Internet Engineering Task Force IETF 115, 9 Nov 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF115-RTGWG-20221109-1500

Description

RTGWG meeting session at IETF115
2022/11/09 1500

https://datatracker.ietf.org/meeting/115/proceedings/

A

uh Please make sure you sign into the meeting, so we get a larger room next time.

B

uh Before you come to the mic, please raise your hand so.

B

Okay, welcome everyone to ATF 115. We are so happy to see you in person. It's going to be a busy meeting. Please be on time. Please respect time of others and let us open the meeting.

B

If you haven't read it before, please read not well by participating in ATF. You agree on everything that is defined in BCPS and text below.

B

Ipr disclosures we've got a routing working specific process. We require you to disclose APR before document get adopted by the working group. We require you to declare IPR before document can go in towards your plus call. Unless all authors and contributors have declared explicit their IPR document is not going to progress. Please make sure you do so.

B

We haven't had any adopted working group document between two atfs existing documents: the young crib extended. We closed working group loss calls and Shepherd right up has been uploaded, so we are waiting for 80 to review document. Bgp Peak uh commands yet to be addressed. Atn bgp, there's some reviews been done. It's a long document. We expect to progress at some point and there's also related. Routers are probably since six men TI LFA Stewart. Do you want to say a few words and thank you again for doing this.

C

All right, sir, well I apologize to the working group which has taken such a long time to get there, but had a very productive meeting with the author team before uh short earlier this afternoon and I believe we've got a way forward with the last remaining issues. So hopefully this will soon be with the isg.

B

Great and again, thank you so much for doing the work and I believe after the upcoming update. We are already to progress with style, faith and Qs model will be updated today, so agenda for today, I'm not going to read it uh again. We've got busy agenda, please be on time.

B

Agenda for tomorrow's meeting and we are going to have guests uh from sitcom. We are going to get the best paper award presentation and again busy day and we are ready for the presenter to come. uh Please do take a look at Wiki. We are start to populate data there and all additional information will be hosted in Wiki.

D

E

Okay, so you'll be moving the slides right: okay, okay, so hello, everyone, my uh name is Asim. Chaudhary uh I will be providing updates on qscn model.

E

So, in the uh latest version of the model, uh we have added one more module, uh which is Aina cost types uh which contains uh uh all the identities from the different cost modules.

E

So essentially, what we are doing is that we have taken the identities from the cost modules and moved to the Ina cost type module so that going forward, uh we don't have to change the cost modules and we can simply change the Ina cost type module uh which will be a much simpler process.

E

Also, in the uh last uh version of the model, uh we had uh included um operational model. So in this version of the model, we have added the action statement to clear the cost counters.

E

We have also added the security considerations. uh Also, uh there had been lot of uh young doctor comments on uh regarding the difference filters and how they operate or connect with each other. We have given an example in appendix T and in the next slide. I will be presenting or including uh how they are correlated and will be pretty clear from there uh that how the different filter correlates, with each other, uh to create classifiers and uh in the policy model. uh How the inline filters are included uh to create the policy.

E

uh So also, uh we have created a presence container uh for inline classification uh that has helped to include the uh to get rid of the line uh flag and has simplified the model for us.

E

uh In addition to that, uh we have taken care of numerous young doctor comments and uh based on those modification, we are able to come up with the cost model- zero. Nine next thing.

E

So if we look into this particular example, uh it's example for classifiers and the name of the classifier is my classifier, and this is a filter operation match all now. What does that mean?

E

That means all the filters uh need to be matching to the incoming packet uh for the classified to be classified for the packet to be classified under that classifier, and so when I say match all that means it will be matching all the filters and if we say match any, then any of the filter matching uh to the packet will be considered classified uh for the packet.

E

uh So in this particular example, uh if we uh we can see that we have included two filters. uh One is of the type dscp uh other is of the type Source port, and uh if we look into the dscp filter uh there is a logical not as true what does that mean. That means that whatever parameter, which is defined under the uh DHCP filter, uh it will be the knot of that. So, for example, in this particular example, if the dscp value is 11 to 13, that means and The Logical not is true.

E

That means 0 to 10 and the packet incoming packet if it has dscp value 0 to 10 or 14 to 63 that in that case, only it will be matching to that particular filter value. uh If the logical not is false the then it will be matching for the value 11 to 13.. Okay.

E

Similarly, for the source Port, we have multiple range values. uh In this particular example, we have Source put range from 10 000 to 10 300 and 17 500 to 90 800. That means incoming packet. uh If it has Source put range inside in these two particular ranges, then it will be classified to that particular filter.

E

So if we look overall, classifier example, which we have presented here, that means an incoming packet which is coming um has to match to all the filters which are defined here and the source Port value will must be in this particular range and for DHCP value. It will be the reverse of the range. uh Now there had been lot of confusion, a lot of young dot comments regarding this one.

E

uh We have not only provided the text to clarify this particular details about the classification, but also provided this particular example, which will help to clarify or to understand how the particular classifier behaves.

F

E

Similarly, this is policy configuration example, and in this particular example, the policy name is my policy, and um if we look further down it has an inline container. Now that inline container means that the configuration parameter has to be defined in line uh into the classifier itself. Now, if we look into the various vendor implementations, we will typically see that a classifier has been defined as a template or has been defined as inline. Now it also depend on customer requirements.

E

I have personally seen lot of customer who wants to segregate the transportation into a separate classifier, and that way they are able to segregate filters with the actions now to typically, the cost has much more complicated actions, and so they want to segregate filters with the action, and so they Define the classifier as a separate template, but at the same time, uh I have seen a lot of customer who wants to Define uh the classification parameter as part of the in line into the policy uh for them.

