Internet Engineering Task Force 104, 28 Mar 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF104-PCE-20190328-0900

Description

PCE meeting session at IETF104
2019/03/28 0900

https://datatracker.ietf.org/meeting/104/proceedings/

A

B

That side of the building is for the is G.

B

You should stretch out on CT sleep.

C

Good morning, this is the PCU session.

C

Okay, let's start but the usual not well, remember that the IGF has IPA rules. Please buy them. Forget that anything say right here is recorded and involves C's, as you seen as a contribution from the 80s perspective.

C

C

It should be going through, so you know process. We are through a lot here, so we try to have a look at job or participant questions. What you have to remote presentation so work as expected. If you come to the mic or to the presenter mic, please make sure to use it correctly and we have pink cross over here down make sure. But you know even the cross when you present for the video to be recorded properly.

C

So, as you noticed, you John stepped down from his chair position. So this is the opportunity for me to thank him again. I wasn't there last time, but I already told him directly. We hope to see him again. Vhf, maybe not as rigorous, is a before. Oh, we have now two rookies who just on me.

C

First, one on my right, his drive I think that if you put on a newcomer tie on your badge, you certainly know already so.

C

Is Adrian but it needs quite fun to maybe someone who's been the motorist.

C

Excited to serve as a soldier in to help to to solve to two topics. One of them is the transitioning for drove who is involved or used to be involved on many documents in the group. So to is this transitioning in a step by step basis, Adrian will be able to to act on ensure that the load is properly balanced because before he leaves for for drove and second goal, help flushing some of the cues we have for the PC document.

C

So we hope that this will progress quite efficiently and the in term is to see everyone stepping down from his position before the next ATF meeting. Montreal.

C

As you may be aware, there is an Asian responded by ETF. Quite recently, it involves the world so working group from the routing area and ETF. It was a response to Qt communication, which came a while ago.

C

We roughly just include some of the recent PC document there to make sure the AG, which is aware of the walkway with progress. A shorty minor on using the main list, will have again this time, multiple draft that came out without any discussion on the list just sometimes and not each time, announcements and the mailing list. Please remember that I first want to progress. The work is the mailing list.

C

Don't wait for a meeting to discuss draft issues and try to send message as much as possible on the list among the new things we've got the wiki, so I can want by Mexico shell. For that too, one of them suggested the other one make it real. So there is a wiki page that has been, which was there is been greatly improved to include all the tombs and all the commands we manage.

C

The good thing is that anyone can update it. So if you have a document that which significant milestone pass something and that would share- are late in updating the wiki, you might do it yourself so try to make the best use of it on I think we have good tool to start with on this wiki.

C

You may be aware also that in the routing area there is an implementation policy discussion. We have to decide how we will manage that, for the PC working group will certainly start a discussion on implementation policy over the mailing list or in the coming weeks. We have to talk all together on decide what would be the reasonable way to end an implementation policy for the PC working group, yet gender anyone wants to bash something look standing over the road, nothing to do. Okay,.

C

um Let's start.

D

The status of all our documents, so these are the documents that are beyond the working group. We have no new RFC's since Bangkok, but nothing to despair. You can see there is a lot coming. Your way, we have two documents which are in RFC editor queue. Both are pending a mr. f1 from a document from spring and another document from PC itself, which is the gmpls and that's the next document in the drafts with is G, which is on the telly chat next week, which is the gmpls piece of extension.

D

So who is happy to see? Finally, this going through, apart from that, we shipped various documents to is G in the past one past months, there's they say there is stateful P to MP her I key extension, applicability of PC and a CTN, a very important document, Association group, which is being used, and they are a bunch of other documents which are built up on the generic Association group and yesterday, with the help of John.

D

We also shipped the inter area ace applicability document, so bunch of documents have been sent to the as she so also request. Authors to when the document goes to the is JB, please responsive when the comments are posted and let's try and resolve those comments from the direct rate as well as from the from our 80s quickly.

D

These are the early code pointer locations which have been done in the past. Luckily, all these documents have been sent to the SG, so there is nothing to worry and most of the things will be taken care of matically. This errata that you see is is been pending from the last last idea of meeting, but finally, we closed it. It was raised technical, but after discussion we decided that's actually an editorial and we fixed the our BNF issues, and this errata is now closed and verified.

D

Now moving to working group documents, which are near last call, we have one document which is waiting, Shepard report. This is the auto bandwidth document. We had a change in the personnel for Shepherd. That's the reason for this to be pending, but we plan to act on it pretty soon. There is an ongoing last call for stateful HPC, which will end next week. So this is also an opportunity that, if you care about this, please respond.

D

We have not seen much response for this on the list yet so please do that, and- and this will also help us chairs- to make sure that yes, there is enough consensus in moving this work forward.

D

What you see is the at the bottom is the working group glass called Q, so this is the Q in which we will start progressing the next set of documents. Two of them are related to Association protection and diversity, and one is a simple stateful extension which allows PC to take control over LSP, so these has been discussed in the past extensively and they are almost ready. So this is the Q for your information and you can find all that information in wiki as well. So please keep checking to know how your work is progressing.

D

Okay, so these are our rest of hard-working group documents. This time we asked everyone to update the status in the wiki, most mostly I made the update and I think everybody was happy with that update because nobody came and shouted at me. So this is what I have put on the slides as well. If you have disagree, you still have an opportunity to come up to the mic and correct the status. So we have an association policy document. There is no update and there is no pending actions from the author's as well.

