Internet Engineering Task Force 103, 7 Nov 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF103-MPLS-20181107-1120

Description

MPLS meeting session at IETF103
2018/11/07 1120

https://datatracker.ietf.org/meeting/103/proceedings/

A

It's actually time, I.

B

Given close the door.

B

And you can see if you can see Gregor.

C

So it's obviously something with the brakes here in Bangkok people are very, very reluctant to leave them and come to the meetings. So if we are quick about it, we can take a lot of decisions now and then go and do something else. Yeah I was about to say that we have something out here here to share. That has nothing to say. We want to start the only problem with that is that the first presenter is Greg and he's far far as I see not in the room.

C

Can someone like Martin help us to to find him.

D

C

Have one further comment: I presented a long list of sleeping beauties at the last meeting at the meeting yesterday. I have got some response. I want more, are given to send the list out in the mail, so you can review it and see if you have anything there, it's actually I look at the list and now and then and see that there are things that I think we done right there offering in there that I think we done wrong and there are documents that actually would like to progress, even if they do four years old.

A

E

A

A

C

I've been that all morning.

C

F

Morning, thank you for the patience.

B

F

As you can see from the draft short name, its om for SRM Phyllis over IP or over UDP, so what's in scope, scope, it's a fault management on demand for detection, localization, characterization, proactive monitoring path, continuity and performance, monitoring, delay measurement and packet loss measurement.

F

Fault management, so we have RFC 8287, which explains applicability of LSB ping for asar MPLS and the pretty nicely describes ping and traceroute over SR MPLS and works asaram LS over IP seamlessly drag nurse key spring. Pfd also describes a applicability of BFD over LS LS, pianist Toronto as the main and again there is no issues to use it for a syrup. The LS over IP.

F

Both documents, so RFC 8287 uses recommendation for 88 in Hein IP UDP encapsulation. If someone is looking for non IP encapsulation, then there is an issue with the preserving source. Id source ID can be preserved using IP address Tildy proposed in draft that discusses MPLS applicability of point to multi-point BFD.

F

Performance monitoring, RFC, 63, 74 packet was and delay measurement in MPLS networks is.

F

Entirely applicable to SRM TLS, and there is no issues using it in a serum pls over IP.

F

So the conclusion is that existing Orion mechanisms in MPLS networks they work in a sarong POS with very minor extension that doesn't require really a separate specification, but there are some other issues, interesting issues coming from the fact that when we're doing SRM, POS / AP we're creating multi segment tunnels and with multi segment tunnels. If we lose identity of the source, then there might be some interesting issues that I'm planning to explore and see what happens?

F

Then I discussed with the one of the offers of SR MPLS over IP whether there is a reason to have a separate document or Abbott as a section in SR Intel s over IP document. But suggestion was that, since this document is mature in close to the working group quo, so probably it's not a good idea just to merge anything in and restart the discussion. If anything that will be of substance in this document, it can proceed as a separate document.

F

So for now welcome questions comments, participation, contributions. Yes, please I.

G

Just have only one comment for SR, MPLS or SR in general: they've also SP FTP solution for Esther policy, and that also works, as is yeah.

F

It's it's. A very common thing is a far I'll add this reference and in the document so that it works seamlessly. Yes,.

A

F

Okay, thank you.

A

Next is so far.

H

G

Thank you, so this is.

C

Okay, we have to get a new set of slides, they have updated them and we haven't really uploaded.

G

G

Thank you, that's good yeah. This is thank you. Thank you much. So this is a word that we started for extension of MPLS LS, beeping and and for SR SR MPLS based LS piece I'm, going to present it on behalf of my quarters.

G

So the main thing is that initially our 382 880 287 was defined factoids for to be using SR angeles network for prefix, it n agency said, and they were related procedures that are fully defined there. This draft addresses some factoids. There are not address in the base RFC 80 to 87, specifically the base, as he doesn't talk about bgp prefix exact. So we added a statement for that.

G

Bgp epe CID, incidentally, like us, found this gap as well as Shraddha who presented this yesterday in NK s, working group found the gap as fall, so there are two drafts that talks about VP EPP, that type for BGP EPE sites and as I go. I would go to just to highlight some differences, but there is some discussion going on with the sugar for cooperating on that and then last but not least, is the binding set. So the so.

