Internet Engineering Task Force Interim Meetings, 7 Sep 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: IETF-MPLS-20230907-1400

Description

MPLS interim meeting session
2023/09/07 1400

A

A

Hello, everyone can you hear me.

B

Hi Loa, yes, I can hear you.

B

Yeah I I said I can hear you I didn't get a response. Can you hear me I.

A

Can hear you now but I didn't.

B

A

Anything before.

B

A

Must be it's a little bit slow, starting off up, I! Think, kids, I, don't know why exactly.

B

Right um but I can hear you well now one's about.

C

B

Good once it starts rolling, it's okay.

C

B

Yes, I can hear you. Oh.

A

Okay, I know that um Nick is in Vietnam.

A

There is a oh, which one over there, one of the right, not right, I, don't know which organization it was, but in one of the internet registers has a meeting there.

B

Oh okay, um yeah that we have 10 attendees at the moment, uh but I'm expecting more to join Maybe.

B

Okay, so we're six minutes after.

A

Oh yes, checking one more see if I get the response.

A

A

It doesn't work, I think we're ready to go. We are 11 now.

D

B

All right, let's, uh let's progress, let's proceed! I guess then uh welcome everyone. I was attending uh to this uh interim. This is a uh part of a set of series of interims we've been having on um mpls Network actions. uh It's the usual collaboration between the three different working groups.

B

This is the note 12 I'm, presenting it to make sure that everyone is aware about the rules uh of contributing to iedf.

B

uh Some useful links you might find on this slide, including the action items we're recording and tracking for uh for these interims I've, added that on one of the bullets there. um The rest is.

C

B

Pointers for different material.

B

Today's agenda is is short: we we will review the action items that we are tracking and then we'll move on to present an update on draft mpls m, a header and that will be by request and Jag. And, lastly, we have uh any other business as usual.

B

And this is the time to oh I. Have someone okay, Tony go ahead.

E

Yes, Mr chairman I would like to request a changing the agenda. There has been a momentous event and we should not let it pass unnoticed. Well, we should spend some time to wish Mr Stewart Brandt a happy birthday.

B

Anticipation for but nevertheless happy birthday, Steward I'm.

A

Not just saying happy birthday thank.

F

A

Are you younger or older than I am.

B

A

F

In physical years, I'm 71., oh yeah, but.

G

A

Own them, you are anything.

B

Okay, um yeah I mean we could throw a party or but let's uh defer that till we meet face to face um and then I can move on to the action items that we have. Hopefully we can update them. We have something to update. uh Let me log in so that, oh yes, asking me to log, and so one moment.

B

Sorry um it'll be up in a minute.

B

Okay, so we have two action items that I have and that I'm tracking. The first is uh in potential meeting between Tony and Matthew to discuss the progress on the requirements and framework draft any outstanding comments.

B

Do we have any update on that meeting if it ever happened,.

E

That meeting did happen and uh we did have the discussion and I think all of the changes are already in the framework draft. As of the latest version.

D

Yeah I agree with that yeah we had the maintain I think uh it's all looking good.

D

A

Close the action item.

B

Yeah I'm going to record the update, and then we can talk about closing it. So you meant we have two drafts in fact, so the requirements and the framework and the changes were made in both.

E

The changes that I think we discussed were all in the framework draft.

H

D

Yeah we went we weren't addressing changes to the to the requirements, so it was just um some things that came up during the routing area directory review of the um requirements, draft that that probably impacted the framework.

B

Okay, um I can move this action item to closed now unless someone objects.

B

All right all right. The next action item we have is in fact on the agenda. uh There were some suggested texts, um Joel and others have contributed to it and I. Think Jags and Rakesh have an update for us so, um depending on how discussions today go I might update this and scrub it and as well me close it if everything is sorted out.

B

So that was the last action item. I have on record um I'm happy to track another one. If anyone has one now or as we progress in on the call.

B

Okay, just saved update should be visible if anything anybody's looking at it and I, don't see anyone in the queue. So we can move on to the next uh item on the agenda and with this I will switch to the slides and not sure as a trackish or Jag will be doing the flipping of slides, so that I can give you the privilege.

B

I

I'll be looking at uh track. Okay,.

B

So I'll give you a slight control. I think you should have it now.

D

I

Yeah I have a control thanks, hey good morning, good evening, good afternoon everybody. um So uh we got a few comments from the working group and then from the uh our co-authors.

I

um So here actually I'm gonna um present uh the updates uh of those comments and- uh and how did we address in our newly published craft um yeah.

D

So the overall actually.

I

Like we took all the commands and then we tried to uh accommodate and then see a valid uh and then validity, and then we added to our draft. So this is our agenda, so these are the main main comments recorded from the working group and the our co-authors.

