.

C

U

C

You're afraid of David, it's been possible, it's not been described and people it has been described, I liked your book and it was missing the well. We can.

U

Just get that offline, but you don't indicate that you need to swap late now. There is a difference between an implementation technique in their knowledge of document that defines a protocol. There is an implementation technique that avoids point one. Yes, okay. Now, of course, we still have to work with those other points, no point, since we have to think about what does it actually mean for the first fragment to be sent the next router actually forwarding that first fragment and at exactly the same time, the original sender sending the second fragment.

C

U

C

Point that we have done everything that there's nothing to do for the first right I can do you want to keep that my own sort of for later? The point is I agree that it's look mostly a local behavior that doesn't need it's done at least to the be done well, but it's about creating a state for the next fragments and, and so what is what what requires a protocol is. When do you know that you can clean that state? Yes right and when.

P

U

C

Well, the last point about fragments recovery. If you, if you have a fragment recovery, then you can use that piece to kinder state and that's when there is a relation. That's may also to you. Okay, so.

U

The there is a number of implementation issues or timing here.

O

Yes and number only.

U

Would be a good thing to think about those yeah and if I hear what you just said, you want to conflate them with retransmission I want.

C

To confuse well now that really the two points in the middle I really look close to one another. What is conflated is the recovery and the solution to the recomposition, because you can use the the message back, which enable recovery to also signal undone I got.

U

C

Can clean the states in the middle.

U

What observation is that this message back already has been there for a long time. So we have. We do radio layer X of all of our packets in all of our adaptation layer, packets, except for Mallika's, including fragments, so it has.

Q

U

There so you are asking for a different kind of additional retransmission. It's.

C

The okay 6lowpan wide and 12: it's not a one hop thing, but it's a multi hop thing needs to be very, very clear right. Yes, that's why we talk man, it's a lot of a pro right, yeah and.

U

I'm going to say the next thing, when you're at the slide that yes.

C

Exactly and that would that would be next slide exactly so, let's keep that falling slide.

S

Spencer Dawkins, just as a nosy individual, just to help me understand because I haven't read any more of this wrath and David has are we talking about many I mean like most packets being fragmented, or is this a relatively small number there.

C

Are four use cases in there? Mostly two of them in once in each direction requires fragmentation when is reflashing of devices quite rare, and so that's going down to the devices and the other one is. There are some example of IOT devices which will capture a lot of data in one bird-like vibration, stow, that in a file, and then you need to stream that file all the way up. Then you want to do that. Some blocks most of the normal IH activity, will never never fragment.

S

So, and and the last question just to help me keep up so if you do have the loss or you today, if you do have a loss, are you retransmitting the whole packet or just a fragment the.

C

S

C

S

C

Know it's not even that it's today's cost of energy associated to that a lot of latency, because this timer switch right timers are infinitely long because this network is so slow and it can vary. It be very variable, as well so to timeout values about huge, but but the other drawback is mostly what touches us at the sixth upper layer. Is the devices are usually constrained right.

D

C

You're, sending a new image to to this IOT device and you're losing just one fragment the timeout is actually very long and usually all the buffers are clogged with things like that, and now you can't receive anything and that's really what created this RFC 623 de H. Yet because he was wondering Jurgen was very wondering why he had so poor performance on his six level. Network and the reason was once the world is a packet being fragmented. The noise ratio beam being what it is.

C

He would consume all his buffers and they would drop the other packets you see, and so so he could debug that because he introduced as an MP in that in that device and that's that's the cool.

B

C

U

Could see cramming there we say coming one thing: of course there is a retransmission per-fragment, which is based on the radio lahic on the one hop. Yes, all.

C

Right, so we cover one hop. We don't cover the mesh, I, hope I wrote it in the first slide, otherwise I hope it's very clear document. Now, if you progressed a little bit on the network and you don't have a knack back I, don't know what then, if the packet gets lost in the middle of the network, you're screwed, if if it's after a few retry, you can see progress, you're, not screwed,.

J

And so I I think it's an important problem to solve. The bond statement is important, but I have some reservations towards the solution. We have not your assertion, but they would take care of the meaning list. So so one one of the primary disadvantage Caston already highlighted about the sender and receiver boat for forwarding at the same time. So while the first fragment has been sent from the receiver towards the next table, so it increases the collision rate. You mean the acknowledgments back or American.

C

J

A sender is sending a first fragment. Yes, then we see what receives it and then sends it forward. But by the time the the sender has.