G

That's the problem statement here so for the BGP prefix said there is basically no need to define a new fact type. We can use existing FAQ and just define that we can use existing factor. This is just maybe clarification, a statement because implementations are already using this implicitly so I think this should be a non-issue then. The second one is the BGP for EPE. So EP has three different type of set. One is the pair node set, and then you have their agency said, and then you have a pair succeeds.

G

So that is define the EP be TPP draft now here this is an encoding for the FAQ for btp EP. We are notes it here.

G

The information you find would be very similar to what red represented yesterday with the in the procedure. There is certain changes and that basically, we also try to highlight or address the case where, when you go to the neighbor of the BGP EP, the the node that is I, wrote, I think the big appease it and you go to neighbor of it and that neighbor may not have because the different yes, it may not have a route back to the source or it may not be even an MPLS router.

G

So in these cases the neighbor cannot respond to this request, and then we put some consideration that I will go at the end of the talk. So this is something that were different from the drafts that represented yesterday on the APB's it, and the other thing is also that what we notice is some time, because this is a peer node said it is possible that the ingress was trying to validate this peer nodes, said may not know the interface addresses it may or may not know.

G

So that's where the I flag is set to say that interface addresses to be nolde, because the peer the sender may not know that, and that's what not there in the encoding that was pregnant, just to reassure them, but minor difference and then the same thing with respect to the BGP peer agencies. It is, there is no need for that flag here, because this is a chance he said.

G

So you know the interface for sure, so you notice that there is no I flag in this case, but in terms of whether, where you go to the PR agency, the problem is still remains there. The pier agency, the BGP Pier, mean maybe an IP router or may not have a route back to the source and then last is the BGP peer sets it. So we had discussion just during last two hours in a spring working group with the shredder.

G

What mike and she agreed that when we have the pier we're set said we need to. Basically we we have the local information that the the node that is advertising this pair we're set set is the same entity. It's the node. So the we don't have to repeat the local IDs, local, our ID and the local, yes in in and being coding again and again. So what here is done is that is through defining these sub TLV, which which are actually encoding same way as as previously encoded for the pair agency or as pair node.

G

Those can be carried for all the member that are part of this set, but these member may be in a different s. So this is where this encoding can take care of that, because there is no one pair digit. All the peer doesn't have to be in the same in the same remote. Yes, so that is that is addressed here as well, so the encoding straightforward and is almost similar, always close to shelter draft then second or another sip type that we were addressing was the bindings it so bindings.

G

It represents some interesting problem because or or interesting question which happen to be similar to question that you will face. If you have EPC, then you go to a router which is non MPLS, so we made sweep and then here is solution by the same. But the portion is that if you want to validate the path the binding said, and if you said packet with the binding seat, then you're going to get switched it's a tunnel and the end point.

G

The LSP will appear as IP at the node, which is a last set of the binding set. So here we try to address that a way that we can get to the head end and just try to validate from that. Heaven that do you have this binding set and the bindings is identified by had an address color and the end point.

G

So basically the procedure wise. We say that if you just want to validate the binding said- and you don't want the packet to be switched to the egress of the binding set, then you can set it a loved one, and this is something is: is there in the base? Rsc's are loud is used so that packet doesn't get switched the inner mode label packets set to one is mentioned, but what is not mentioned is that the fact validation, so in this procedure wide.

G

What we'll be saying is that when node receives this binding said and TTL is set to one, they very clearly validate that that do I have again this label or binding said: do I, have a corresponding control, pin state that matching my head and outer ID.

G

My the color and the end point in the fact: that's what is the rasa sayang and then we can do the same thing when we know that the neighbor of EPE note, if the neighbor of the EP in order the IP neighbor or doesn't have route to the source, then we can terminate that on the polarity madhawk by setting the TTL value to one and the fact validation would just be basically checking that if that particular set is a local set any that I.

G

So with that would like to get your feedback, we actually was not sure whether this work will happen in spring or or in MPLS working group, and this writes some it for the spring since right, so this would be really up to the chairs to decide where they help the way the work we go, and the second point is that we are talking with each other on on the other draft on calibrating, so I think at the moment may be best tested. We get the comments on the list and then asked shares what they.

G

What is their opinion?