D

There was an update made for Pisa flowspec, based on the dis between the last meeting and now. The main thing that we would require here is: we need to gather more implementation experience for this document, so if you are, if you have- or if you are planning to implement this, please reach out to chairs provide some information on the implementation part. This is a very important piece of work, so we would like a little implementation feedback as well next document, which is the piece F yang. We send this document for only yang dr. review.

D

My ish provided those reviews during the week, I posted the updated version, I'm still the editor for this document, and still back and forth. You would have seen their messages going on to resolve this quickly, but this document imputs lot of other documents which are still not RFC so anyway, we would have to wait for those document to progress before we can take piece F yang to RFC publication final. We have bidirectional.

D

This document was also updated, based on the comments that were received on the mailing list, basically mainly from marina, and this document is also progressing quite well. Next, we have set of documents which are on the agenda today. The one that I wanted to highlight was the stateful gmpls. We made a request to the authors of stateful, gmpls and remote initiated gmpls.

D

If these two documents can be merged, and luckily we reached a conclusion and that will be presented as well in the so I will skip that for now and other documents, the three that you see on the top. They are all on the agenda, so let's get them and then we have documents which are not on the agenda, which is PCL, SP scheduling again. This document is progressing well, but needs a little bit more implementation feedback. So please provide that. We have two recent documents that we adopted since the last meeting.

D

One is the PC flex grid and all the comments that we received during the adoption call those have also been handled. So that's very good to see that the document already has a zero one version which handles those comment, and we have SRV six document that we are adopted. We are still awaiting a zero one version that handles the comments that were raised during the adoption call.

D

This is the pull queue adoption pole, Q that the chairs believe that, based on the last couple of meetings that there is, can there is enough information for to us that we can start the adoption poll. So we have the VN association, pc, c c sr and the binding said which is in your adoption poll Q. So please look out for these documents. If you this is also good. Don't wait for adoption poll if you have comments on these document reach out now, let's fix them and resolve them quickly and that will be the status.

D

So if somebody has a disagreement on any of the working group status or want to add some information, this is your time going once going twice and thank.

E

C

While we try to run the slides as inyoung your first presenter.

F

Good morning, everyone I'm young.

F

I think as true explain, this one is actually a merger of two draft.

F

F

Pizza stay for PC gmpls extend RFC 82 31 Adrienne donor is no laser in.

F

And this this draft is basically extension to cover stay for PC for gmpls, and there was another RFC 8281 which is called PCE, initiated LSP set up in a stateful PCE.

F

This one is basically you know: PCC delegation, controller, locally, configured LSP to PCE, and this draft is extension for gmpls and I. Think this draft is quite mature and likewise this one but I think working group chairs felt this might just merge together for the benefit of reducing the process, overhead and so on. So we basically merged that the second draft into the first one, because first one covers with a bigger scope and we published the published version and ready for working of Lascaux I think through mr. Dutchman Sheppard's, right and I.

F

Believe I deflected your comments on the first one before more 0 and after mojo, you have some comments to make it smooth, as if you know you suggest earnest to similarly now so that it's one document rather than merger of two documents.

D

Me just clarify this is true, but much is done. We now just need to make sure that the document feel as it's a one document and not a disjoint two documents which I just put in one place, so just a editorial pass and movement. You do that, since this document was already in the working group last you it will go back into that queue and we can even let you know, get it a little bit more higher preference than the other documents that were added later.

D

One thing that I would also suggest is that the state, a normal gmpls piece of extension, is also in the is G. So please look out for any comments that you that we received there and align this work with it. So just also keep looking for this thing in the next week. So.

F

D

You finish that I think it's back on the cable, okay,.

C

As I don't see any question from the room, we move to next presentation by how many.

G

Okay, this is how me speaking and I'm, going to present of this work is extension to the PCE communication protocol, for he has Tara and notification. So this is a road work.

G

So let me go through something background, because this work has not been presented in recent nursing, five or six years. So we need to emphasize again that we need this work, because this is motivated by that we issue the address Eric, how the handling issues between the PCs originally in the fifty forty I, feel you.

G

There is ever code that sniffing handling process SS in between the PC and PC C's, but this one is a kind of more intricacy needed and we also try to review all the error code has been become defined in so far in the working group and we also define how to process between the PCs. So we assume that every draft to who has been working on the multiple pcs in areas, including McCann, be multi layer or multi domain or hierarchical whatever.

G

So we need to specify okay, whether we should handle the errors in this way, and this work was studied in 2010 and the doctor in 2012. So some of the authors- I ever don't know, but it is now reactive we still alive again and so currently, I feel that there are quite not many attention so and but we believe this one is still necessary, so I'm presenting here to draw your attentions to take a look at this one.

G

So what's what's in the trust, actually, we have specified two different section: one is forever handling and the other for the notification, so error handling. Will we have two separate objects and we decide whether we propagate this kind of error? Yes or no, and what is a critically critically T level of this kind of error can be low middle high. So we will comprise the existing every County to all the combination of the propagate categories and there's a credibility level classes for notification part.

G

It would be a little bit simpler and it only need to notify whether it is a local or it's a kind of best request. The specific.

G

So here we provide very detailed handling rule how to allocate the propagation value and the critic ability level for each given PC era, a.

D

Clarification question: what do you mean by on demand here all.

