Open Telemetry Uncategorized, 9 Aug 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: 2022-08-09 meeting

Description

Instrumentation: Messaging

A

Hey morning good morning, how are you.

B

Yeah not too bad. uh I see from the agenda martin's out this week, so uh I don't know whether santosh is going to join or ram the other.

A

One I should be joining from our demon's message: yeah, but yeah, there's not much on the agenda for today.

B

A

A

Are you able to hear me now yep, I can.

C

Can you guys hear me.

A

Yeah, I'm able to hear you okay,.

B

Okay, so yeah not much on the agenda. uh Martin's out he's one of the main people here, um santosh isn't here. um I don't think we have much to talk about today.

B

Unless beneath have you got anything, you want to bring up.

A

No, I do not have anything mad, uh so it looks like matt says he's in the other zoom okay. I just need to.

A

B

Other calendar entry I've got here is for ruby this time.

C

Are you looking at the calendar to see what's going on now.

B

Yeah look at the the public calendar, so the only other, open, telemetry public calendar entry for now is um is ruby.

A

C

This realm six before um which company is he from now.

A

Really, I don't know.

C

C

A

Sanders hi yeah.

D

A

I have some issues with my zoom, so I'm just will be joining back. It's crashing for some reason. Okay, all.

D

A

You all good, how are you good good.

D

Yeah, I think, uh last week, uh ram uh sorry uh nev showed us uh a document that you wanted us to fill uh well. Yeah.

C

We can, we can chat a little bit more about it. Sorry I couldn't make it. You know it was just the you know. First, uh you know uh I chatted about it uh the day before. I think we slapped some together and then filling out a few more sections. We can check about it um and go from there.

D

Okay, um I wanted to. uh We can wait a couple of minutes. I think martin, I don't know, uh is he.

A

Martin's out he's he's.

B

Got new generation out this week, so he won't be here today and tomorrow.

D

Oh, I see I see. Okay, um guys, can you uh give us a little uh more details on what how you want to use that info like as in what is the motivation? uh I I know of a couple, but I just wanted to reiterate uh the same thing.

C

I think you know sharing the screen, you know going through the doc. You know I wrote the motivation there. Also. I think uh let me just check my screen and then we can go through that.

C

Let me know when you guys see the screen yeah.

D

I can see it here: okay,.

C

All right, so, uh essentially, what it is is uh we we all know. You know uh at least the notable vendors that have been participating in the rum signing step right. We all have our own schemas that we've been using. You know for the last, how many years or whatever it is, and we want to maintain those use cases. Essentially, you know, even when we switch to open elementary and things the concepts are common across the board.

C

I think you know this is from experience that we have had within microsoft, and we've had a lot of different sdks collecting data for a client, and we worked on unifying all these things several years ago and a similar exercise was performed.

C

We essentially went to all the sdk owners, and you know people that have been doing data collection for different pipelines and ask them just tell me: what is the data you're collecting show me the shape of the data, and we put them all slap them all together and then uh sort of patterns emerge right. We all want to collect name, we all want to collect url, basically,.

D

Mostly most of the parameters will be similar. um Maybe some few things will be different. Yeah.

C

And then part of that uh some interesting things also came out. um You know some teams found out that um hey. This is a cool thing that we should collect, could use this way that way whatever it is, that's a side benefit. I think um I I I see we are up against the similar thing. Also you're going and defining an hotel event schema is not possible without understanding what the use cases are, what data everybody's collecting in what shape and things. So the idea.

D

C

This is you know. I also read through your document and martin's document I think about a trying to create event, schemas and stuff. This just fits nicely into that. It's just an execution type stuff. You know logistics on how we can go about doing this. My idea is these steps here. So, let's first you know work on. You know uh basically contacting everybody who's interested, and you know figuring out who wants to contribute to this. um uh The you know so essentially collect a list of uh vendors that are interested in contributing their schemas.

C

um I don't see anybody would hold back unless it's completely proprietary or whatever it is, and then um event names, there's two things that we have to unify and we have to all come up with an event name for whatever the thing that we're trying to do. Example of that is in microsoft. We have something called page action.

C

A

C

Of sense for us, but it might not make sense for you so understanding what the heck is page action. Pjs is nothing but user interaction. I think somebody else wrote about it or even timings. For example, martin's coming up with event timings, we call it pageview performance and, I believe, um did I have um or even page view, I'm pretty sure some.