G

That's a given that we agree. If we agree, would you go ahead and adopt the work or any like any guidance? If you can provide.

A

So can I ask who s register aft.

C

Almost a handful, the I have a couple of question: can you go back to the first definition of the theory.

I

First, one: well, anyone would actually do, but this is the prefix. If this.

G

Is the ignition key.

C

To you have four or six in octaves or.

G

We go to the interface that you have with the and then the interface that you would have with your VDP egress fear could be a ipv4 interface or ipv6 I.

C

Understand how do you know which one it is I hope? The answer is obvious. I.

G

Mean that's something that the center? If sender knows, then it puts therefore the note said and it if it is a know, then it put the I flack that I don't know the address so because this is a I mean really. The FAQ is really pure node. So it should go to this. To the correct node I mean from a fact definition point of view. It needs to land on the correct. So that's where the ie flag will be set. The.

C

I'll flood flag indicates whether it's in for octet or 6. No, no.

G

The I'll flag says that whether it is to be ignored or to be considered but I think your point is different. You are saying that how do I know that whether this is the 4 or or 6, and that could be done by the LAN or we can put something exquisite hide assembly EFT.

G

Type at a f-type would say: yeah.

C

You had a may in one of the slides and you go tada.

I

G

I

C

So we have no I unpack it and we say it may not be leaked out of the Duma domain. I thought that there is a kind of a general that the way impact it must not be leaked out of the oral domain. Yeah I would like you to consider it. What you should have here a insured or must the.

G

Sanguine s could contain.

C

No further questions for me. Anyone else so.

C

We should coordinate with the spring chairs and see where this work goes. So I have a feeling, I know where it will go, but we should make a decision on that and.

I

Okay, thank you. Thank you. Thank.

C

I

C

J

Hello, everyone, children from Hawaii I present about if he and the SVG protocol extension for Pia.

J

Okay, the two draft has been updated to zero one way she addressed the comments received from ITF Oh 1:01 about MVP and also allow the comments be a working group. They say the Leo defined the FEC. We all needed to overlay change, so I will also have some have some considerations on this.

K

J

Okay, make before break it's a strong requirement for Marcus.

J

The k2 game, the MVP in my cast- we can we can review about this.

J

We need a very.

J

Konami clean changing on there on the incoming interface, for example, post P, my mom and I'm LDP use this technology to support MDF, but idea PHP based appear. We have revealed about this, it's difficult. It may be difficult to support and BB, because PR, don't don't use RPF or upstream check mechanism and one pfr, it's a responsible for staring at many downstream, be fer, and the different 9 cards may be responsible for different, downstream behavior.

J

So it's it may be difficult to tow and one short change and through different language, but it is seem to be a fairly easy for P, 2 and P LS v based appear because the already defined the mechanism, it's useful for building P 2 and P LS. He justa with eco extension, always be away the create who the am/pm he MVP is that a router allocated different levels for different absolutely interfaces to the same pmpf. You see. ah So the MPP seem to be a significant benefit of this solution and also one possible impact.

J

A change of forward impede mask need to be knowing through the path to the root, so the convergence may be slower than the normal Peter MPSP.

J

Another probe problem- this is from be a walking group. They say we defined a lil FEC, so we have to change our PDPA, a median PTP American protocol, because a little Leo type of FEC you can say in the future PTP as I PMS Iowa as p.m. SAT route needed to carry the FEC. If we define value. Obviously, then we'll need to change at the bill attribute in P DP.

J

So how about using existing PDF? You see I think it's possible and the here. When is the message, the message, for example, T centered the met him message to see with the label 400 and also a Pho, a with label 401 this label is for is for Pia set set at 0 and the label 484. The inclusive are inclusive P to impede hollow, also F sender, to see years a little P and the season that would be business way.

J

We can say this will build totally for P 2 and P 3 1. It's for the inclusive, would P 2 and P and well that other sway is for every PS set. For example, PF RIT one is the UNICEF 0pf RIT.

J

257 is the set to use instead of one and the PF RIT.

J

513, if you set a set to, then we have a 3 P, 2, ampere OSP for them separately.