E

It's a simple mechanism to configuration of a Qs policy, so it may depend on a customer requirement number one number two also for vendors. Typically, they support one or two format or even both. uh But essentially, if you look into this example uh in this video example, the con the classification parameter defined in line uh for the DSP value, which is in the range of 21 to 22.

E

And if you look into the action, uh it's a marking down action from uh 2122 range to 23 range, to 23 value so and so for vendors to implement this. uh They may take this inline parameters and convert into their internal template implementation, or they can, if they have a template, configuration here instead of inline, and they are referring the classifier uh as a template. And if vendors don't support that, then they can take the classifier template parameters and convert into the inline uh in their internal implementation.

G

E

So uh essentially, people are hoping that we'll be uh able to cover all the young doctor comments by this ITF, but uh even despite we working very very hard on that, we still have some comments to be taken. Care of. uh We are very, very hopeful that by next ITF will be able to cover those.

H

That's it questions.

B

Okay, so thank you for the products really good work and we are getting closer to working with classical, definitely yes, yeah. Thank.

E

I

Foreign I'm, going to present a Sr V6 pass equals protection next page.

I

First, there was over document would like to thank AC, Peter, Bruno, eaming, Jeff, Grace, Alex, Kayton and Greek for their comment and suggestions.

I

One year ago uh we present this document and ask for working for the last call, but we are suggested to have LSR working group to review this document, so we presented this document in LSR intermitting and we get a comment there. In addition, AC also carefully review the document.

I

We've got available comments from AC, so in this current version we just all the comments received. uh Basically, we remove one optional subtr V under some editorial changes. So that's the main updates to this current version. So next page, so we would like to have ask for a working Math course. Any comments welcome. Thank you.

B

Any questions comments, tomatoes.

A

J

Go ahead um since the contents of this draft majority is LSR and spring related, uh and you said this has been reviewed in NSR. uh How about spring review? Has it been reviewed there because it defines a new uh behavior for the mirror set.

I

Give me for the spring review, so in fact uh yeah you're right, also I, think I. Remember you suggest we send that this uh document to Spring proof or review and as you suggested I, send this document to screen. Ask for for comments a couple years ago.

A

At least before the last call, we still need some feedbacks from Spring yeah.

A

Let's try again try.

I

Again: okay, I'll try again: okay yeah.

B

So, actually, there's a couple of documents that kind of share working groups or protection that definitely belongs here, while encapsulation and Technology someplace else and to okay, we I mean we will be talking to Spring specifically to get more reviews there and make sure that it can progress here.

I

Okay, I'll try again: okay.

A

So the the next presentation is also covering the same topic so.

I

Two uh two uh Eagle Productions, in fact a different one. So maybe I just answer a difference here. Just just answer some questions, maybe further. So.

A

I

K

H

Hello ciao: do you hear me.

A

Yes, I can hear you and I passed the control to you, so you can control the slide.

H

uh Okay, so uh hello, everyone from Channel mobile I will present the draft on behavior all the courses so.

H

It's a little bit slow right.

H

So ciao, could you help me control the.

A

At the bottom of your screen, you have yeah.

H

I see that, but when I change that how about it now it moves? Oh yeah, it's working! Okay, thank you. uh So the draft introduced a very simple uh multi-home protection uh for the i76 uh I, think it it's different from uh the draft. uh Why not just the present?

H

uh This is a uh on the VPN library to provide some protection solution, so maybe quite more will give more uh his understanding, yeah and the the basic uh idea is that we introduced to new item the first one is a backup seat which will be put on the top of the city list, and the second item is a new flavor. We call that Panama segment decapsulation liver, which indicated, as are the sixth parameter segment, the capital lesion processing.

H

uh The basic idea is that when the parameter and the point receives the packet and the fact that it's not the Hope derived the neighbor uh is the egress node and it's a unreachable, then the parameter in the point will direct the packet to the backup node based on the backup sheet.

H

uh So, uh let's see the encapsulation of the backup seat, it will be put on top of the seed list and the uh let's see the PSD flavor I. Think uh I I just have a message that it will indicate the parameter segment to remove the uh encapsulation and then for the packet.

H

To the uh primary note or the backup node, the flavor will be applied to some represented, such as the and Dot dt4 and Dot DTS, six and so on. uh We just needed some uh minus engine to the side. The side to code of the Ice Age processing, which defined the FC 8986, as shown in this page sub uh condition, will be added to s02 line.

H

So we also give an example, um so uh the basic mechanism at the Ingress node, uh it will insert the backup seat on top of the seed list for the pandemic endpoint.

H

H

At the normal condition, the traffic will be forwarded to the primary egress node. When the uh primary egress note receiver said with a PSD flavor, it will remove the encapsulation of the packet and forward the packet according to the payload.

H

Packet height and if the pandemic and the point the primary U.S node is unreachable, it will modify the uh destination address of the packet to the backup seat directly. Then the traffic will be forwarded to the backup node, so the multi-homing protection will be achieved.

H

So here is the example. We will ignore that and uh compared to the previous version is instructor. We introduce the PSD flavor uh to uh indicate uh the uh not made segment uh to remove the encapsulation directly and we also changes the position of the backup seat instead of the bottom of the seed list. In this version we put it on the top of the zip list.

H

So uh that's all in the any comments or questions.

L

Yeah Darren Dukes from Cisco Systems um I hadn't, seen the PSD Sid uh discussed in Spring at all. Have you have you discussed that there I might have missed it?

L

H

uh So this is a good question. Thank you, darling, I think we just the president has a draft in uh RTG working group at this moment, I think, if uh it's necessary, of course, it's needed. We should maybe present to the draft in the spring. Also.