C

uh You know, google, uh I think calls it the impression or something like that so essentially, step one is to identify high level the event names and we can all come together and you know come to an agreement on what do we want to call this in uh open, telemetry, uh rom events moving forward key mapping and then you know double clicking on each one and then collecting actual schema uh in sort of this format. I think this has enough information for us to go.

C

Go back the name of the field, description of what it's being used for um you know, type in case you know it's different than absolute. um It could be no.

D

I I got it one one question I had, though I think I I uh understand you know uh the importance of collecting this, so you.

C

D

Basically, we are trying to unify uh and get a common schema, but um the other question I had was uh like: how do you want to use this schema once we have the schema once we have identified common set of fields.

C

What I want us to do is that we want to. I want us to go, define these rom events uh essentially call them so domain, slash whatever so you know, the assumption is that we're going to call the impression event the page view uh in um in open telemetry. That's that's. Basically the event name and we'll define these these things and essentially say this is what um any client sdk should produce for.

C

Please do and then the particular vendor they can do the mapping from this to their schema, and then you know continue with the rest of the pipeline. Their exporters would do it or their back-end could do the mapping or something like that. So it's the the the net would be the the outcome of this exercise should be a defined um hotel room events. That's basically what it is. I don't know. If uh is that that doesn't make sense, it's a pretty simple outcome and there's nothing can't be complicated about it. No.

D

No, I I, um I think, I'm still missing something so um there are like. Maybe you know, I can think of a couple ways um where the schema will be helpful. One is, of course you know in the um generation like when you generate data, you you make sure you confirm to this schema right on on the agent side and on the receiver side. um You know some of us uh like aws and uh we at cisco.

D

You know we also want to do validation of the data, because this data comes unauthenticated and we cannot trust it. uh So we wanted some uh predefined schema so that we can validate the data against it. So we could use that, for you know, for that purpose as well.

C

Okay yeah, so when you talk about validation, is a data, type validation or some sort of scrubbing or what level of validation are we talking about.

D

At this point we don't know we have to see what all is possible. uh I think the only motivation is that the rum agents- it's very hard to uh you know, give them any authentication, token yeah and therefore yeah and therefore yeah yeah correct, and therefore we have to accept data. You know, as is yes and, and uh you know there is a good chance that uh you know anybody could just send random data. So we want to do some some, you know whatever validation possible and that needs to be defined.

C

Makes sense um so the you know again talking about experience.

C

There's some validations that we can perform, so uh it's um uh ensuring that the data that's getting produced from the client side is actually a good usable. You know, structure uh confirms data type and what is one thing? I don't see us doing it in broad and in fact we have a component that does even validation similar to what you talked about it.

C

It'll say if it's main field it has to well, you know if it's a page view event, the required fields must all be present right, optional fields or whatever it is, and the size conforms to it or whatever we can even throw on some more validation. It's usually.

A

C

Type validation and required field validation. That's well! That's all we do how we do that use that is, as far as we know, there's nobody uses that event, validation in production. It's really a design time, development time thing they add the plugin. They ensure that their instrumentation is all good. When I say instrumentation, let's not confuse with open elementary instrumentation, it's really the the the collection. You know, the the thing that's producing, instrumentations uh telemetry is all valid. It's it's a part of the life cycle.

C

Once it's done, the code is all good, even validation plug-in goes out, and then you put it in product because it is expensive to do this in every client. That's sending data and then the data validation you're talking about in the back end. um The the question that we have to ask ourselves is that what is that they're worried about? Yes, it's an authenticated endpoint. What is that they're worried about the primary thing that we are? You know that we're worried about that. We know of that.

C

Our customers are worried about and stuff is usually uh data's queuing right. So you know some.

A

Sort of a ddos type.

C

Thing so somebody just spams the heck out of them and then uh that's all they could do unauthorized endpoints. They would send um you know page views. While they are not really happening, they would just simply just send pageview pageview page view from you know a fake uh uh thing or something like postman or whatever it is. They write a script and they're just spewing.

C

The event. Validation is probably not going to be super helpful there. It's. uh You know some some other way of training that you have to do. You have to look through the data. It'll. Look like a perfect data set. If you, if you take a look at somebody, some are not that you get a transcending. It could look exactly how you expect it it could conform to the schema.

C

uh I don't know how the schema could help that problem. The problem that schema could help is like, like I talked about uh it, just doesn't conform to the structure, uh data type and required fields and stuff. um Is there something else that you're worried about.

D

C

D

um No, I think it's mostly that in our case we noticed in production. We noticed uh data coming with.