C

Otherwise, they rely right yeah in the receiver. This this was not the problem. Currently it's one of this, it's actually fragment flow interfere with one.

J

C

J

C

J

Exchange, it won't be a problem.

V

J

The ANA at the problem that I see is with regards to single buffer strategy. If any of the device in the net in the in the path has a single buffer. Okay or let's say the sender has a single buffer, then, unless or until the complete, all the fragments are been received by the end receiver, it has to wait yes and.

C

That would take substantial long time asking anything about the buffers right I mean. So it's not prime again solution. It's a prime.

J

Prime right I'm just saying that you know there's a scenario where the sender has only a single buffer. Let's just a constraint: node, oh yeah,.

C

That's the three: what goes priority since right.

J

So let's say it has ten fragments to forward and now, in the current case, it forwards all the ten fragments in a single the single stage, without.

D

Any weight, yes, it pushes their service yeah. There is the yeah. There will.

J

Be layer, two acknowledgments there will be all so, but the point is now it has to forward it five hops down the line. Oh, it will take.

C

J

If you slow them, the other, the source- yes and you won't be so the source has to now keep more a bigger queue for all the other packets which are offering over the time through that's right, yeah.

A

I just want to good take into consideration it's just just a higher orbit here. I'm gonna cut the line right now so after Thomas, because we need some time at the end. We want to actually ask some questions. This is not about designing a solution. This is not about a solution. This is just about. Here. Are.

D

Some potential.

A

Issues and then the questions will be more like. Should we do something about it? You know that sort of thing right, so we want to get to get a sense of the room of what whether we should do something about this problem. I'll.

R

Be very quick, Thomas.

Q

I think this is an important issue that we need to address. It's a major major pain to have to reassemble at every hop. It's all kinds of problems with buffers. How long do I wait? Stuff like this and I've read your draft right before the meeting I've sent a kind of a very detailed PDF with melamine aiming points. If carton says it has been done great, but I like to understand how either we can have a side mini I mean we need to do something about this I think we agree.

Q

It's important there's been coming up so many times, I'm waiting for something I can use, and and and so either we come together this week and we figure it out or I. Don't know we do an interim, but I I mean this cannot be coming up again and Singapore and again after that, we just have to solve this thing. It's not hard.

C

Can we move to the next slide and then we'll have a lot of time for question response so I, based on the points that were released on the Prairie side? I basically made that list don't know if it's complete, but my view of it first thing: we need to provide fragment forwarding so, whether it's so secret source, you know you know node or we actually explained we there are. There- are pitfalls, I described them last time.

C

There are ways of doing this very wrong which appear to work until they don't so so those people have known some of them. At least we I documented a meager slides, a dresser proposed a way that everybody understand that everybody validates that does the fragment forwarding.

C

Well, okay, even if people have been doing that for 10 years in the hoenn implementation I'd like the ATF to come up and say there is way that we recognize work and we will do the trick for you and and yes, my draft sighs, you can do it with MPLS switching of the label. What did they tack on that? Pointing point is, if you fragment the first packet with, if, if you follow the first packet with the IP address, then you need to leave a state for the next fragment that you need to clean later.

C

Second thing we may or may not want to solve the second problem in the previous list, which was kind of this. This burst of fragments going on if we send this burst, like how just said we free memory on the sender, but we bloat the next hub and then the next stop which may get different births coming from different places, but the end of the day packet last fragment loss all the bad things.

C

Do we want to do some pacing windowing to avoid loading the network beyond its capacity to avoid congestion loss, same thing for well next step? Sorry fragment reliability. Do we want to be able to retry individually fragments which have been lost somewhere in the middle of this mesh right so exhausted, which rise in the middle of the mesh? What do we do and last but not least easy, and do we want to be able to SWAT or more or less efficiently, based on the current state of the network?

C

Do we want to provide some form of explicit congestion notification, so my my surgeon dropped, obviously as them all but clearly also people might think it's a huge overkill. So so that's that's where the discussion lies and then is maybe other solutions in mind, but at least my draft sighs all this. There is a way to do it next line.

C

So I just said the first line. This is a path forward, like Thomas said, which we've been there for so much time. I mean this promised in there. It.

A

C

A

The purchase a weakness, okay,.

F

C

And you're, a police you'll find my slides for last time with some drawings and stuff David. If you want to go through them, Oh.