L

Yeah sure, just and and also just uh I mean you could alert spring event at any time to send an email, uh because I think we should probably talk about the impact of that sit type thanks. Okay,.

H

I

Yeah, so I just want to explain the difference between the two equal protections, so this one. My answer is that this is a for VPN some kind of surface protection, so in order to provide this uh equals protection. So on the backup equation, node, we need a configure the backup seat so that backup seat sell the same space in the back cover note right and then the behavior of that state should be mirror or stem or similar to the seat on the on the primary equation. Node.

I

So under this condition at the English node, and then we can put another back up seed as the last seed in the city list and then we can achieve the equals the node protection.

I

So our approach is that so our you can you could production is a normal equals protection, so all the seeds on the primary equation node.

I

So we don't need a configuration on the bicar for the equation node, so we use a mirror sheet as a context and then all the behavior of that seed on the primary equals node. We just Barrel to the context node of a primer node, and then we use data table to forward uh for the the package to the destinations. So that means it's, so it's different from different approaches. Why is this focus on Surface and why is for normal protections.

H

Okay, thank you, one more uh I, I. Think in uh yes, the post draft is different. I think it solves a different problem, uh answers deep in the library. So thank you.

B

B

G

Can you hear me now.

B

G

I have a few comments on this work. First, it's laudable. um It is a a problem that needs to be solved, but there were a few impediments first, along with bringing it to Spring. You probably should bring it to six man because it changes the semantic of the SRH.

G

um Second, it does change the meaning of fields in the routing header. How do you intend to maintain backwards? Compatibility in in light of that.

H

uh Okay, so thank you, rob uh yeah, uh firstly, I I think I agree with you. We're not only needed to uh ask for the comments from the spring. We also need to do the same thing in the sixth, my working group uh and the the second question. I think uh the current uh solution uh is a compatible with the sih processing rules.

H

I, don't think we introduced some new, but anyway we should.

G

The semantics of segments left and last entry.

H

Oh uh yeah yeah correct. uh We introduced a new flavor for that I think it uh maybe the flavor uh can can be used at that.

H

Scenario right because the fever can indicate the different.

G

H

G

More question: how does this interact with the season.

H

G

Yes, one final question: how does this interact with Seaside.

G

Can you hear me.

G

Can you hear me.

B

No avail of the presenter will give an admin to reconnect and otherwise we'll take all the questions to the list. Okay and obviously, this needs to be discussed in six men and spring.

A

Let's do the next.

B

One yeah, so if your connect will try to to do this, if time allows hello Jazz, could you hear me yep so last question from Ron was: how does it interact with ccit, okay,.

H

I I just was lost due to the banned Network.

H

uh I did not hear the question from raw the last question. I did not hear.

B

The last question is: how does this interact with c-seed.

H

uh I think uh this moment uh you know the CC. The draft can support the mixed, uh comprise the seed and as a normal seed. uh You know this flavor just uh for the uh vaping seed, so when we keep the VPN seated as a normal seed, there's no any other uh problem, so I think it's com compatible with a cc truck.

B

Ron does it answer a question.

A

Not probably a better brain.

G

A

G

I think it might require a little more discussion.

B

Yeah I definitely have some discussion.

H

Kitten, maybe after the meeting we can't do more uh detailed discussion. Yeah, please occur all right.

J

I had actually a question on slide number six. If somebody could bring that up. Please.

C

J

This slide talks about uh yeah. This slide talks about penultimate uh node. Now, when you say penultimate here in this example, uh the SRH has A2 colon colon 1.. So did you mean the penultimate endpoint, because that could be several hops away or it could even be the Ingress P when there is no SRH right. So how does this work? uh How does this egress protection mechanism work when the penultimate segment endpoint is several hops away?.

H

uh So uh we have a uh the flavor uh PSD here right when the uh load, the P2, received a seat with.

J

So let me let me sorry, let me interrupt that assume that the SRH does not have a two colon colon 1 in this example.

J

H

J

Protection work in that case.

H

Sorry, let me see I.

J

I do assume that yeah yeah correct so assume that the penultimate uh segment endpoint was uh A1 colon colon.

H

J

P1 router is the penultimate. Yes, it can be right.

H

Yeah uh yeah, you are correct, yeah yeah, so thank you. We wish you the correct. Okay.

J

So a second, so please see or work out how this works. Second thing is: uh how does uh okay, let's say it is P2 and it detects that P1 node P3 node is down right uh now, P3 node is down. uh How does it know that that particular Behavior uh was an N dot uh DT uh with PSD, because pe3 is in the overlay, but P2 is a underlay node. How would it learn that the next seed is now this flavor and that it should do this?

J

You know funky Behavior, to uh you, know, Bypass or skip that hop and go to the next one with all the encapsulation and all of these things. So how would it know.

H

uh Yum, meaning uh how P2 no right, how.

J

How does how does P2 know about the overlay said, which is this n dot DT with PSD? How does it know about that number one, and how does it know that that PSD is the behavior associated with that said when it arrives at P2.

J

J

Yes, I I meant a failure, condition I meant a failure, condition.

J

Yeah, it could be, it could also be. It could also be a yeah, but it could also be true, but then.

J

J

Okay, so this is assuming this is. This is assuming that this.

J

Okay, I think in the interest of time we.

B

Are going in rounds and.

J

B

Please send your questions to the list and we'll take it from there.

H

Okay, so thank you. Thank you, chairs to all the authors.

B

Of the draft, please make it a priority to answer, because you actually didn't provide any answers.