D

You know invalid data, like like application names. You know it's all garbled garbage gotcha yeah yeah, so so we wanted and then you know we don't notice it, but our customers notice it and they complain right, and uh so so we at least want to do some basic fields, validation uh and uh and- and that's all like, for example, for a page view. We want to make sure that it has all the required fields, because if somebody is constructing these things manually, you know they may or may not.

D

uh I don't know all this schema correct.

C

C

D

And it also affects performance, so we want to do very. You know simple validations to begin.

A

D

And, depending on and and like you said, I think the ddos uh challenge, I don't think we can address that so.

C

D

Yeah yeah, so that we're not uh attempting to address it's purely we want to avoid garbage. That's all.

C

Got it yeah, uh I think uh so. um Are there any open elementary constructs that we have to, or maybe just before,.

D

There is one of those there is so the schema there is something uh in open. Telemetry called schema.

D

But there the scope of that schema is, is very limited. Today it is only limited uh to to version the set of attributes uh so that, if somebody, if the convention changes, let's say some field used to be called a and now it's called b, then uh they are capturing that in that schema today, so that uh when you receive you say that hey my um I can as a receiver, I can accept until version three.

C

D

If you are sending version two, I can, I know how to map it to version three so something like that is what exists today and for our requirements. You know uh that for this object, you know these are the required fields. That is not present right so that we need to uh submit the notep, and you know, enhance the spec about that.

B

But yeah, I think in terms of what rams are highlighting here, is exactly the basis of what you build. Your back end, validation on by saying, uh if you get this domain, this name, these are the things you expect which, which is what happens in our back end. Today, um it's gotta conform to the basic shape if it doesn't conform to the shape or has uh bogus data it gets dropped, which is, I think, the same concern you have.

C

uh An example, I believe is it might be wrong. I think it's ib address. uh If the client sends an ipa address, I think it comes as a string. uh You essentially just to make sure that you've got four parts. You know a separate byte, so that's kind of matching and stuff so that that's done at the at the high level. It can also get expensive. So it's it's a balance.

C

You cannot do this, for you know, um because the moment you talk about clients you're talking about millions, if not billions, of uh events coming through and the back end would have a huge scale problem. Yeah.

D

So scale is definitely uh a concern. Yeah.

C

So it's got to be simple, uh but this is this: this could capture that information and then um the way I see it, this is step one. We first need to understand what is that we want to define and then, if operativity doesn't have constructs that allow us to represent those things, then we need to get another type in and say uh we we want to. You know you know push that through or something like that.

C

But this is a larger level high level step, one we need to first come up with what the heck do. We want to capture collect and what are the things that we want to um use.

B

C

For step, one, I think um yeah.

B

Yeah, I I guess that's where I was going with pushing all events to land in event data, because it limits some of the discussion that we're going to need in the general otep and the time that will take to get that through. If we just say an event is event, name event, domain and event data.

B

It's then just limited to okay, now we're defining. For this event, this is the schema for this event. Rather than trying to have uh you know a large set of fields being dumped into the common schema.

B

We're just saying for this event, this is this: is that the schema and how it's defined for that single event?.

C

So uh you so you, the top level, you know you're trying to keep the top level uh minimum. Is that what you're saying? Okay? So essentially, you take any any event: you'll probably have three fields or four fields, or something like that.

C

B

Three or four fields at the hotel level, uh whether there's other attributes, because the customers are sending other attributes or other vendors. That's in addition, um like we.

A

D

Yeah yeah, so so the number of required attributes we keep it to minimum yeah minimum essentials.

B

Yeah um because you know, depending on how ted's ephemeral uh session id stuff goes, um you know one of those top level fields may have to be session. Id.

A

B

That don't conform directly to the event itself, but or extra data that's carried along with it.

C

What would lead a min bar for something to meet uh or the bar for something neat in order to become a top level field? uh In your mind, um we name everybody and agrees. It's got to be a top level stuff. So what is the thing? That's going to drive the decision to promote something to a top-level field or leave it inside.

B

um Part of it's probably going to be the transport level like if we have something that we would normally transport in a header that for some reason we can't um it might live there, um I'm still in two minds about uh like where we've talked about like the part c, the custom data should that be a top-level field, or should we define that as a sub-field in data um yeah like I, I see an event itself?

B

Is there wouldn't be too many, but it things where they're common, where today they're like resource things that might want to change, they may be top level. So the things that come to mind would be. If we can't get a few more resources, it would be session id, maybe user id.