G

The bound a is a kind of we have not specified anything, it's a first time, for example, if the class 11, the quick particularly, is recommended to be low, that's the propagation! You can decide whether you want or you don't want.

D

Basically, you want unspecified as the right word like we are not giving a specification you can put 0 or 1 based on your requirement right.

G

So actually, this is not exclusive and seem because some of them have multiple values, depending on the urgency or some kind of that.

G

And yeah, another motivation with we we had to present here is because we have some quick check on the existing path, and this will definitely impact as a kind of work, and we are assume all of the works will be good and some of the inter PC. We assume that we need this kind of error handling, but we currently don't have it so far. So this is some potential comments there can be, but I just find a issue here. I wasn't aware the first one is an informational one.

G

So the objective is to have every multiple PC related craft in standard track, to have this kind of descriptors in session, indicating whether the new arrow types would be needed and how the propagation type and critically Atilla tea level would be encoded. I mean.

B

Are you expecting that the authors of drafts will do this, in which case you may be ambitious, or will you talk to the authors of the drafts and help them by suggesting text yeah.

G

We can do that offline and I can, but I believe this part should be in the separate in their cosmonium draft are not present in the draft.

G

So we're looking forward to see some discussions or whether this is kind of agreement. We can reach in the working group to do the error handling sessions in corresponding drafts. It's certainly an important work, but yeah we have too many important works, so it may not be an urgent one. I agree, but I'm just wondering one I want to understand from the working group whether this kind of inter PC publication may have this kind of dependency all this work.

G

If, yes, it may become more urgent than before, and we're also disclosing discussing how to close this work, because everything current so far is stable and we hear requests for the working group to have a some attention have some review and we are preparing for the last call of this work.

G

That's all for presentation comments. Please.

D

D

While we are waiting for someone to stand I second, what Adrian is saying I suggested that earlier as well, that let's just provide text on the mailing list to these authors. That I would like you to add this section, and this is the text and if they disagree, then we really are seeing that your your work, users, the people that you are writing this document, for if they are not convinced that this is needed, then yeah I think our work is not going to be that successful.

D

But we know that there are a lot of documents. Hpc is there. We have one document today on the agenda as well, so maybe we can use this time also to talk to them and see if the error codes that are being defined there are the mechanism. What is the criticality level, and this can also be specified now, at least and all the new documents that are coming through, and we have our ad.

H

Yes, Deborah I would like to suggest, since you have now wikipage possible, you could put some other guidelines there, and this would be very useful.

H

Security considerations would be useful. That would really be helpful, so.

G

Let me try to confirm whether the last item here can be consensus. D in the working group saying that we only focus on those kind of not spoke, PC related crafting standard FAC, so I don't think we need to specify too much on the informational or a single PC. You want I think.

B

You're all right about not needing to do too much, but I, don't think that's nothing so I the comparison I would make is with security considerations. An informational document is not going to need to do protocol specification for security, but it probably points out security issues and I think the same thing here. So the an informational, Monte PC thing might just have one line that says this is multi PC, so it also consider.

G

I

C

We're moving to the United I presentation by each oh yeah.

J

All right her just to give some description for appetite habitation for the PC PC extension for the networking.

J

So here are just the what's the proper role for the PC peace. A prejudgment were your PC to assist an assist our traffic engineering, our network network network. So here we proper to UM TCP facility for two reasons. First, is the PID MPP pure dynamically and rapidly sir, the? Secondly, the proper popular different prefix epithelium, certainly is many clearly to the past each evening. Kapa on demand based on the network. Consider is, this is main aim for the for the released protocol release drugs and the employers document the way just the propeller suite.

J

Please have object, which is charity. The PC in this LSP request, the following is real detail of project, but after we discussed with our observed, her least physically attention for PC controller proposed the PCC architecture for the for the instruction controller instruction download. So we think peace and unity, IP network held also some flavor of the pieces, is a character. So we, after discuss with each page of with other, also of the job that we had to take the similar procedure to proton, is a kinky object.

J

So here we just the proper define my new and since I objected, type and I put the information into unity, newly defined the CCI object, as it was Subterra way. So the main content of the way not trained here is the key field. The first day the apparel will be downloaded, appear, IDs and IP address. The second is previa courses in PPA. What we were terrible idea, NOS circita prefix, the shoulder is for a PR for your magna purely deposit will next hopper towards a PT, PP R.

J

So here we just look for some comment. Australian for the newly proposed and clearly especially from the implementation, will point. We also considering the the future called reason between the commander from with tradition on it, nakum, young or Sarah, and the dynamic piece aybe protocol.

J

So this is all for the application.

C

Any comment question.

B

Thanks agent, has this been circulated to the idea of working group not for them to work on, but so that they know we're doing it.

J

I just ascended the the leader person to for DeMeo these document, oh sure, and also discusses the medium of line of our solution. So maybe.

B

Just having a back of your head periodically just to send an to their mailing lists, just to say, don't forget we're doing this work because we clearly it's related to how BGP works yeah and we we need them to be aware.

J

D

I was just going through and I realized that in the teas working group, all the related documents are experimental, but in the PC we are aiming for this in standard track. So is that is that the plan, or is it just a mistake in picking the track, but.

J

D

Then we need to talk to T's, yes as well that there the work is going experimental, ideally, I, think we have oven there.