B

I

I presented foreign, so I'm going to present a Sr physics, midpoint protection next page.

I

First of all, we thank all will give us a comments and suggestions, so I suggested we present this draft in the spring group and the way I will also the initiate the discussions in six month mainly needs. So basically, we focus on transun node as a Repel node, which cannot change a second loading header.

I

So originally, in a document we have a section to talk about a transient node as a Repel node. So in fact this is not in complying with the RFC 8200.

I

So we address that we remove this section trying to node as a repair node. You know in our draft and now we also address all the other comments.

I

So in the current uh version of the draft I think we are ready for adoption so because this one is also experimental. So next page, so we would like request for adoptions. So any comments.

B

Have you confirmed that your changes are acceptable for spring and six men.

I

Yeah I think so I think in the Middle East we, these are the trying to know, there's a repair node. So right now only endpoint node can change. Simran had basically changed the destination, so the those one violet.

A

All the packet is encrypt, uh is uh you have an extra layer of encapsulation or you.

I

A

Sri header is now clear: no.

I

Not to change the SR SRH header just change. The second left this one right now, because we have a segment list so when one node fails so the under pointer, node, abstract and anywhere so in private hope, and then that node nodes or this one failed, not reachable. We just pop up that one, the document that the segment left and then we go to the left next one right. So there's a very simple change: there's no, no is not finite. uh Rfc 8200 is simple change. It's almost the same as uh same routing.

I

Behavior is same as a second round behaviors, so trans just now in at the end point originally, we have also we have a protection on the under Point node, also transfer node, so just try to notify that the RFC 8200. So we will remove that one.

A

If you only decrement the point of what, if there are it's gonna cause loops.

A

Only have the next half right, you actually skip the failed one. You go to the next one directly, but you cannot uh guarantee. There is no Loop.

I

Well, that one yeah that's a good question, so in fact that's the after IDP convergence, after igb conversion conversion, so we use a DOTA. We use the endpoint under Point seats, so we go go along the shorter path of igp after IDP convergence right. So that's the solution. So the student have a loop because I have like igb convergence that it works. A.

B

Gp converts is slow, there's significant Delta in time where you have failed linked to the next hub, but your igp hasn't converged yet right.

I

So uh basically, this solution have two two stages.

I

One stage is that before IDP convergence, so before igp converges and then the package will deliver to the adjacent node of data fail node right, because why not failed igp is not converging. So the 14 table on Evernote still think that node is not failed right. Only the Json node node. That's not failed because it has a PFD detection, whatever Fast Direct detections, so that adjacent node will do TI, LFA, whatever frr, so that data education, node fastfully, detect the failure of that guy and then get around that failed note such as before igb convergence.

I

So after the IDP converge then Evernote knows or that guy is died, so the remote node will know. I cannot reach that failed node. So in this case this second stage, the endpoint node will pop up at a further node, see the seat of that fair, node using and then the last seat.

I

That's opposite of along the along the Sr pass, so that one right now is converged shortest path. And then you know there's no dope right so at a certain stage. So.

B

Your assuming field be preempting, uh tlfa and local, not by changing the list on not before.

I

Not changing the list just pop yeah.

B

I understand yeah.

L

Yeah I couldn't put my hand up on the tool because the Network's not working in here for me right now, but during Duke Cisco Systems um I, don't think that particular change was discussed on any mailing list, because, if you say you're, just gonna ignore the next segment and decrement segment list without carrying what it was you're just going to break it. You can't do that.

I

I think there's a we send it to the sixth amendment list and then we have a discussion recorder. I can send to you.

L

Yeah that that aspect, okay.

M

I

Send it out very clearly and I'll requests of this document that collect a lot of a record of a wedding list. Yeah.

B

So uh you, the update, is a draft and you assume that the update has addressed the commands right, but you actually didn't get any confirmations that people are happy with new changes right. So I would really advise you to send another email to spring with the updated version of the draft and explicitly ask whether people are okay with it.

I

Okay and then we can talk about adoption. Okay, thank you again, okay, to see whether further comments, okay, thank you.

A

Then you won't ask no question.

J

Again, it's kind of the same first question that I asked previously. uh There seems to be an assumption that uh with srv6 that the penultimate or the previous segment uh endpoint node is directly connected to the next segment endpoint node, so that it can detect. You know a failure as a in a tlfa or a local thing.

J

This is probably not going to be all always the case, so uh I mean I I'm, really wondering how uh practical or useful this is, or if it is only applicable for a hub by hop kind of uh path. Right.

I

So we're cut into your question, so this is not assume the the note is I mean the endpoint. Node is a directly connected to fill the note. So this is a draft.

I

Is a law node and the pointer node is the upper string of the filter. Node can be multiple by way it can be. The endpoint node directly connects the fair node. That's okay! It also allow the endpoint node, which is the upper string of the field node of multiple away. Both case is covered.

B

So your assumption is that every not participation, igp will flood update information and denote. That is the previous Hub in the list will react accordingly. Yeah.

I

The difference, so is the case whether in case of the node and the pointer, node Direct Connect, it's just a fair note. Also the Enterprise node is away to the fair note. Those are work because the independent node after igp convergence knows that I cannot reach the further node right, because that note is done. I did be converged after this, and the pointer note notes that I cannot reach the third note. I will put half the seat for that fail to note and then the likes.

I

The next seat will be the first at the top of the stem list right and then we'll forward the traffic to the last seat using sorted paths after IDP converged. So it's okay right, whatever.

I

A

As Darren suggested, this Behavior needs to be discussed in Spring yeah. We.

I

Can discussion no problem? We can discuss the whatever issue you can come up or reach. You can resolve.