J

This is the Peter. He pays the PFOA team and we can say that the special is is is this: the label is pulse P, 2 and P and P a label. So this can be changed. Hoho just add the p2 and p4 deal and another important it's the bitstream carried in the PA header can be unchangeable also can be changed, but this capability of unchanging, the bitstream it's important to bypass. If this loader don't have the ability to pass the period appeals dream, then it can bypass this packet to the next to the downstream router.

J

So this is the forwarding and the encapsulation summary. We believe that it's a simple way to introduce peer in the current PTM heat deployment, for at least the following reasons. Why is make before break? I TPPA may be very difficult to support and the 2nd 8th March is peer deployment, because if we want to deploy peer, mr tas, then we need to change our SPF eius eius phobia, while I'm out. If he it's the protocol, independent and the recursive AVC high easily reaches the root. I p occurs in any area.

J

Yes and the secondary is the papacy, so many effort has been throwing into the bypassing of some peer, incapable, daughters and turns out to be very complex and the side effect. But this simple way only needed to extend the MLT p and the way leverages are already supported.

J

The PDMP fall ready, so, let's step where we or updates the amount of GPA expansion using the original PDMP AVC instead of a lil, a few CPAP and updated the beauty of a Academy and also way, we plan to add the PLT of a used with recursive FEC for building in Terraria, p2 and PSP with beer.

J

J

L

Hi torik Francisco, so my question is: we are signalling AP to MPLS fee for every tree here, right, yeah, I, think this is what you're presenting it's not clear: how or what how this compares to the existing v2 MPLS fees. Signaling. Is it better or is it equal? What are the advantages you're bringing in that's one and the other thing is I sent an email or on the alias and I think you responded, but it wasn't clear your response.

L

The architectures there's states that we shouldn't creating states, control, plane states when we are following this multicast tree in the core and you are creating States in this case. Yes,.

D

L

Not sure what how is it not violating or violating that assumption.

J

Okay, thank you. The first vesicles hear about compare corporation of this solution and the this resolution I think the LePage have for some summarized this this point and for the second question about for the second question about whether this solution violate the pier architecture, because there is some describe a description in beer or say or Beatrice that piya doesn't need any to a building protocol, but but it it is not a constraint I'm here.

J

It is not a submission that pierre-pierre want to appear Peter I'm here or something also there is the text in the pier architecture. Rc said it aloud ran pier over some mark has the specific tree. The question is only that the magic has the specific tree it's built by about, if he or by ITP I did here, of course, it's possible, but ml DP. It's also I existing existing protocol that easy to be reused.

J

So summary I, don't think it valid any of the pier architecture.

E

Tony P juniper, so architectural site, human rationalizations, are wonderful mechanism. This goes pretty far. One of the major reasons that beer happened was that we wanted to get pin and ml DP out of the network's. Well, customers wanted. So if you now introduce an OTP to run beer to get the rate of ml DP, you ended up in a pretty indefensible spot. But you know good luck.

I

C

You any other questions.

A

C

Last but not least, curated and kiryat is actually talking about a set of drafts that is not entirely within the ambulance working group on resilient MPLS rings. The reason I asked created to do this is that we will need to do some type of coordination. Work in the working group chairs.

C

Okay, yeah, so yeah there are armored grafts have now been spread over a few working groups. The base draft is in MPLS, there's some other drafts in MPLS, where there some drafts in T's, and there are some drafts in there's one draft in spring, and these are also a draft. That's missing that I'll talk about at the end, but these are the things that we want to make sure we have the right allocation and that that's why we want to keep them and LU at the back of the room is smiling away so whoops.

C

So we have the base and that's almost ready for a working group. Last call there were some comments on the list that have been addressed.

C

There is one small change that I want to make here, which is the set of supported signaling protocols, for this has right now, LDP and rsvp-te, it doesn't have, is a place for spring I understand the spring is not a signaling protocol, but it's a way of creating LSPs, and we want to add that to this before we progress this to working group last well, the LDP extension graft is pretty much ready and there were several comments from the list and I'm not going to go through the idea. I mean.

C

Hopefully you know what that is, but the basic idea for all of these drafts is that you create these two counter-rotating recipes and then you use the other direction in case the direction that you're going in doesn't work. So the LDP draft became a working group document in july. All the MPs are key. Comments have been addressed, the diagrams have been connected corrected and so there's a new version out, and so the next steps I can draw. Others is that this already for working group last fall. We believe this is the right place for this.