B

um Then we have like the unauthenticated and the authenticated user id.

B

Device id potentially yeah stuff like that stuff that isn't directly identifying the detail of the event itself, but it would help glue the events together.

D

But would they be present in the resource and not in the event itself.

B

Depends that's what I said if the ephemeral resources gets in, that would be the resource if they're not.

A

B

Our fallback would be your top level field.

D

Okay: okay, by the way, your your data proposal uh event.data, I think it will have to wait until the necessary attributes is- is clarified.

B

Yeah, it's not going to clarify, um uh had christian uh chime in today or in the last couple of days and his comment this morning or yeah from a few hours ago. uh He's really adamant. uh So I don't think it's getting clarified at all.

B

um Unless we get some other people jumping on the bandwagon, but at the moment I've got three people commenting and those three people are very adamant against polluting polluting my word, not theirs. uh Spans and metrics they're, not interested in logs.

D

No, no for for metrics I can understand, but for spans uh I collected uh some links based on our last week's call, and uh I'm sorry, for uh you know we'll come back to the original topic and um now so I collected some based on our last week. Logs sig call. I collected some data on uh you know so different logging, libraries.

D

You know uh that they they also support uh nested attributes, uh so that information I collect like electric feelings, and uh I also put together uh you know a couple of examples where spans also would benefit from having nested attributes yeah. So with these two, I'm thinking that we should create a notep this week.

D

But then I was thinking that metrics uh still doesn't uh uh I mean the original challenge was that only one signal uh in a needs necessary attribute right and the two other signals did not didn't. Therefore, uh since they're using common code, uh you know nobody is getting nested attributes today and let's say you know, we introduced nested attributes for spans and uh logs.

D

Now, how do you, how do the implementations prevent it for metrics, yeah right so so either ways I think uh using a common code base is a problem, um but at least we we want to get people to agree that for spans it's useful and then you know let people start thinking and give ideas on how to handle it.

B

Yeah, I think if you look at the issue 376, which is linked off my um thing: that's there's a bunch of people that say it's useful and a bunch of people saying no avoiding like the plane so.

C

I thought uh uh was, I thought, making it optional, you know, making uh nested attributes optional. uh Was the sound like a weight of a way forward.

D

And the concern is the moment you make it optional or in any type the moment you allow it. They are worried that people will start abusing it. Yeah.

A

B

Yeah they don't know if you, my spec thing, is saying it is optional, because today in the protocol it is optional um and they're saying no. No, we don't want to have anything associated nested anywhere near the existing definition, um so yeah- and I I think a lot of it comes from uh previous back-end discussions. Where back-ends just don't there are the major back-end, jaeger and prometheus, don't support, storing, nested and they see the api is storing nested, which, from my point of view, that's wrong. Yeah.

D

I I also feel those back and those uh other aeger and zipkin. I I don't know why we need to consider them all the time. I think they they either they need to evolve. Or you know if they are not evolving, then you know we shouldn't be uh like open, telemetry shouldn't, uh you know be held back.

A

B

And just just because it doesn't store, nested objects doesn't mean it can't represent. Nested objects. I think that's the bit, that's being lost, you can still represent nested objects either as a json blob or it gets flattened for those back ends. So yeah exactly.

D

What if we uh uh settle with using a json string.

B

Yeah, if we use a if we change and say, okay event, data is a json string. um It would then that the other, the shape of an event, would then work for a span event as well.

A

C

The problem would be uh performance on the.

A

B

Yeah that there's there's an advantage and a disadvantage. The advantage is a adjacent string. Representation of a nested object is probably going to be smaller than a json string representation of an uh nested attributes, but you then got the you know: conversion either way both times um in terms of a grpc and an otlp representation, so binary form um the method attributes would be, I think, less than adjacent string, representation.

B

D

How bad is is an additional uh computation to convert to json uh I I know you have provided data before, but uh well.

B

It's in terms of having an object and converting it into a string representation. Json.Stringify is the fastest way to do it um if you have to iterate over the object, depending on the number of fields in that object and how you created the object.

B

That's expensive, um uh like in the examples that we do today where I have a switch to enable or disable that feature, there's a 20 to 30 cpu saving, um so we're talking like milliseconds for lots of it, but it can drift up to tens of milliseconds when you've got to iterate over every object which, if you've got um if you don't represent.

D

Json.Stringify is supported by all browsers right, so.