J

It just a short follow-up form or whatever you discussed. It would be good if you would put it in the draft before you send it out for him. So this way people have the right expectation on what kind of uh how fast the uh protection mechanism works.

I

Yeah I think this idea also will present in the NASA ITF. We have a picture there and then these two cases that I present there are you know because that's it will print in the spring and then after that prison here somewhere, as your question can present again in the spring. Whatever the more comments, that's okay,.

B

So you have updated the draft and you significantly changed the logic right. No.

I

No, no, not significant just remove this transgender node as a Repel node.

B

Which is significant change so again without going around? Please do address the comments. Please uh have a discussion spring contestment.

I

B

B

Robin is going to present in three months all right.

K

Hello, hello, everyone. This is.

K

The generalized IPv6 tunnel, okay, next, so while we need the Gip uh generalize, the IPv6 tunnel, because countries there are many types of internals such as you can see the list in the labsible how to tt3, like yourself, Electro and the on IPv6 Network, it's hard to Define extensions or for all the all these tunnels to support the new features.

K

It's not recommended to attention new features, but it's the ipv4. This plan for the tunnels, new features such as the API and the alternate mapping and our ammu okay.

K

Next, so what's the challenge need to be supporting new features such as the last the previous before, and so the following problem will be existing first, there is a lot of standard worker I'll, add all of the IP tunnels mentioned about mutualism, accordingly, resulting a lot of standard work, and the second is hard to keep the consistent because between the IPv6 and ipv4 for this opportunity, since the possible extensions are recommended to be only on the IPv6, so you can't um it kind of keeps the consistency, and so there is a function.

K

Reduction ibv6 can directly support some functions of these articles, which can't be done with a pv4, and this means some function becomes reduction over the rtv6. For example, we we, as line, takes use of the UDT to do the USM key, but.

K

In the IPv6, header also can be used to support the same key and, of course, the zip code to extension, based on the existing format, some actually tunnels, such as have their own headers. If these tunnels need to support new features over the ip6 they will, they will face the challenge of the choice between reducing the IP resistance, IPv6 information for this in future, based on the IPv6 extension header and Define, a new extension based on their own header.

K

So they have two problems. The first is: if the tonal header extended, it will be returning with the execution IPv6 in Evolution, for the new feature is on the ipv63 header and the second is for some existing funnels as a tune to have their own other, so they have to use the IPv6 in calculation for these new features based on IPv6 header, so extensions need to be redefined in the IPv6 extension header. As a result, the extension May differ from that of the IP tunnels, which have their own header. Okay. Next.

K

So we Define, we defined a new tunnels called the generalized IPv6 tonal, and the tunnel is defined to use the IPv6 header and rb6 extension handle to sort both acquisition applications function and the new features under the cavaliation completely. The party has the following format: okay, with any V6 header and ipv16 extensions.

K

Next is the pillow the option below the packaging. This is option one and the next option two is the IPv6 header and IPv6 extension is.

K

Okay, that's what.

K

The function of the UDP is replaced by the flow label of the IPv6 header in the grp6 tunnel to ensure capability. The value of the flow level cap calculated for the powers of zmp should be the same as that of the source Port of UDP until the definition of the main option is defined, and this includes the big Baseline had the information and the option must only encapsulate in the destination option. Header.

K

Okay for people, it's also an example, and we Define a new Peak header and the quicker also to Define, to cover the quick credit information, the quick option, and also you uh also include in capitalizes, in the destination options: header, okay,.

K

Okay and uh for I'm here, so we just use, give a use case used for the ipv jp6 for MPS next.

K

And what's the challenge is is helpful challenges first, the MPS is the lack of the source and indication and the MP 2p connection may occur, so this will cause the difficult and the compliance when process is also SFL is defined, but there is a few in position and the second is the pillow type can't be directly and determined, because there is no pillow, the indication and third there's no medita extensibility and distributed to encapsulate the new sodium attributes for the new future, such as a feature, selection and APN Etc unfold is the process of it.

K

It's a p function is complex, complex and the vasopodium performance. Intro pay level is used for the example at the bottom of the legal stance. So we information how to be passed for the purpose of the SMP and that's so when you use the grp6 tunnel for an foreign we can have the volume the first is. Social address can be used to form a source identification and the next is the IPv6. Network header can increase the payload time and third is the IPv6 flow labels?

K

Can you can be used to implementing the gsmp and, of course, the encapsulation for the new future defined well in IPv6 and can be used? Okay, that's what is the format? Okay? Next?

K

Okay, so we have two type of Ms idea, and here is the faster and faster. We have two actions one is to pop and one swap it would pop up with just the shift. The uh the picture open that underfo the tab. Two I'm here uh see them and we use the IPv6 using Hydra and so use the Lego stack. Also, we Define the procedure. It depends to home house multiple labels in composition, the phone this is also the same as and the type 2 is used as a segment as a June header.

K

After all, the label in pollution in the IPv6 definition address are housing, so maybe it's the first label interpolation in the second indicates uh SL of the IH will be processed. Something else so uh also have some control plan consideration. The first is exiting uh I'm. Health control plan doesn't need to be changed, so I'm has a label on the control plan can still be destroyed for IPv6 I. Believe four and R2 next.

K

Okay, so let's use some requirements of the grpv6 terminal and Suppose. There had two problems: the con, firstly, is the country, many new features, I imagine and the cost between interpolation over the IPv6 and has defined, and the second is in the process of deployment of this new future, because Network demands have different capabilities of Apex detention, header, so the following issue: maybe okay? Maybe you can you can check this okay?