K

As well, so you have any concerns. No the two documents that we have both our own experimental track. Yeah there's, there's no plan to move it out of experimental media. So unless there is some feedback received here, we don't see it getting out of experimental anytime soon. So I put you on the spot as to why that's experimental.

K

There were a few reasons before adoption. There wasn't much support for the document, but there was some interest from a few people to see fair. This focus. So given the lack of interest and yeah, we decided that was the best way to make take it forward. How.

J

Can he change the later yeah yeah.

B

And everything can change right right until versus possibly.

A

B

K

I I don't see why this needs to be a scientist track until the problem.

D

Yeah I would also prefer if this is remains experimental but I- think it's up to the working group as well. So I think we can take that discussion to the list as well, that what should be the ideal track for this work. Another suggestion would be- and first I would like to thank you for my personal comment that I gave that, if, instead of introducing new objects, if he couldn't put it within the PCC architecture which you kinda greedy, so which actually makes the document also much easier to follow.

D

The second comment that I would have is without my chair hat on first time, I said it that that we need some more description in the sense you have three different parameters: one is to create dynamic, bgp session, another to put prefix, which are, in my point of view, aiming towards the towards the end points. But then you also have one more TL v which might be needed on all the nodes in between, so that clarity is not there in the document.

D

Right now we have justified the tlvs, but we need to go a step further and explain the procedures and the interaction between piece, F and BGP. If there is any and error codes, all those things so there's still a lot of work. That needs to be done in clarifying itself that what how this will is actually used beyond what is the messages and what is the object? Thank you.

H

Toppers beginners ad, yet experimental, just clarify what Adrienne said: it's: it's simply um that you can go through and publish it as experimental and then come back three months later, five years later and take it standards track once you once this has some maturity. You know it just so so don't worry that it it's experimental, but what needs to be shown is interest maturity, implementations that can be stable, so.

B

I got a little more on this that the t's documents are more in the position that being framework and use cases. So there I think the choice was experimental or informational.

B

Here, it's protocol said the choice is more experimental or standards track. I think this whole question is going to come back to what the Russian group thinks about implementation status if we move forward, but ultimately we don't have to make this decision until we make a publication request. So when we go to the ad in section we have an RFC, then we need to make a proper decision. Now, it's just just it's just a flag.

B

C

And now cross your fingers we're moving to a remote presentation. The one on PC is central controller.

E

I

Yeah I could see, you could hear me drew yes.

E

I

D

Good good yeah well morning, everyone, this is Mahinda Meggie from India I, just have buzz, update on PC CC extensions. These two IDs quickly go through that.

D

Can we move the next slide, I! Think I'm, once I do yeah.

F

D

About background, PC has a central controller, the main RFC, the architecture RFC is already published, and we have basic PCC draft recently, I'm not adopted by PC WG and the PCC see Assad the specific to segment routing extensions for PCP. It is in adoption, q2 next slide, roofies.

D

Yes, a major update on these two IDs are supporting binding, sida location by PC, and we are reusing. The path binding PLA, which is already there in use and PCC, is a PCC, a PCE request, PCC to allocate binding level using this t path, binding TLV. So we are still using the same TLB bye-bye in which PCC request PC to allocate level by PC, request, message or PC itself can allocate the level and binding level and push to PCC. So this we are reusing and the another one is PCC a location.

D

They label a location by PCC on PC PCC sees request, so this one will see a bit in detail. The first part we are just reusing the another part we will see in detail next. One please.

D

Are you seeing.

I

The dislike yeah.

D

Yeah, in this case, in this case, PC requests PCC to allocate label instead of having explicit you know, control over pcc's level, level range or level block, so this we feel could be.

D

So disable the video and continue I think he's dropped off. Ok, so save you time. Let me take over and of this without taking too much time. So this is Ruth without the chair hat without the auto' hat, just a contributor to this document, and and so as Mahindra was saying that we have a see flag, this flag basically says I'm, not gonna allocate the label, but I would like the PCC to do their location.

D

On behalf of me and to do that, I set the flag, don't put the label value, send the message to the node and node allocates the label and send it back to me. So how does this work? The PCC C goes to the egress goes to the egress says: can you allocate a label for me, but this is my instruction: the egress allocates a label and reported back to the PCC, see now at the transit.

D

You have two instructions, one where you are specifying a particular label, which is the green, which is the out label and on the in label side. You request the node to make their location and the node makes a new location, which is the label yellow and kind of report. It back to the PCC, see and same thing happens at the at the ingress as well. So in this way you basically have the label allocation going from egress transit ingress in a sequential manner.

D

This is different from the case that we were we had earlier in the draft where all the label management was done by PC CC itself, so the things can be done in a more parallel fashion. In this case, labels are coming from the node, so you have to do it in the sequential fashion. From the back to the front.

D

The main reason why we added that in some cases it was difficult for us to know the delegation of the label space at the PCE for some reason, because the node may not be in directly or control, it could be a border router or it could be various different things. So we wanted to allow ECCC to also just tell I would like you to execute this instruction, but I'm not specifying what be able to use. You pick a label from your space and reported back to me.

D

That's the small change that the author's have done in this latest update. So the question I think the yeah manga. Do you wanna place yeah.

I

Yeah yeah yeah.

D

Almost you finished it right. Yes,.

L

I

D

For that, so so authors, you know, expect comments on the the recent updates and feedback from other implementers and we have we request the adoption call for PC CC s, our draft suit. So any any question on this particular update case.