I

uh So the first one is the you know like how to handle the uh label stack, push um or hph Mna Network action Push by a mid node on a received packet with a hbh, M, A subtract and the next one is I'll: go in detail, one by one I'll, just flash this one first um and then now like we have some editorial commands we got and then we have updated it and the next one is the pending that that is one pending command.

I

Like probably we can talk today and then we can address it um and the final is the next step. So let me go into the first uh First Command, what we addressed it. So uh thanks uh Joe and others.

I

You know like uh um help us in putting together the um the words and sentences together and then form a very, very good uh way of presenting how uh what to do when, when we receive a packet with the HP h m, a sub stack and if the mid node wants to add any additional hph Network action. So this this is the um the text we added to our new draft. uh So I'll just explain. You know like uh what that mean.

I

So like, um if you say like I, receive a packet on the on the left, hand, side right uh and then, if you want to add a additional uh additional labels uh on on top of it. So the idea is that you know like uh we want to duplicate the um uh the the HP h m, a uh depending on the rld on the downstream path, so that uh the downstream nodes could uh have access to the hbh MN substance. So.

J

I

Of the example, another example is like, uh um if you, uh if you want to add an additional m, a m, a uh that is m m position, which is hph, then actually like. uh We need to take the uh the existing M A up codes and then up and uh the uh the new opcode, uh which has been which the mid node is going to to impose on.

I

So these are the uh two examples which we covered in sound dot. uh One I'll go to the next any questions.

I

Okay, I'll move to the next one. uh So uh then there was some discussion about you know like um this uh Arabic. um So in this document, actually we say that uh this must be transmitted as zero and uh ignored upon uh receipt, um but um in one of our post attack uh PST uh draft, uh we are trying to reuse this orbit.

I

So there was a discussion like how to uh word this specific orbit and reserve it. um So any any questions or anything on this one.

I

Okay, I will leave it as yeah any.

B

Question yes, Jack uh one question, so the reserved bits are three right.

I

uh It's uh two three, four five for.

B

For what about oh, okay, so you're, you're setting okay, so you're, basically saying that uh the okay you're setting them to zero them must be zero and in case we want to uh use one of them in the future. uh We'll have to update this trap uh because okay.

I

Okay, let's go to the next one: okay, no problem, okay! So next one is the uh then. Actually we had some discussion about the uh readable uh label. That definition um and then we agreed upon uh to use the uh rld instead of uh maximum label stack depth um and then so. This is the the Bold whatever the Bold is there, so that there is a new uh text. We added to uh uh describe the rld information.

I

Okay, any questions this one: okay, let's move to the next one okay. um So this is a there's. Also a nice comment given by Greg uh yeah. We also agree that you know we need to add something with respect to the maximum set depth like rld, and then we added in the section 9.1.

I

G

I

This one, okay, yes Tony,.

I

Yeah yeah Tony, you can talk.

I

Any questions Tony.

E

Can you hear me now, yeah I could hear you. Okay, sorry meet Echo is failing to turn on. um Why is this described in terms of MSD and not rld?.

I

The MSDS and the inpution case right uh the rld uh in the mid notes, actually like we were talking about foreign.

I

Position- and this is actually like here actually in the head mode, uh impotion case.

E

Okay, I, don't get it, but okay.

F

The architecturally we're dealing with it in those terms, aren't we.

F

Sorry did I trample someone.

I

Yeah, you can talk, go ahead.

H

C

H

It's a bit rld and MSD, but um we need to respect, but MSD is used quite a bit in path computation. So the comment was: what does it mean for the MSD uh when the path is computed? So if there is a limit, uh then- and we know we're going to add M and a sub stack of four lses or something like that, then uh you know part computation uh public should know that um it cannot. uh You know if the limit is 10 and this sub stack is 4.

H

Then path computation can only go up to six uh LLCs and things like that. So there's already this existing term and limitation. So our tax comment was to say what does it mean? So this is the text for it.

I

I

If you don't have any questions.

F

Confused about this, so you have to worry about the maximum number a node can impose, which is what this is about, but you also have to worry about the stack deck that something is going to be able to read later on. Don't you.

H

F

So I I I can't I I, haven't read this yet this for a while I'm, certainly haven't read this new text, but I am worried that uh we have two concepts and we need to run both of them in order to make sure that everything works.

I

So this is for the uh path competition right like uh uh whether uh we can impose on the head end or not. uh That's a hidden capability right.

I

Another one is the. You know like whether the mid notes can read, read it. This is the push and then that's another. One is a read.

F

Right and you need to make sure that you don't push so many that the read doesn't work um yeah, but.

I

Just need it yeah.

F

I

You can go ahead, yeah.

B

uh I, don't know if you can hear me, but I'm waiting a little bit. uh You can hear me okay, so normally this MSD is passed to the computation engine as you're, highlighting so that you know a a path with uh so many segments can be considered. uh You know imposable on the packet or not.