K

So we have some some new comments. The first is the way to advertise the community and maybe uh the uh two different ways. One is use the RDP sphere and the next is use the controller so PCS and the next time young may be used and the interest we mean. So the paths can, of course most to me. So we we this time we can use a TPU to support this.

K

You know it's Neo and the TCP or if and that's exactly, okay and the company about the ability is essentially and the community option of the header. Also, we need to send it and the the last thing is the completed about the specific future and like a solution out or alternate marking is okay.

K

Okay, that's all uh comments are welcome. Thank you.

N

Greg um Greg musky Erickson. um Well, thank you for sharing your imagination. um I have a question. So um are you aware of the work on mpls network actions to uh extend uh mpls architecture.

K

And you mean the extension I mean us.

N

It's not extension, it's called mpls Network actions.

K

I I'm not I'm, not on not uh don't just do.

N

That okay I'll encourage you uh to check uh the work of mpls files and that uh user uh working groups and the documents they adopted already and not to duplicate it elsewhere. Yeah.

K

But I think I think that that is not conflict.

N

uh Since you are not aware of this work, I don't think that you can really uh compare your proposals and what the community is working on for some time. Okay, so uh I, don't think that what you are proposing uh really is uh required. Thank you, foreign.

L

Duke Cisco Systems um this one's, maybe maybe more- for the working group chairs this. uh This draft is defining a bunch of new tunnel encapsulations. Is this the right location for this type of work.

B

Yeah, so there's a set of requirements and usually requirements are there because someone requires them I'm yet to hear from NVR people. They are unhappy with the transports layer used for uh loud sharing or load balancing it's. You know there are billions of endpoints during the summer ip6 today and they're using UDP, so the cost of implementation. What you're proposing is enormous and benefits are absolutely clear. Whole use of flow level for load sharing, is, you know, toxic topic, so in general, I mean I understand what you're trying to do.

B

I'm, not sure industry needs it, and, and you definitely need to go to all the stakeholders start with mpls and nvo and tell them what you are doing and here's their reaction to it. If they don't need it, maybe you shouldn't be doing it.

K

We can discuss later.

B

I think there's a what.

A

B

O

The long road to the mic, AC land of Cisco, Systems I, think this is interesting- that you've taken like a green field and mapped all these existing uh end caps to IPv6 around hampers, and it was a nice it's a nice exercise. It would be a nice like research paper, but given how long all these things have evolved and been deployed and where they are, it's something I mean it'll, never see the light of day. I can tell you that right now, so you're just spending a lot of time, especially with the scope.

O

You know it might be. If, if you hadn't I'm, not saying you do, but if you had an end cap that tried to replace those that would be one rather than mapping every one of them and you still have all the different cases, because you really haven't gotten rid of you know you haven't, got rid of uh genev or mpls or anything else. You just mapped them and it's you know like I, said well, it's kind of nice to do a green field. It's not it's just never going to happen.

K

Okay, I think the advantage for this methadone- we maybe the even we- have also have money harder to do the harder. The format may be at the same format.

B

The three phrase is see you know if you would have focused on completely new use cases. Maybe there's a value, but looking at what you're trying to do is just ocean boiling exercise.

B

Okay, well done right, yeah! Thank you.

B

A

D

So if AC thinks the previous draft will never see the light of day, he's gonna love this one.

D

So um this trap is a little bit uh unique in that we're not proposing anything we're simply giving a summary of what we've seen being proposed out in the industry, the authors we were involved in the iic and a little bit of some of the IEEE groups that are working on blockchain with regards to iot and governance, and so- um and we presented here previously on the impact of dlts on the network and some measurements and testing we've done.

D

So this is kind of just an evolution that we've had it's probably too experimental and forward-thinking for IDR, Sue and um insiders. So we're just going to start here and see and see where it goes, and we probably could have changed. We could probably could have called it um append only ledgers for bgp, but blockchain's, much more provocative and gets people upset. So we're going to just uh go for this.

D

um Just in one slide. I want to just cut to the chase, to see it, tell you what we've found and then I'm going to meander a bit and then we'll come back to it.

D

So just want to make sure that everybody's clear as possible on ways that something like a blockchain could be used for a routing protocol like bgp, although it could be used for probably any writing protocol, um so a DCS, a distributed consensus system that uses something like blockchain, it doesn't have to be blockchain again, but um a lot of us are getting more familiar with that could be used out of bands. Something like rpki does for bgp to to supplement existing bgp management, but perhaps by using smart contracts.

D

So smart contracts are if then, programs that are stored on blockchains to run uh when predetermined conditions are met so like. If Bob signs a document and that's sent to the blockchain, then the smart contract would initiate payment or a transfer of a particular asset, so automate automates those those agreements.

D

um So a bgp related smart contract could be executed immediately when some conditions, such as receiving an update with too many pre-pens or or some sort of a hijacking, is detected, and so the action could be to deny an update after Consulting a blockchain that you've created.

D

It could be something as simple as using blockchain to securely store, config files or row, as maybe the irrs could use blockchain to store these and before they distribute the the row as um and we we dealt with that in in the draft as well. Tools do exist to do this today, but a blockchain does have potential advantages in that it's transparent, it's immutable, it's secure, it could be used globally um or within the domain. So that's kind of cutting the chase that just kind of gives you overview.

D

Let me Meander just for a little bit next leg.

D

So there's as I mentioned a variety proposals that have been happening for those of you that may just despise talking about blockchain and the ITF, it is starting to infiltrate the IHF. There are some potential working groups, um one involving a secure asset transfer between blockchains being able to provide interoperability between blockchains.