M

Yes, hi, my name is Eric from juniper. In the case, why are you mixing the allocation mode from PC, PC, ECC and distributed on the device? There could be cases where you have collision. Are you thinking of carving the label space? Like you know something managed by the controller or something managed by the device? Is that the recommendation and okay thanks.

D

My no you can answer, but I kind of shake my head, which we communicate by head movements.

I

Yes, yes, yes, yes, you're right.

C

You know comment or question if not we'll move to the next presentation. Thank you. Oh the next one is a tune again well.

I

Thank you. Thank you very much here. Thank.

C

I

J

Here, I, just probe negativism for the tckontis I give space and the culture.

J

Motivation for this raft is, you know the PC. The pvp extension for PC controller specifies the procedure and protocol system for the PC Ellison controller, so your label can be disappeared. The town stands for device select any PC P and there are no draft deleted for the SR deployment also specifies the procedure for a piece of Education for to Daniel is our label. However, the document assembly to the label around you to be used by PC.

J

You know and the second piece of approach it appears by the Samsung- have some mechanism to spotted their hunger, other situation from holdup, identical location and the parity. Second epistle only the PC controller. Now your ID space reversely.

J

So in order to a local IT directly from a PC and the underlying, he says it should be so the under under under Isis control, ice-t space and twenty pieces. This is the main motivation for the draft. This is a hurry. The message flow for the fully staffed so forth, allocated in the controller control IT space, the delete your ID ID, especially on away, must be included. You know open up message and for each each. Each carrot hope it should be included once you note message. Currently, we can consider mainly that token.

J

The theory for the first day's labor, contrast based and here in the second area, is only six function. Id control, space, TR away.

J

This is the format of the control lever, controls bestie only and he's a chance to describe a block as the start, while you're under around you. So after knowing this controller space, your PC can assign the label, such as the bride incident. I said, formerly the space.

J

The second is SLE six songs, ID control ID space- is he handsome? Your structure is for materially her. He defines the valid is no say mattifying the hello Frank to indicate the look with her either exists or not. So if this is a fact easily flag, he said then look at her tyranny. Associative incision, look at her fulfilled our societal regarding creating. That is the order. If it not said, then it will not increase again survey.

J

Her listen, the carpet, we just element a flake, let's define the universe which, after the the main aim of a flag adjuster until.

J

Indicated pc-12 dedicate all space to look easy, but after his cousin is such a scenario not really common. So it is a very increment. The PC needed to define the labor want you to hold it pcs or it is not necessary for this flag and the secondly, the we under your culture problem under Cisco and with other small needs. So the experiment may only see missing the meeting in function. You can't PC and GCC and is all related after so convexity inquire useful for the derivative work.

J

So is asking for a comment and the question walking adoption.

C

Any comment from the room I went question related to the previous one in your other document. Currently PID is flagged as experimental. Can we confirm this is on purpose.

J

C

Eid document is a experimental on the header.

D

Instead of standards track, this document is currently under experimental, so Julian is asking. What is the reason is that is that a thought behind making this document an experimental.

N

Actually, from always occurs, first of all, we choose the type of experimental because we think he's maybe no so easy for cold for dead track. But if you want we can go for stair jack. That would be better.

C

Thank you, for you must answer.

C

In a recomment okay, so let's move on! Thank you.

C

The following one is Mike.

C

We don't clean yeah.

C

Yes, we can hear me: okay,.

O

Great yes, I'll be presenting this draft piece of extensions to support segment routing a policy candidate passed next slide. Please.

O

D

O

There's been three versions of this draft. Oh sorry, previous slide, there's.

D

A little bit of lag between the slide change so right now you are on history of the draft right.

O

Yep, so there's been three publications in this draft. First, we introduced the overall structure of you know tying the piece of LSPs into an association group to represent the SR policy. In the second publication, we defined the three TVs that signal the different policy attributes and the in the third or yeah in the third publication. We just did some clarifications next slide. Please.

O

So the purpose of this draft is there was recently introduced this concept of segment routing policy, which adds extra structure to segment routing. It basically unifies kind of different candidate paths into a single SR policy, and it adds the concept of color and other things, and so piece up was lacking a way to signal this information.

O

So our current draft is meant to address this gap and piece up by allowing both the PCE and the BCC to signal these extra pieces of information that were introduced in the second routing policy. Draft next slide, please so.

O

Basically, this is just like an overview of of the draft, so we define a piece up LSP to correspond to a segment out in candidate path and.

O

The group of psi powers piece is tied into an association group, and that represents the SR policy, and the SR segment list corresponds to the P sub arrow.

O

And we define the TVs to signal these various values next slide. Please.

O

So this is the TV that gives the ID of the policy within the context of the headend.

O

So basically, all the piece of LSPs that are within the same as our policy must have the same value of this govt and correspondingly.

O

Opa power speeds that are not within the same as our policy must have different values of this govt.

O

So this yeah thanks- and this is the candidate path- govt gendered path, identifier too.

O

So basically, it must be different among different candidate paths.

O

In the same as our policy excuse me so this this encodes stuff, like the protocol, origin and the originator in the case that the PCC may want to report non PISA candidate paths to the PC. For example, if you have a configured candidate path, the PCC can report that configured candidate path to the PCE and specify the protocol origin as configuration.

O

And similarly, if you have a BGP te instantiated candidate path on the PCC, then the PCC can report that to the PCE and set the protocol origin to be BGP.