B

If we're going to encode some Mna uh lses. Are you going to decrement? So let's say the Ingress can push five labels, but you want uh two labels for uh M A, the then the MSD will be modified.

I

So whatever yeah.

B

I

Msd will be used to determine whether I can add the m a stack or not to the existing.

B

Ah That's okay, that is uh I'm thinking about it differently. So what I'm thinking is, you will add the m a and then decrement MSD, so that the computation you know so, which one happens? First, I guess uh the computation of the path or the addition of m a uh that is what I'm thinking.

I

uh That depends on the local uh policy, but uh uh when, when we compute the computation right, so uh we we get the MSD right so uh but uh uh the.

B

Mst, yes, if you did the computation first, then you will have the number of segments on the MSD right.

I

Yes correct, so so, if you say once once we get the number of uh sets, we need to impose uh then actually like uh um decrementing uh the actual path computed and then see like how much uh um uh uh spare, since you can impose with uh using our m a sub stack.

H

I think the draft just says that MSD needs to include the m, a substance uh right so and now how it is done is uh it's a local thing? Maybe there is additional avoiding needed, but it just means that you could stand and energy is poor, then now path. Competition can only go up to six, but.

B

Yeah yeah, that's exactly the point that I was the MSD is negotiated. uh Let's say you're talking with a pce that MSD you provide from the Ingress as a static value. You say I, you know, I can push 10 labels, but then there are some packets that have M A some packets, don't have I mean or one tunnel has a mini. Another tunnel doesn't have a mini.

B

How do we deal with this I? Guess.

H

Yes, we just need to include them in a substance. Basically, so if you stand then uh in a minute is four: then it needs to can include support. So MSD becomes six. The real.

I

No MSD is constant, actually well.

B

Msc is about segment, that's not about label and position. So, let's.

K

Be specifically, the labels are segments for this purpose. Sorry,.

D

K

I'm behind in the queue, but this is very straightforward.

K

If I may Rakesh yeah.

H

K

You have the cast computation thing: whatever it is, gets the MSD it then computes a label stock. It's going to ascend that label stack, includes M, A things or doesn't include M A things it doesn't matter. It must comply with the MSD for the thing with to which it is sending it that has to include all of the entries in the label stack the fact that we call it segments in one case, and labels in one case is absolutely irrelevant. The meaning is very, very clear and very simple.

K

In fact, the only thing that should be said in past computation about MSD is path. Computation entities should continue to comply with MSD restrictions.

K

There is nothing different about Mna entries, they're part of the label stack and MSD. Even if it says segments is about how many label stack entries, can you push that's what it is.

H

Yeah I think we are saying the same thing, but does the text reflect uh Joel the what you just mentioned.

C

K

C

J

K

Think the text here is fine. It belongs in a section on path, computation engines, not encapsulating node responsibilities, but the text is fine in that regard. The rld issue is a separate issue that should also be mentioned, as somebody else said,.

H

Yeah, yes, so the Saxon 7 uh Steward that mentions the RL deep, and this one mentioned MSD I, wonder if there is um some refactoring needs to be done, but uh okay I think we are on the same page thanks.

I

C

I

I

H

I

A

You can hear me now.

H

A

So, between Russia and Jewel, I think I agree with what they said. The way I think of it is that already it's a property for a node and the when you set up an LSP. You have to check that the number of labels that you want to push actually fits within the rld for each and every node on the on the LSP, and then we need to say the same thing for the maximum speed, depth and I think we did so I think I think we're okay.

A

F

So I I think encapsulating node responsibilities really needs to refer to both. Doesn't it because it's your responsibility as the encapsulating node to make sure you um you, you don't push more than you can actually push and that everything that needs to be accessible at every point is accessible at every point. So it's not just MSD. You worry about there. It's it's, presumably both of them.

F

That's right, so, presumably the text will reflect this.

H

So I think that split into two different sections, and maybe we need to put them together or do something else. You.

F

Certainly need to at least reference it because it is definitely an encapsulating those responsibility to comply with both of these Network constraints.

F

I

H

I

Section here to say that it should be complaining with both MSD and rld.

I

Directly can go right now.

B

Okay, um uh so back to the you know: negotiating the the capability of Ingress with the computation engine. So today we passed this maximum segment depth to the computation engine. Are we going to pass a new attribute called m a sub stack size because that sub stack size is a variable. uh You know, depending on the features you're going to carry or enable uh to the computation engine.

B

So that's question number one um I'm, not sure of the last statement there that says the MSD needs to include the Mna sub stack to be added, um not sure why segments maximum segments need to include something else, and maybe it's just uh parsing it I'm finding it hard.

I

Yeah the maximum set depth is, uh is a total known capability, so it should be both whatever said. Communication is abstract.