D

It very well may become a working group, I'm, not sure if that's happened or if that will happen this week and supply chain, Integrity, transparency and Trust, which is also based upon a penned only Ledger.

D

So there are proposals happening, we've received, as I said, questions about how DLT in networking could be handled, and so we've we've been discussing that in some other forums we did hold a side meeting to discuss this a couple High tips ago. This ITF there was a networking in metaverse side meeting uh and blockchain and metaverse work hand in hand. So you can know who owns what in some sort of a metaverse.

D

uh This is just an informational draft. It's not an endorsement and we're not again asking for anything, except for advice as to whether this is something that we should be working on in this working group or elsewhere, and it ends it asks the question: is it possible to use a distributed consensus system like blockchain um to manage bgp? And the answer is yes, whether it's a good idea or not. That's another question next slide.

D

um So, just let me just fly through these, these ones, real, quick! um If you don't, if you're not familiar with the LTS just then you just kind of get it on your own, but um cryptocurrencies and DLCs, they don't care much about the underlying provider Network. They just assume that it works. They have their own peer-to-peer networking.

D

um You know with a pool of Transport layer connections and they've done a pretty good job of securing their application next slide uh there. This is what basically, their view of it is is that they have this peer-to-peer network with many nodes. Sometimes it's millions of nodes in the case of the global blockchain that are very verifying transactions and executing smart contracts and things, and it just runs over this IP network next slide.

D

This is the architecture that, within the IEEE anyway, that we've kind of been using in most of the uh the groups where you have your session layer, whether transactions and blocks are being created um above that is a consensus layer where you're either either using proof of work or proof of stake to uh create these blocks and the transactions that are in these blocks and then the contract, layer and um application protocol layer in the application layer. Next slide.

D

um On my flight over here, I, listen to a podcast, which I think is one of the best podcasts out there. My only argument is uh that they should have more of them, but it's uh Leslie daigle's um consequences, and this last one that they had, which was on uh um cyber criminal gangs. At the end of it they start talking about using blockchain, and so my ears picked up about how Maybe um you know. Criminals have their own ecosystem.

D

It's gotten very professional, even accepted and blockchain could potentially be used as a base for the good guys ecosystem to help counteract that, because blockchain helped show proof of where criminal activity is occurring.

D

um It'll make the bad guys expend more effort than perhaps intelligence gained, and the whole idea of blockchain is to make it publicly visible, and perhaps we can use that to our advantage looks like you could use regular database like a SQL or nosql database to do these kind of things. But these are some of the potential drawbacks that they are controlled by administrator. They are a client server in nature that malicious actors can more easily alter the data than you could on a blockchain um administrator decides, which date is accessible and visible.

D

The good thing about them is that they're easy to implement and maintain it very fast. Next slide.

D

um There's a lot of opportunities outside of using for bgp, including trusting capture uh packet datas. If you're capturing data from your network, you can maybe store that on a blockchain, because you may trust that more and a variety of other opportunities as well. um Dino has worked on this with regards to lisp. He worked with the Nexus community and um he used the lisp to help DLT applications Define the best nodes.

D

Okay, uh I'm gonna um skip this one go to the very last one. So how about just look at that last that diagram right there um and I'll stand on this. It's a kind of very similar to rpki. You could have bgp clients that are also utilizing a distributed consensus system which may have a bgp block table in there, including smart contracts that they look to before they uh act on certain things. It's kind of similar to using rpki. It could be replacement for rppi or an addendum to it.

D

um That's it good time, correct questions or.

B

Yeah David.

M

I'm putter, um so a fundamental aspect of blockchains is that you have proof of work in some form, which can be proof of CPU usage, proof of storage, proof of stake um the way or the actually only way I can see this being useful is, if you have proof of network ownership, of something where that something could be addresses or an ASM. Are you actively looking into making that the base of the blockchain, or is that? Are you still too early for this or.

D

We're not actively really looking at anything, except for what I just described so but very good point.

P

Jeff has uh hi to Eric here: user bgp, a protocol advice says, is very slow to start with, and blockchain is not noted to be fast. True, um so for your use cases having this as part of the protocol, control loops is probably not a great thing. You know, I would suggest just simply taking those use cases throwing them into the fire, and you know stay warm by that fair.

D

P

But bgp also is about policy and policies about money so places where you know policy can actually, you know, take things as input to know whether it's you know, people agreeing to you know announce this route after you've paid your bill and your bills registered in a blockchain. You know something like that. That's a reasonable thing. People wanting to actually have their address delegation contracts, no move through a blockchain. Those way of saying that they're transferring addresses so people like Heron and ripe at other places, obviously have their own systems.

P

Maybe they're good targets for you to have this conversation, but uh mostly look at this as use cases not as a transfer of centralized things that have to go through some sort of delegated infrastructure. Pki makes sense for a lot of what pgp does for the things that are more provided to provider type relationships. That's not a bad thing. Thanks.

B

Dean one minute question not more.

F

Be quick, okay, I'll be very quick, um so um I'm just going to respond to your point. uh He didn't say he was going to use bgp and have the rib B in blockchain, because thank you for not doing that because it is, but it could be used for administrative things and I was going to ask. The same question is if this would could be a good thing to prove ownership of address prefix that have been allocated by the Registries. That would have some really good use yeah. Thank you all right.

B

Mike, so please do think what you're going to do with this work. uh I mean we are happily we're happy to give you an opportunity to introduce the world, but practically you really need to identify your focus and start putting more meat on the buns. That's otherwise.

D

Very good, if anybody's interested, please contact me and we'll go thanks. Thank.