O

Okay, next slide, please.

O

And this is just the attributes of the candidate path that do not serve as the key and they I believe they can theoretically be even more during the lifetime of the of the PC powers b-but, the previous other TVs must not be modified during the lifetime of the LSP next slide. Please.

O

So next step is we plan to initiate workgroup adoption and that's it. That's it for the presentation time for questions.

P

Question coming just interrupt so number of comments you've been using to the policy terminals P term interchangeably, I think you should put a big letter that one means another I.

O

Mean no I mean policy s. Our policy is a group of piece of LSPs. That's how so, as our policy equals P sub Association. That's what like this. If you go to the overview slide, the.

P

Number of places where you specifically sr l SP term so.

O

S our LSB, yes, is that, like in the actual text of the draft, let.

P

Me get my gloves.

O

P

Q

O

If that is the case, we will correct that it shouldn't. It should be P sub L SP, not srl SP.

P

Any thinking about using binding seats in case where you cannot instantiate LS space, / policy / and to end.

O

Binding Sid should work as before, like there is no, we do not kind of add any new process for the binding SID, so binding said this still belongs to the P sub LSP, which for us is a candidate path. So so, basically it's the same.

P

The turning point might not be the same as end point where bagging would be expanded to reach again. If you cannot push enough labels, instantiate full policy.

O

Sorry, could you clarify the question, I didn't think I understood the question so.

P

You'll, be human SDS, one of the constraints to compute the path right. If you cannot.

Q

Push enough flavor.

P

Then you shouldn't be doing it and you'll know your MSD a time of computation right.

O

P

You might decide to use binding Sid instead in order to segment the tunnel. Okay,.

D

I think what's happening is that all you're saying is valid, but it might be there in the normal SR document itself, because this document is not talking about how to compute the candidate path and what computation to can consider it is talking about when you have multiple candidate paths, how to group them into an SR policy using an association object, understand.

P

However, yes lower, when you implement it, it should interwork yeah.

D

It's sure I think you're, putting a right hook that, yes, it should into work. In fact, I wanted to also ask the room that one thing that the document currently says also is that they assume that within a candidate path, they have only one CID list, and but the SR policy allows a candidate part to have multiple set lists. So there is still lot of work to be done in this space. Okay,.

P

C

C

Thank you, making that you.

D

Point I also wanted to use this opportunity to, like you know, talk to the all the author's working on s. Our document spatially make sure that the there is some alignment between BGP work and the piece of work. Recently we had come in from Tariq earlier also, while we were progressing s our document there were some issues between the some flags are different between the two documents and they share the authors between the BGP and piece F. So we request specially those authors who are involved in both documents.

D

Please help in making sure the alignment is there. Thank you.

C

Yeah I think the next one is.

D

Michael, my question is related to candidate paths. Now can PCC initiated and PCC initiated candidate paths can coexist in this yeah.

O

O

The protocol origin can be set to piece piece up if we go back a couple of slides. If we go back a couple of slides, please.

O

D

Question I think we can move on yeah.

I

Thank you, yeah thanks. You.

C

Okay, thank you. We move to the next presentation, then.

C

Which is change.

N

I'm chun-li from Horry and my topic today is about a suppose segment and my bad Russian oppose in p-set and I, won't repeat the whole details of the draft assume that you have read the jobs already since I have presented these for several times. So, let's, let me repeat the motivation why we do this? First of all, in other words, you know use cases like n, 2, n, +, 1, +, 1 and pass protection and and like bi-directional, pass correlation.

N

We need the power segment you identified as a path, especially in as I'm Pierce network, so we proposed a draft in spring to change it. Instead and recently, these charts has been adopted as I, each F working through a draft, and then we also proposed a draft to introduce our segment in hobby 6 network so for allocating the power segment.

N

We suppose that the piece of extensions in piece F annotate right, so we proposed this to document, to describe the the mechanism to allocate the power segment and the use cases of by the rational pass a correlation.

N

So the mainly the bank updates of this document as we move the RSVP flag to PC allocation flag, since we noticed that this flag can be used for multi for multiple use cases, s such as biting seed allocation and power seed allocation, some time, some something like that. So we we change the flag from post segment allocation flag to a PC allocation track. Also, this has been mentioned in PC ECC drafts already, so just update to sync up and the second is we we adding.

N

We add some text to describe a mesquite consideration in the first segment, an allocation, so the rest is some. The rest are some minor changes such as reference sync up and some needs. So basically we add a new cause, a junction from ZT to join this draft and for implementation status. Basically, a highway is implementing this in PC and PC C products and I suppose that he is if he would do the same thing as well, I suppose and the updates for this draft.

N

We basically we mentioned we described how to propose a new Association group to bite the the the s passed to be a bi-directional house in Assad networks and this term. We we update the jobs to address the Commons from Marina, marina and Chen, Chen, Xiang and and then we we fixed the the the earth based in figure, which represents the direction of LSP from PCC point of view instead of PCE point of view and the others are some minor auditory changes.

N

Also, we added Chen Chen shown as a new Corsa and next plan since the draft stable for a wire. So we suppose that we are ready for working through adoption, because the commercial delivery are going on and and this mechanism are this mechanisms supported by the operators and vendors. So so we click request. The hooking group adoption, just like we did in on the mailing list and comments and contributors are very welcome.