B

The maximum segment depth is the maximum I can push right. A maximum number of segments I can push. Why is it related to the m? A sub stack size.

I

Because there is a notes capability right, like uh how much node can push if a node cannot push the m a sub stack information, then.

B

Yes, that's what I said you can decrement? If you have a mini sub stack, you can decrement the MSD. So let's say you had without M A you had 10 MSDS, 10 and let's say Mna sub stack is four.

B

uh When you enable that m a sub stack, then MSD will be six, but then Joel made a comment and I'm still not I'm confused about.

K

It so may I may I reply Carrick.

G

K

I from your last question, I think I figured out where the disconnect is the Assumption I am making. Is that when a path computation entity either in the head end node or separate it doesn't matter, compute a label stack to be imposed.

K

It includes the m, a sub stack in that. Therefore it knows how big that m. A sub stack, is and makes sure that the total stack it's delivering, which includes the m. A sub stack meets the MSD requirement.

K

If you instead assume that the path computation engine is going to deliver a path and then the head end is going to separately. Decide. Oh I want to add some actions to it. Life gets really complicated because, as you say, substack compliance becomes impossible. Msd compliance becomes almost impossible because you need a piece of information you're going to compute after getting the path that you must have in order to compute the path. Well, you can't do that. It doesn't work tarek.

K

That's why I said the pce is Computing, the the stack it computes the whole stack to be imposed. That's what it does that's going to include any actions when it includes the actions it makes sure it meets the MSD, as well as the rld restrictions of the past. Yes, that assumes that the pce, whether it's in the head, end, node or separate, knows the rld as well. Somehow it has to otherwise it can't compute the stack it's it's created.

K

If you try to do this in any other order, you're going to produce a mess because the MSD becomes variable based on an unknown quantity and that just doesn't work.

K

B

Can I try to clarify Joel, uh sorry, I'm, not sure. If you can hear me, don't go ahead. You go ahead. Okay, um I meant to me it's simple, I mean with normally MSD is 10, but I want to set aside four labels for m a so I will just modify the MSD to be six. Why.

K

Would you but, but you put as a head end, you don't know which actions you're going to put on which paths you don't set aside. Four labels for m a as a head end because you're not putting M A on every note on.

B

Everything it's a on purpose. You can pass. You know specific tabs.

K

If you know it per path you can make, then the pce can know it. No.

B

No I mean I mean before Computing the path on path, request on path, request, yeah,.

K

Well, that would be a big change to pce, because I.

B

Don't know today, I can pull it up. You can pass a limitation on number segments so.

K

The fundamental disconnect here is who is responsible for deciding that an act, Ive sub stack put in this in the in this label? Stack, that's being computed, I had assumed that the pce was responsible for it because well, it knows what it's Computing and what one might want to do, and it knows all the policies and it's it's the policy engine.

K

It knows what the policies are you're, assuming that the head end will want to insert mnas after the pce does its job based just on knowing the path well, frankly, I don't I that doesn't match any deployment model. I can understand.

B

Misunderstood what I'm asking Joel I am with you that um you compute the path after knowing that I want to insert a mini. Well, not you yeah! So.

F

Don't you res I mean there are operational cases where you might want to put in a in or not put the Mna in, depending on what you're trying to to do so you might want the path with and without the m a there and you don't want to keep Computing paths.

F

Surely, um but I think what that means is that um you know a model is that the head end reserves the right to um put in a certain number of labels for its operational purposes, and it lies about what the MSD capability is, so that the pce stays within its constraint. But I think all of this. We really need to be sort of out of the weeds of how various implementers have implemented various different flavors of BCE and describe in architectural terms. Don't we.

K

Fair enough Stuart yeah. If it is true that that a head end could lie and then the things would happen, and it could do what it wanted, and it is true that, if that you could get some pretty strange results, if both the pce and the head end are trying to add M A's, but so be it.

F

Well, well, I think we need a uh some agreements about how this how this works, or at least the two entities in a particular instance, need an agreement about how it works.

K

Well, I'm not sure they do even.

F

um Oh well, I mean you could have a I suppose a parameter that goes between them. That says whether the head end has the right to push in to put Mna put M A in well.

K

In fact, from what you said before, there's nothing, the pce can do to stop the head end from doing so and there's nothing this the head end can do to stop the pce from doing so and if they both do so, they both do so.

K

F

Indeed, I use the term lying I mean it's really a question of allocation. Isn't it the.

C

F

Hardware can allocate a certain fraction of its MSD to the pce.

K

And and the normal case I would expect would be that it allocates all of it, but that's we don't have to mandate that normal can be different. Implementers want, indeed,.

I

In today's world, right so uh who, who is going to insert the uh entropy labels? uh Is it a pce or the local device.