B

Q

Hello, do you hear me yeah, it's okay, okay, so welcome to the cyan update presentation uh in this presentation. We are going to tell you. Oh sorry, okay, we are going to tell you what we have done since we were here at the ietf for the first time in that was at the ietf 113 in.

G

Q

So there we there were two events where we uh presented sign. There was a side meeting about sign where we gave introduction, and we asked you for input, uh suggestions about uh design, ietf strategy and I was also a sign presentation, a very high level overview of the routing area open meeting, and then we got a lot of feedback. We realized there is also interest, but people told us sign is really a very big system and the ietf is not really good at the existence but more at small components.

Q

So that's uh we got the advice to try to break sign down in smaller junks and that's what we started to do or try to do next slide, please. So what did we do so far? uh We have written and published three internet rest. uh One is a sign overview drafts uh here we try to explain, sign a bit more in detail and also a bit more per uh core components uh than the sine component.

Q

Analysis uh is also about the core components, but then more how they've interact with each other, how much they are, depending on each other and also how much they use existing protocols extended, reuse, it or maybe not use it, uh and then the third draft is about to sign control, pane a pki where we already in detail Uh Kevin specification about one of the sign core components, and we presented these drafts in several energy entry meetings.

Q

uh As a base for discussion, we chose penergy because we do not have a home yet at ietf and the Passover Network and research group seemed for us the most close, but we also want to share our work with you with the rooting Area Community. So that's also why I'm here next slide and now I want to very shortly give you an idea what these three address are about. This sign overview draft discusses the motivations behind the design architecture.

Q

It gives a high level overview of the fundamental components and it concludes with an old view of current sign deployments, because it is not only a salary but really already in use here is the link to the draft. If you want to read it for yourself next slide, so very short sign in a nutshell: a sign is based a part based interdomain Network architecture, The Three, core components are the control, pane PPI, which is working on responsible for authentication and which builds the base for the unique sign trust model.

Q

And then we have the control plane component, which is doing the routing in that it's constructs and the test image path, segments and it authenticates path, information and then, in the data playing the password forwarding happens, uh and what here uh takes place is that here these path segments are combined into paths.

Q

This pass information is put in the header of the packets and the packet forwarding is done based on these past information, so the the forwarding is is based on the information in a packet and not the routing tables, and that means that you use that sign, uses simple routers and a stateless operation.

Q

Next slide, please, uh as I already told you, it is really already in new sign. There are several service providers in Switzerland, Europe and also in Asia, which provides iron, and this is growing.

Q

There is an internet exchange, Point C6, which offers signed peerings, and we also there are also customers working with sign, especially the Swiss financial institutions, the secure stress, Financial Network, where the Swiss National Bank, and also this risk clearing Institute and the big bigger space banks are working together, sharing their data based on sign, also education government, as well as a healthcare Network, then also research networks having a connect based on sign, and we have the sign lab. This is more like a research test bet where we can try out new features.

Q

So design component analysis is really analyzing. This core components from a functionality perspective. It describes how that depends, and it interacts with each other outputs properties provided, and it tries to answer the questions. What are the main components of cyan? Can they be used independently and that what existing Protocols are used? Reuse extended or maybe why are they not used? And it briefly touches on the maturity level of components and extensions? If you are interested in this draft here is the link? uh Okay very shortly. Now the this dependency graph.

Q

uh On top, we have this control plane, uh public key infrastructure, which is responsible for the authentication. This gets us input, uh bootstrapping information, uh among others, uh some kind of trustful contracts and lose time sync and as the outputs uh it it delivers a sign trust model which I will shortly describe later and authenticated control messages. This is the input for the control plane for the routing part and then as an outcome. You have this authenticated path segments which the data plane uses to construct.

Q

These paths which it puts in the past header and the the outcome, then in the end, is secure into domain. Multi-Pass communication next slide. Please so, and then our next draft is uh more like really a specification of this control pen, PK I visit, which we think is the most independent part uh core component of sign. uh It presents the trust, concept and design of this control plane pki. It provides a short overview of the certificates, keys and the roles involved.

Q

It gives also detailed specifications of these building blocks and it concludes with several considerations about deploying it here. I can link to the craft. Can you please send a comment so in here is this trust model very shortly? It is based on isolation, domains isds. This consists of a logical grouping of an automa systems that share a uniform trust environment like, for instance, a common jurisdiction.

Q

Each isolation domain is administered by several core as they compile the ISD core, and this ISD core negotiates this trust policy or contract, which we call the trust, root, configuration so and then the certification authorities in an ISD that can only create certificates for the autonomous systems in their own isolation domain. So you see also the picture. These are three isolation domains. They have the core core ISD course, and each isolation domain has also his own trust food Constructor configuration with the next one please so that was in a very short, fast tempo.

Q

uh What we have done so far uh that we are doing uh now is uh to address the received feedback, and we still welcome very much reviews and commands from you uh based on this feedback. We want to improve the existing draft because they are really the very first versions of these drafts and there is a lot of room for improvements.

Q

Then we also want to write specifications for the other, two core components, the control plane and the data plane and in the meantime, we keep coming to the iatf, share our progress and source with you and we try to build up uh yes, support and a community um and yeah. That was that was it. Thank you for your attention and if there are questions or any feedback, please welcome.

G

B

Thank you for being here and uh sharing what's going on with sun, so I believe you need to focus on those things. One is service specification, because if someone else is interested in building systems that can interrupt with your system, you need to clear specify what you are doing. An idea from the right place to do so. Number two is potentially interworking between common internet run by bgps and and immigration scenarios. Tunneling scenarios how you authenticate non-san demand software yeah and thank you. Okay,.

A

B

Thanks everyone, we are giving you three minutes back and we'll see you tomorrow.