N

M

Question carick from Jennifer, so it's about.

N

M

Terminology actually SR path, LSP and what is it? Everyone, oh yeah, I, Candida path or maybe segment list? That's what I was looking for segment list. So could this be the segment list ID for example, mm-hmm? Or is it the LSP ID or is it what basically.

N

The past, maybe we.

M

Just need terminology.

N

The past document can be useful identifier, see least Oh Canada pass Oh as oh and as a policy as you want.

M

The way I understand it is the candidate path can be many segment lists. Yes, so how can you identify with one path, ID many segment list for.

N

Example, there are two parsec Canada pause within as a policy, and then we use one.

M

Candidate path, many segments lists one.

N

Can infer many okay, so I suppose that you, okay.

R

From Cisco, so it depends on the use case right. It's a path segments, depending on your use case. You can have it for the whole policy or for the candidate pack or for the segment list, so the it defines the concept and how you allocate it it it's based on how you want to use it. Yeah.

N

M

Okay, maybe expand on this in the draft so that it's well understood and.

D

I think the right place, for that would be the spring document where this thing is defined. Thank.

N

You thank you anyway. Any comment. That's it.

D

Thank you last time when this was presented, I think John asked the question that we will progress. This work once the spring working group has adopted. So luckily that has happened so I think this would also be a right time for us to see whether in the PC working group, whether we should adopt this work.

S

D

Do you give for one of them that before other or do you want both of them? What's your preference, I suppose.

N

The boost of them would you like to expand in it? Sorry would.

D

You like to clarify.

N

Why she's? The perfect man is the basic mechanism for allocating the segments in this set, and this is the basic solution and then the key use case for post camp. Our segment is about the rationale: post, correlation and those two jobs has been stable for a wire, so I suppose that they can be yeah.

J

B

It sounds reasonable, so we've got these now in the wiki in in our pending queue and then we will move them to the pending.

B

So there are, there are a few documents for adoption in the queue before then yeah. We should be rippling them up. Okay, understand.

Q

B

Working group that does not prevent you from making comments now you are allowed to read the drafts and talk about them. No.

N

Problem, thank you. So that's all. Thank you.

C

Okay, so this is a summary of the latest change in the draft that has been presented.

C

It's actually merging to draft one which was focusing on the state rule will enter the main PC on the chain. Bizzy's. A new RPC like approach under the one which was focusing on the euro Kiko method pools, are addressing very similar problem space inter-domain a mistake for pc context, so we managed to match, pose it to a common document on the point that it clearly clarified where the various approach that can be under there, there have been some improvement. Also, we know draft itself verification fixes, correction.

C

Context, we believe that the text now is quite clear to be considered for further move.

C

Welcome to my little show your interest, there was a major change in the section number.

Q

C

On the selection between VP te and segment, booting types when for a yes type, the the chain of also this was piece of key information which was part of the draft and it's greatly improved now. We still believe that there is a hot topic behind this problem statement. There are some proposals at heart trying to address some data, synthetic connections or all those good fit that benefit from this solution.

C

We see also some interest coming from the community when it comes to combination, stitching nesting by ninive pass type of pass or tunnels from a mix of meteors emigrating.

C

We also noticed that there are some common ones on proposal, even though it doesn't fully addressed our problem here. It's also a bit connected and we felt that the full picture wasn't fully there. So we propose to refine the solution to the problem coming here in for social I, see someone queuing, you have a question for me.

B

C

C

Technical solution that has been included in the graph is one option. We believe it was reasonable enough because so far it doesn't extend the protocol at all. It just allocate I propose to allocate some new code bite over existing registries, and those footprints are triggering behavior that are defined properly in. The draft, of course, is working with feels that this is a problem we want to work on. Has a working group, any other alternative option to specify the way to incur those is follow up on.

C

It is the solution that we consider reasonable for the first step. Others are possible on.

C

Hopefully, if we adopt I will talk more about the details of whatever the idea is to work on the problem define how it behaves also work on your calling and then consider that the proposal coming here are the only way, even though the good news is that we have an implementation using those, including you, is that it's not the commission of the rotation, so it will be updated according to any modification of the specified own career if the draft becomes working on the.

C

um If you want, if you are interested, there is a YouTube chain on YouTube content that you can click onto to watch the demo firmly so currently, the draft as far as our concern is really for what will be production Oh, since it's too late working to join the queue for orkut interruption.

C

There is a technical solution to the problem, knowing that is not fixed. It's open to discussion for any further work, it's applicable to stay for context, but we try to keep this pair autonomous system policy. Naming that I mean yes can evolve using SVT signaling. Maybe one may use second routing. It allows to combine boss on a common service into n with the two technologies, and there are some interesting use case on the road. So we request to join the walking a bit from Q question.

C

B

This is Adrian proving we need three working group chair, so Rankine chair or one present. Someone asks a question: I have two questions: is there an assumption about the end domains and and whether they are under the same administrative control or not? There is no on the.

C

Contrary the RPC based approach was built on the assumption that different, as opposed to be working alone, who usually on that in situation where you have a common entity controlling all the child's the children.

C

You know pc-like approach, doesn't assume any client server or parent or child relationship. It's a flat which, if you like approach like in the autonomous systems over Internet.

B

Second members: are you aware that there's a document in Bess on rather it's it's a bit suspiciously called domain gateway discovery, but actually it's much more about how you get end-to-end connectivity, similar to the picture that you've got on that slide, with the additional thing of finding out what the points of attachment are from those edge domains to the transport motor.