K

I don't know what current nodes do. I had always assumed the pce did, because otherwise you get a mess.

I

Yeah, okay, so it.

K

C

I

Same as that, one, that's what I'm trying to come right, yeah, but.

B

I agree, but El is inserted by the Ingress, uh uh but I don't think today we pass it to the pce as okay. We need to double check that there is no indication that I'm enabling the Eli, uh so it might be a mess so.

K

How does if, if the head ends are adding Eli? How does that currently interact with MSD for pces.

B

Right, I, I'm, I'm thinking it's not somehow passed uh to pce. We need to double check that we.

K

G

We should try yeah.

C

I

Okay, so can we move the next one.

I

Okay, I'll move the next one okay. So there was a comment you know like in our 7.1 right. uh We have. We have said that uh um the new network action added must not conflict with the Network action and they received uh Nas or Nas with a HPS scoping right. So the question was like: uh how do how does a node find the uh conflict?

I

So uh actually uh our thought was that now like um this would be specific to network action definition so yeah we can talk on this. One.

I

So any uh clarification someone can give or.

K

um There is a small sorry should raise my hand first. There.

I

K

Problem here, as written you've left open the.

K

The case of well what if the pushing node, doesn't understand some action but I I believe that the implica, the assumption is that no node will ever be pushing an action that it doesn't understand and therefore there can never be a conflict of that type. You might want to add just a sentence to to.

C

K

Out so people understand there isn't a gap.

I

Okay, so we can say that um the node, which is pushing any M A substance, should know about the network actions that is well.

K

The thing is, what it should know about is the network actions it is pushing. It doesn't, have a conflict with unknown Network actions.

K

Because resolving the conflict of unknown Network actions in accordance with the behavior of the action would be impossible, but fortunately the only conflicts are with actions. It is pushing.

K

Now, if we Define two actions to conflict, we've made a mess for ourselves and we deserve what we get yeah.

I

So when you say conflict right action, one and action two, so the conflict should be defined in the specific Network action definition right when they come up with yep. Okay,.

I

H

C

I

A

To me, yeah I think I I'm still confused about that I think I got what we're trying to do, but the text Joan. Can you help us with the text and see that we get the Texas is correct. It should now not be much Tech, just the just the sentence or so, but.

K

Joel yeah I was waiting for the microphone to turn on. I will be happy to risk to off list correspond with the authors to see about Minor cleanups to the text. To make sure it's right, they've got the right goal, we'll we'll clean up the text.

A

Okay, thank you. Thank.

I

H

uh June, maybe I didn't understand, because the text says that M A capable node must understand the received Network accents. Are you saying it also must understand the network accents it's uh pushing uh and it's not pushing unknown Network actions um so.

K

I I, as part of what I'm going to reword with the with Jags and Company. It doesn't actually matter if it understands the received Network actions. What matters is, if there's a conflict between an action, it is pushing and they receive Network action, and it must understand its Network. The network actions. It is pushing enough to resolve such conflicts.

K

Okay, that seems to be a necessary condition if we're going to leave it to the node. That's pushing.

H

Okay, but it's.

K

I, don't really like it I just don't know any better answer.

H

Yeah just that it will depend on the network action itself um and every time we Define the network action I think we reject that's what Jack said that maybe it should say that this network cannot coexist with that Network action and things like that, because then it will conflict. um If.

K

We have oh well, we can run into trouble if we start having actions which are mute, different but mutually exclusive, because then we literally would be requiring something to understand, requiring a pushing node to understand something it doesn't understand, and that is not achievable.

K

Yeah I I don't like to ask nodes to do the impossible.

K

But that's how I'll work with the guys on on getting this text a little clearer so that we know what we're dealing with and the the corollary Rakesh is that if there are two actions which conflict we need to specify what happens if both are present in the stack rather than specifying that they must not be both present, and that seems to be pretty easy to do. If somebody understands only one of them, they do the one they understand. If somebody understands both of them, they do what the draft says they should do.

K

And that doesn't end up in this document at all.

I

G

uh Can you hear me.

I

Yeah I could hear.

G

You uh yeah so I here I'm also confused. What's the could be a potential confliction? uh Could it uh you provide a solid example for that, because I I I I I can see. um What's what's the real issue we got concerned about here.

K

So simple example, and by no means the only possible one, a node is planning to push an m, a sub stack with some ioam on it for tracing just maybe maybe tracing the fast reroute patch.

K

Meanwhile, it receives a sub stack and it's it's using direct export and it says to direct export some set of information to node a meanwhile. The stack it receives has a ioam direct export, saying export some different set of things to node B.

K

Does it do both? Does it do one? Does it do the other we need? Obviously, in the ioam case, I think I know what the right answer is, but there may be other parameters where the right answer isn't do both. So we need. So it's the specific. The common conflict is going to be the same occurring in the push sub stack and in the received packet.