A

J

D

I can go through some part of the slide sense. Let's see how we can handle it questions, so thank you pass go for elaborating the issues with route over so when the RFC 4944 is defined is called attitude are 15 . 14. It took fifteen that fall. It was mainly designed for mesh under forwarding, and this is mostly at the 2.5 layer packet forwarding with the 6lowpan adaptation layer and English nation header and the problems that, if you want, we use the same technique for route over.

D

You will have the problems that Pascal already mentioned. So the next few slides we are going to learn and I'm going to put together some of the questions and see what should we scope and what shouldn't be in scope. So we know that we have solved mesh under, but we want to ask the working group if the current definition for fragmentation forwarding fragmentation format under mesh under is sufficient for other 6lo technologies.

D

If there is an interest in the working group to revisit and reevaluate that, partly you know that some of the other technologies that we have specified in this group- many of them don't want to use fragmentation because there, because after emphasize or that's what they have specified and if there is any implementation in there to that, they may want to use some mesh, inter with fragmentation.

D

That is something they need and welcome, and that second question is doubt over: it looks like you know: Pascal has been promoting this presentation pollution for some time and today in the working group, we want to I think take a consensus whether we are going to work on this route over fragmentation solution, knowing that the current specification is not sufficient, so Debian do you want to.

A

Get on the on to the questions yo Carson. Do you have a clarification question or something very quick? So what.

U

I, do is go to make sure that we actually discuss things that exist, so no actually, the solution for the info troll works forward over so okay.

A

So let me ask the questions and for first we'll ask question about mesh under how many people here think that 49:44 and the collection of you know associated document. Thank you goes. Six, no model is fine for mesh under work or other than 15.4, or maybe even four freaking. Therefore, basically, do we need to do something extra for particular mesh unders.

A

Anybody interested in that things that it's okay, so I, see no nobody raising their their hand for route over how many people in the audience think that the working group should do something or the IDF whatever? Maybe it ends up somewhere else that the IDF should do something about improving fragmentation issues for our okay, raise your hand, please.

I

Okay, well, three.

A

Okay, so that's.

D

A

Was lit more than 20 21 22 ish definitely looks like more than half of the people right now hmm yeah. We will yeah both of these questions. We will report what we saw here, but we will confirm on the list, so you didn't get a chance to raise our your hand here. You can do that on the loose.

A

A

Well, the big question with the three seconds that we have right now is: it seems reasonable to think that there's going to be potentially some work on route over Carson disagrees with that. No.

U

I'm not disagree with it and we need to properly describe what we are doing yeah. That would be part of what we would do. I.

A

Think and then and then I you know actually do it so the what do you mean mother? You mean there.

U

Are three parts that are muddled together here? One is describe what people have been doing for a decade. That's probably a good thing to just write it up that.

A

May be a good thing, but this exercise is not about an informational document, no one. That is that's a great thing to do. That.

U

Goes to go to your auric requirement to get out of this moment, not thinking the second one is: do we have recommendations for implementations to do good timing in the presence of these optimizations?

U

It's probably also informational, because it just describes good practice all all to do things, but it could also be knowledge of I, don't know, and the third one is do we want to add a completely new thing, which is the fragment retransmission mechanism that goes over multiple hops within the six toe pad. That's a completely new thing and and one where we may be getting some transport pushback before we have described exactly what what we want to be.

A

I also saw other other new things like ecn, so there's different potential new things that and I guess that's the last question. A design team should go off and figure out and then for each one of those new things say yeah. We think we should do it and here's how I'm I don't think I mean it would be great to have some background info. But the main output of this effort, I, think here from the point of view of working group, is, if there's something to be done.

A

What is it was the normative text that we're getting we're gonna add into the collection of six love documents you know in history is great and we can use it, and maybe we can even reuse it. As is that would be great. We just need to solve the problem or problems so I think a lot of the questions you're asking a right, but that might be better for the design can to go after each one of those and that's something we can. We have to continue. Definitely we have to continue so.

A

A final question I want to ask is how many people here would be interested in forming a design team to work on this specifically.

A

The other thing I have is the people who are raising their the people are raising their hands. Please send email to the chairs so that we have the composition of that design, team and.

U

That way we can say we can finish on time, so the one thing that I would like that design team to do first is collect the research that we have available to actually that and if.

A

That would be a good input to the design team, but.

U

There may be other.

A

Opinions but definitely I think we have to close at this moment, we're officially at one minute are over and there's a social and people have to go grab it. So at this point we will confirm these things in the mailing list, but you we're done hey both blue sheets right here. Is there another blue sheet in the room somewhere.

B

D

You guys thank.

A