C

Oldest one, but something.

B

Definitely need to talk is that's approaching. Last call, okay,.

C

Good to know, I talked to the best chair and.

B

A question for the chair: has everybody in the room sign the blue sheets.

C

T

Good morning, everyone I'm then promised a te. The topic I want to talk, is P CEP Ignace for parity.

T

In Ferguson, yes, this this document specifies a consistent to the past compute PC. Sorry. This document specify citizens to see if he that allow a stem for PC to compute and initiate the path for the beauty.

T

Extensions in our draft we have defined for in for his essence, the first a spear Kappa the world Testament. We define a new path set up time for beer and that they find the beer te PCE capability, slapped your way to exchange fear capability, the second reactant a P and a sappy object. They find a new path set up type property. The throw is in points object. There are two of options.

T

The first is, we use the P 2 MP and points object protein for ipv4 and ipv6, which depend in FZ 83 the road six and depend on the. The other option is to find a new, pier and point object to can reach a PFR, IT informations, the last a tenth year or object with defines a new and imperial or sub objects to carry artists, assist with strength, psi, au sub domain and si yes. This is the details of the existence for beer capability.

T

The world has meant for extensions beer capability, yes, a new path, setup type and the beauty PC capabilities of Jia Wei are required to form it is below, and this is a key as a key object. The order to set up a beauty, a new path segments, a setup type TL way must be. Content are key as we object, the which means passes said huh Buren beer T.

T

This is the endpoints object for specifying the beer information of the past, for which surpassed computation is required, requests and the points object is required. They are two of since we use the PDMP endpoint object, called a for ipv4 and ipv6, which city we tilted. Finally, up c8 stress: there were six optional, a we defined a new peer and point object for peer, 4pf ID.

T

The format is below account in the contest leaf, type, sauce, tfiid and one or more destination PA parties. Yes, the last a yahoo object for carrying the beauty explicit passes. A new period are all sub. Object is required at the contest PS. They are sub domain. As I and associate a bit stress for PSL, the max last of of the biggest tree is 5 in turn it in the it indicates. The last of rapid train is Westerlund 24, because.

T

Because tellin ya, all object which defined the length is limited to 2 8 bits, so we limited the max the value of psi to the 1020 base in case the beer euro sub object is too long'. Yes, next step comments. We work a welcome and wish think that we should kiss ASSA is to the extent 8 for peered. He is needed and the way well the next walk we. We would like to add this. This part, where chrome option hope to enjoy our work this, how it may be another rough days. Ok, yeah.

D

Any comments from anyone.

D

I was in the beer working group where this document was also presented. Yes and then I further talked to the beer chairs as well. I think BHS suggested that this work it should be done in within the PC working group. So that's something that I need to further talk to my chairs as well and conclude- and the second point also that came up- was that there is the yang work that is going on in the beer working group, so what they suggested.

D

The Bayeux chair suggested since the beer yang work could be the key reference for us when we develop the PC PC extensions and most of the things that is there in the yang can be used as a reference when we are developing the PCF protocol so that there is alignment between the yang work as well as as well as the piece of extension. That's something that we might need to consider I'm, not hundred percent sure how to work that thing out, but that's something that we can take into consideration. The.

P

D

P

Itself, thus far has lots to do this. Beer. It's especially for beauty and about ambiguity, should specify a number to Young's. Mother for beer is not going to include beauty, so the Youngs model is going to be to be defined in future and I agree. It should be cool aligned between acid Jeff.

D

There is a tiny bit of a confusion there. I also was confused by that point. So what has happened is that there is a beer yank. There is a beer tae-young and then this document, as presented, is only for beer teehee, but DCCC for beer could be for a location of bfr IDs that can include both beer and beauty. So that's why there is a bit of overlap happening there, but it needs to be clarified right now. It's a bit confusing. Yes,.

T

U

Burger, this is actually more of a comment for the beer working group, but since it keeps coming up, I feel like I, have to say it. This is a yet another case where we're having confusion over terms, so we went through this with spring now we're going to go through it with beer that we're really talking about your policy and past steering, and that's what what the terminology that we that we have at spring and it's this it we're providing the same function with beer here.

U

We should be consistent in that terminology, and the nice thing is: is the person sitting to your right is yet Adrian who's trying to duck is actually helping us through an update on the what te means in the IETF. So this is going to be a great topic. I, look forward to seeing his future document on the topic that well make it. So we stop having the same conversation over and over again so policy and past steering when we're not doing the resource allocation. Is that what least what we're using in spring and in teas.

Q

Son, isn't he mmm I just still want to clarification as a discussion just happened, because they're te can be used to resolve beer. So it's a separate two things yeah believe me, so you can use PC ECC to allocate tfiid for peer and the also you can allocate the agency ID for peer te. Yes, it can be used in peer to the ECC and but, as this document is the extension for purity pass steering, so it can be used to calculate the pass impurity, but not in peer uvr. We needn't calculate as a pass.

Q

Yes, so it's the difference. Yeah the.

D

Good thing is, we are in agreement.

S

C

Thank you. It's funny to note that more people get involved when it comes to beer. Okay, I think we're done for today.

C

Thank you. Everyone for attending the session tomorrow, hopefully.

B

Thank You Pam where's, the other blue sheet. Please.