K

There may be other cases, but that's the common case and we need to say resolving. That has to be done and it is done by the node which is pushing these and it seems reasonable to say, hey if you're pushing something that is identical to something. On the stack then identical action, then you understand it enough to figure out how to resolve that.

G

um I'm sure I've fully understand uh that, but could you uh or anybody else actually uh write it down to some text about it to um describe his case and we can better understand it.

K

Given there are very few defined actions, it is very hard to write down what the conflict, what example conflict cases are beyond the example I just gave you.

G

K

Happens to be a simple case, it happens that in that particular case, the likely resolution is just do both, but we can't just assume that for all actions.

G

Yeah I just don't worry about. If we don't really have a um real cases for this, then maybe we are talking about something doesn't exist um here. So so we must fully understand this is a common requirement. Then we should solve it right.

K

Tell you the problem is we get? We need existing nodes to do the right thing. They need to be told what they're going to have to do. I agree. We could. We could just write that I for ioam. You do both in the ioam draft, but this is a general problem. If you find two things, what do you do.

K

And if, if there are nodes which you shouldn't have, if there are actions where you shouldn't have two conflicting actions and.

G

The nodes both.

K

Of them it should sort.

G

It out yeah but for iom is just a passive merment protocol right. It's just a collect, whatever it's seen, no on the foreign I I, don't see why this will conflict with anything else.

G

You just collect the data from new parts. It's a that's a aim.

K

Of it, that's that's because you're assuming the resolution of the conflict- and it happens that that case I have a pretty good guess and I agree with you as to what the likely resolution is, but a we're not requiring that and B. We don't think that's the only possible case.

G

So that's why I'm confused I'd like to see a solid example. I still can't see that so, let's, let's make me feeling why we um conserve all this at all. If we don't have it a real example for that even.

K

I asked to put it in specifically because I considered it under specified. If we didn't cover it.

K

I'm having arguments in another working group with folks who have an unspecified Behavior because, as a general rule, I unspecified behaviors lead to interoperability failures.

I

Like I was thinking now, like some example like, if you say, um I want to impose an up code, a okay and then, if already my HPS stack, has upgrade a then actually like. We don't want to do a double encode right so that that could be one of the example. I could think of the common way.

G

uh Excuse me I, think here clear you can say again.

I

What I'm saying is that if you say that mid node wants to impose, uh uh add and uh um op code, uh some X number, okay so, but but my result, packet has the hph option with the same opcode X, then I don't need to do. I know like additional encode of my up codex, so just use the same one, which is already there encoder right.

I

So that is the conflict uh correctly. One of the example for the conflict.

G

You mean um some nodes want to add the same action to the.

D

I

You know I'm just thinking loud and all like that. That could be one of the conflict which Joel was pointing out.

G

So yeah I I want to ask Joe if this is a something you'll considered as a conflict.

K

Full description of the kind of conflict I was talking about yes and the question of. Should you be required to put in the new one and keep the old? Should you be? Should you keep the old and not bother with the new? Should you put it, keep both that's going to depend on the particular action and possibly on local policy, and we don't need to specify it as long as the the node recognizes it and deals with it ignoring. It is bad.

G

All right that that I think that's a case I can buy, um but other than that for two different actions, how they will conflict with each other. I can't see still.

K

I'm not seeing um that's not my primary concern and that we haven't run into and I'm going to work with guys to try to just leave us enough wiggle room that we can deal with that later.

C

I

Hello, you can go ahead now.

A

Okay, I have a I, think I have a problem with the word conflict, but because, if you say that you have three different parameters, a b and c- and you have uh one instruction saying: send uh A and B to one node and B and C to the others. That's not the conflict! Is it it's just something that just works.

A

And that the very first example I I can't construe the conflict out of that.

I

So lawyer, saying that um uh noticing and a and b up chord.

A

uh Now we Gilbert talked about um um direct export, sending uh different from a comments at the parameters, sending two different subsets to two different different uh um two different nodes, but that shouldn't be a problem. Should it.

A

D

I

Laura wants to answer this question regarding the iom example, which he told.

I

Joel, could you hear us obviously.

A

I

H

A

J

I

So so, for this one actually Joel is going to give some uh updated text.

A

uh Well, that's not what not what he said. He said he was going to work with you to actually get the updated exactly yeah. Well, did you hear my question.

K

No, unfortunately, I got a call from my car mechanic who was just telling.

A

Me yeah I heard the playing uh no I was asking on the first example you had. You said you were using direct export to send uh one's office common set of parameters, sending a subset of repairing it is to one management, node and another subset, but overlapping to another node. But that's I, don't cuss through that as a conflict, it would work.