C

This is the only place where we do LD be so that at least is easy.

C

There's a couple of drafts in peas, the arm our RSVP extension and that's a working of document and keys. There is Jeffrey's document which is proposal for how to do point-to-multipoint with RM are using our recipe signaling. We believe that that should be in keys.

C

We're open to you know, arm wrestling between the chairs, but that's where we think it should be comments and not really. A comment is just what we lack is just a decision: okay and yeah I think kind of it. We will give you a good, tease, you're, okay, with that power. Okay, thank you! So let the notes, but the minute state that power is fine with that.

C

Then there's the Audemars graphs that are individual documents, the Jeffries draft on how the multi-cut should work in general, zero zero version was presented last right. If there's some feedback, and they still have to be incorporated, so they will be in the next version at which point and the author's was request that it becomes a working document so right now it's individual and that stays in MPLS.

C

So do you see any reason to actually make this a working group document before we go to work and applause call with the LDP draft they're independent? This is multicast, lgp is unicast and then of course, the last draft that we have in the system is very new drafts. That's in spring, it was presented for the first time. It was unfortunately a little hastily written, but so there's definitely a lot of work. That needs to be done. We think that it should belong in spring. But again you know we have bruno here. We have Rob.

C

Everyone's pointing in the other direction- that's fine! So do we think this should be in spring.

C

I'm not clear yet but I think we should discuss it between the shares. Okay, um it's it's not I haven't, don't have a strong feeling either way.

H

We know the comes in co-chair sauce craft has been presented this morning, so it's very early, no discussion of the draft, so it's very very early the graphic and inventory change. It's not great.

C

H

I, don't think we are at the time of working production, but so far I don't see any impetus, extension and I think it's Menace premium. Yeah.

C

In fact, I should be honest. The reason I don't have an opinion. Yet I haven't read the draft yet and it was not in the spring work group this morning. So I need time to read and then I probably will save spring. But okay. So let's postpone the decision for the oh one. Draft 0:01 version I'll keep the name as spring for now I'm happy to change it. After that, the old one version will probably have some changes in terms of new seed types, in which case it does become a spring draft yeah, yeah. Okay.

C

So that's the current status of the drafts that we have in the system. There's one draft that should be in the system that isn't, which is a draft in lake state routing that says what's exact here, we formats and procedures so that you can, you know, define a ring and have everyone agree on what the direction what's clockwise and anti-clockwise and so on.

C

So the general format of this is is in the base draft in MPLS, with the specifics for ISS and OSPF should be, should be documented and, of course they need to throw things at me and maybe at MPLS. So it's a question. I mean that draft needs to be there. What should we do in terms of timing? Do we want that draft to progress simultaneously with the base draft?

C

Does the base draft? Can it point to this as informative?

C

So my answer to that question- and there might be other answers from other working group shares, but the thing is I: don't want to lock step okay.

I

No requirement.

C

But I want them to progress as close to.

I

C

Possible or reasonable, maybe yeah.

I

C

Yeah, so um there's not a whole lot to do there except to write it down, but yes before the next IETF right. So that's basically the set of documents. So we have the base document in MPLS and a few other argument documents in MPLS a couple of documents in T's, one and spring, and then one analysis.

C

Okay, questions to hear Utley people are hungry, okay, yeah.

I

C

Go around for food as soon as you close me, yeah I have one thing: one more thing: the aura or LDP extensions are, as curator said, almost ready for working group last call I. Think it's a little bit premature. I would like to see some more discussion on that. So please go home. Read the draft comment on the list, and hopefully we can do for a working group last call after draw yeah.

C

Okay, I will also pop some torch to start the conversation yeah okay and Dan I'm happy you're shaking your head, because I don't want negative comments on the pathways.

C

Okay, so and then for the Audemars. I will do one more. The base drops I'll do one more version and then we should be ready to go for that that one as well. Okay, thank you. Thank you. Yeah any comments on this from the spring and teas working group shares or something we missed.

C

So Pawan shakings, head I, don't see Bruno yogi, oh you're, okay with us as this okay I think this was the last presentation for today and for the mpls in in Bangkok, see you all in Prague safety of settlement, safe trips to Prague and have a good time and review our drafts. Thank you.

C

I have one more thing: blue sheets: can you pass them forward.

D