K

um I in the abstract, it is a conflict, since it will look like the same action. The actual risk correct resolution is probably in that particular case. Do both and the ioam draft should say that, and it will be the response. So the natural thing will happen in that case, but it was the only one I could think of as an example of getting two things in the same, but then pushing the same action that was already in that's all I I. Think.

A

It's better you're, saying it's a conflict because it looks like it's the same action.

A

K

That's what I was saying yes, but I I will not be that specific in the wording that I work out with Jax.

A

We have G and the Q.

I

Yeah now we can hear.

J

You, okay, uh actually I, send one question to the chat window.

J

uh This is not about the conflict but more about the size of the uh yes way when a node, a Transit node as some additional actions to the NOS it may, the science may increase to uh uh to some extent that and maybe exceed the ird of constant nodes and how to deal with that uh situation. Do I need to consider to describe it in the document.

I

Okay, so you're saying that in the case of uh rld, is exceeding the the impression is exceeding the rld value. Then what what is what the node should do.

J

uh uh That pushing make the uh updated enough the size of the update in US exceeded iot, something. Yes, we add some new actions to the nuts right in this case,.

I

Yeah, so what we should do so, okay, your thing is that yeah we should address like what the node should do when the uh when it exceeds the rld.

J

I

I

uh Yeah yeah probably should be a local policy. Yeah I, let uh Tony you can.

I

Thank you, but Tony. You want to talk now or.

J

E

I really hate the way meet. Echo fails when you try to turn it on um it's not helping us.

E

um The best thing I can see to do if you run into this situation, is to take and make a earlier copy of the nas somewhere else in earlier in the label stack so that the new Nas is now well within rld.

I

J

That you have to separate, naturally the same scope like.

E

Well, remember, you know we may have done that at uh label and position time anyway right. We make copies of the nas already so that we are always Within rld.

J

um So I didn't quite catch. Your last sentence about.

E

J

C

E

When we first put a Nas on a label stack to respect rld, we may have to put in multiple copies of the nas we've discussed this before.

E

Do you understand that part.

J

Sorry, my little today is not very good, so maybe uh maybe we can put it to the chat window.

E

I'll say it again: thank.

J

E

We need to make multiple copies of the nas okay, if we're imposing the new label or a new action, and it's going to cost the nas size to increase and push outside of the rld, then we may need to keep making copies more copies of the nas inside of the rld.

E

I'll, add this to that too: bye.

J

um Yeah I understand you want to split the nurse into several small pieces, but that is not to recall replication in early discussions. Maybe that makes ishma's a different contestation actions.

J

So maybe they will call some of the nodes only and see a subset of the nurse by the actions.

J

E

I'm sorry I did not say split the Nas I said copy. You would make an entire copy of the new Nas Within rld and you might have to do this up and down the stack depending on what you're trying to do.

J

Yeah, but uh what if didn't us itself is already exceeded the ild after the elections being added?

J

E

If the nas itself alone exceeds the rld, then you're just broken.

J

K

Went to the impossible.

J

That could be possible if you have some nodes which limited IOD on the US one.

E

Then you probably don't want to be using m a.

J

Yeah, that is one special case, but maybe the text to to say something about what, if the nas Galaxy the IOD, what would uh we should do or what is our suggestion about the deployment.

J

And that it's possible.

J

Okay, thank you.

I

Hey Joe you wanna go ahead. All.

K

I was gonna say, is if you're trying to push on stuff won't be understood by the guys who need won't, be received and visible to the guys who need to process it. The only answer we can give it don't do that if you want to send proposed text to the list. Well, anybody can send propose text to the list about anything.

H

B

Dark, hey Jack, because you were on that slide of hop I hop modification for the nas yeah.

I

B

uh So we did describe this slide where you're modifying a node uh modifies the existing Hopper hop Nas by adding actions right.

I

And then adding its.

B

Own there was another slide, maybe that you um described the modification of existing Nas.

I

Not modification copying the existing Nas.

B

So if it did, you address the comment about um a Transit note trying to modify an existing Nas.

I

uh It's not a modifying of existing Nas actually, like is actually like uh copying. The existing mass in this case yeah.

B

I understand the copying, but, let's say I it Transit node, wants to add an action and the packet has a Nas a.

K

There is currently no description in here of modifying an existing Nas in a received stack, there's nothing. We can do to stop an implementation from doing so as long as it meets all the constraints, but it tends to have lots of weird side effects and constraints, and so there is currently no text to describe doing that and to capture all because capturing all of the constraints you would have to meet and knowing that the scope was right and that the implications were correct. We can't stop a node from doing it.

K

Nodes can do whatever they want, but there is currently no effort to describe that behavior in this text in this document anywhere, if you want to, if you believe it is important that we describe that I personally, think it's a bad idea, but if you think of it should be done, even though your chair you're still allowed to send proposed text to the list.

B

Okay, I um I will see what you know. I can do on that. No problem, thanks that yeah that issue in the past and I. um You know I'm fine if we agree collectively that we don't want to address it for now and and specifically the the hierarchical LSP model.

B

You know one one Ingress enables a Nas and there's another Ingress threat. You know down down there, it's a hierarchical LSP, so that second Ingress tries to add another action and I'm. You know that's the thing I was thinking about, but uh uh the other one was the frr. uh The FR is an NFR.

H

B

Confused about that, but setting that action by a Transit node, uh so let's uh I thought this should be captured, but we can talk more about it. Let me think if I need to send an email, sure.

I

Okay, yeah thanks Eric yeah, so um yeah. So this is the next step. Actually you want to talk about.

I

The early allocation of m, a label for pspl and uh are we ready for working group last call.

B

Okay, um I think for the early allocation. We will follow the policy you know for uh early allocation and think it's a working group document, and uh maybe we should do a poll of implementation as undergoing implementations, but we will, you know, there's a policy in the working group that we follow so we'll try to do that. uh Working group last call are we sure we addressed all the comments that uh are outstanding, or maybe we should wait on on this. A little bit and lower can add more on this.

A

uh There is a a process of getting up to the working group class called including a routing director at review, and we haven't done that so I would say we are on the way, but we're not there yet. So it's a couple of steps to take before we can start the workgroup.org.

G

I

A

uh Actually, what we need to consider and um would like to have comments on that, are we where what they have today and the updates we talked about in this meeting? Would we be ready to ask for a routing director at The View on this document.

J

I

Yeah go ahead: okay,.

B

It's very related to what law was asking honestly, so I'm not going to change the context. So we have an open action item uh on the draft and it's basically trying to track anything outstanding. Any comments on this and um you know I, would like to either you know there were a couple of things to follow up on, including what I remember is in the maximum segment depth with interaction with Eli, and you know, on the sub uh m, a sub stack size.

B

Maybe we should follow up on that and the text can either be left as is or be modified to be in line with what the behavior we do today with Eli um other things- and you know there were some clarity needed I. Believe someone asked, uh so are you tracking anything do I need to update the open action item? That's what I'm asking or should we are we comfortable to close it I guess.

I

uh We have uh this comment. You know like uh we need to address. uh Joe said uh he will work with us to come up with.

C

Something extra.

I

Correct, yes, and.

B

I

We are actually in this specific slide right. um We want to uh refer to 7.1 about the rld as well, so the encapsulating node should take care of the MSD, as well as the rld before uh imposing the uh uh our reminisce substance.

I

C

These are the two things.

B

I put one here: sorry, I'm, recording them. Can you repeat the context of the first, uh the first one you you.

I

Described so this one actually like uh the the the group wants to reword this one, the conflict, stuff yeah, so um Joel offered help to change this text reword. It is the first one and the second one actually like in 9.1 section the encapsulating node on top of the MSD they need. We need to add as a sentence saying that it needs to be within the rld which we have described in 7.1 section.

B

Thank you, I recorded these two anything.

I

Else, I missed. This was what actually I remember.

B

um Maybe we should watch the email list as well, but I just tracking these two now, okay and back to lower I, didn't mean to interrupt sorry.

A

uh I actually took down my hand, I think the.

A

No I think I'm done. Thank you.

B

I think you were asking Loa about RTG and being it ready for.

A

B

A

But then what I heard when you were talking about this slide, you have there and there it was 7.1. We want to update the text around that before they actually ask for.

D

H

A

Review um a routing directorate review is not that far away, but we have at least two things. We need to update.

B

A

But actually we could and I'll say we're giving those two two examples, but in general we could leave open issues in the document, even if we do a routing directorate review, because we might get an extra chance to discuss it with someone that has not been that involved in developing the document. So it could be a good thing.

B

Yeah I agree if it shouldn't block yeah I.

A

Don't know it shouldn't Vlog for a long time, but we might stop it just because we know things that we want to update.

B

Okay deal, would you like the editors to of the draft to mark the sections to be edited, uh and then we can or wait until we added you know final.

A

uh I'm I'm actually for waiting in this case. It shouldn't take that long time. Okay,.

B

Okay, anything else, uh Jack I, guess your your next steps. You presented those.

I

Yes, um these are the next steps and then thank you all.

B

Right thanks a lot for the update. uh So if I go back to the agenda, I think that was the last item before any other business.

B

It's the last chance for anyone to raise any concern they have and if we don't have any, we can adjourn for today.

B

B

Okay, I, don't see anyone else in the cube except me and Loa and.

H

A

Think we have fun.

B

Yep I think with this uh thanks. Everyone who joined and we'll see you next interim the watch out for the email updates thanks.

H

Yeah. Thank you. Bye thanks, bye-bye.