B

And we're not going to worry about i8 onwards, but it's been defined for a long time. Okay,.

C

Then maybe sorry, one one other thing: it's not just only a browser problem right. What we're defining you know essentially for all clients. um You know I am never chatting about it that brought it up.

C

Iot is probably another um thing that we should consider yeah.

C

None of us are in that space just yet, I think, but um whatever events you know schema and we define and the protocol and things should work for ifd scenarios also, so if it is expensive for browsers, uh browsers, sandbox and everything you know if you're concerned about performance and browsers, I think we should be more concerned about these cool devices.

C

If we ask them to do heavy processing, it might be a non-starter. So would we? How do we solve that for them.

B

Yeah because you you got low powered low bandwidth devices, which you know there are still some phones out there- that they're still selling today that there are not big feature phones which would have the problem.

D

Well, I think the the problem is.

D

It goes back to the same thing right. I think we are unable to convince these folks like we need some tc member. uh You know uh ram to be working with us closely, like ted used to be that person, but I think he went on vacation.

D

Because some of these uh features even the the session level uh aspect, uh there are several topics that uh you know when we propose. I think that they're not moving forward.

C

Do you know when uh ted is back.

D

He said three months and it's been a month so perfectly another. Six to eight weeks I say.

C

Okay, um nev: have you chatted with riley about any of these issues at all.

B

No, I I like, I was going to bring it up in our internal meeting last week, but there's a whole bunch of other stuff gotten away. um I think.

A

B

We have a bi-weekly. I think I can only attend one of them, because the other one clashes with um other general hotel meetings.

C

uh Basically, for.

B

The internal sync that we have.

C

Oh, you can only do bi-weekly, but the others uh uh not possible yeah.

B

No, no, I can only do one of the two sessions yeah, so I was going to bring up for the next one, but the next one that I can make.

C

Okay, uh I I I think we could you know you could just do a one-on-one with uh riley, you know, he'll be internal and everything should be fairly easier.

C

I can uh I'm going on vacation starting tomorrow, so you know, but I can you know check with them see if he is got the bandwidth and um the reason why I asked about ted was: if we pull in somebody else, and the value has to kind of come up to speed and thing and ted is already up to speed uh some of these things and um he could start from where he left off and things yeah anyway. If we have to keep moving, we will see if value.

D

I think one new member would help.

C

D

C

All right uh uh I'll I'll uh riley is my peer by the way uh yeah.

A

C

Yeah so I'll I'll chat with them and then, if you can connect with them yeah you know pick his brain on. You know, get his thoughts and see if he's supportive or you know, if he's also pushing back, then we know the resistance. Is uh you know uh you know we need to first get people convinced. I don't think he would be. He understands a lot of our use cases and things um yeah, let's, um let's, let's chat with them, see if we can get more help from pc.

D

um One quick question: is there anybody in microsoft uh that we can talk to for some basic knowledge on iot, uh because my understanding with iot is that all the iot devices always have a local gateway, so so the iot devices themselves do not send the telemetry directly to the back ends the the local. uh You know. Some controller, uh you know is the one that's controlling those iot devices.

C

D

Right like hub, yeah, yeah and- and that is you know relatively you know, powerfully.

C

Powerful, yes, I agree, um so you know.

B

Not all the iit devices will have a hub, like you think of your your ring. The ring can have a hub, but it doesn't need one.

A

B

Yeah, you can buy cameras online, which are effective, iot devices. They don't have a hub, you just buy them and they connect directly to the internet. So.

D

Of course, it's not mandatory yeah.

C

For just washing machines, they they're all iot devices yeah, but anyway, we'll we'll see if we can connect with somebody. So so far the um our team has been. uh You know, kind of pushing back on any of the client in a mobile and client things at all. There is a bigger effort going on to see there there's. You know two um there's a group of people that want support for our clients. Internally, um you know support provided from our team.

C

We just don't have the resources, so there is discussions going on about you know, staffing up, and things like that. So right now in a mobile is not top of the priority list. For us, we do support all service scenarios and browsers uh and uh the thick lines if you will right uh the c plus bus lines, but nothing mobile um is supported by our team, but we'll you know ask around see. If we can, uh you know connect with somebody from you know purely from iot.

C

In a point we can, um I will see if we can find somebody, okay, cool, um so coming back to this. I think um so.

D

I got it so I think I'll I'll start putting data from from our product can.

C

We, um you know- probably I was hoping you know a little bit more people, but you guys have been in this uh sig for some time and stuff like who? Can I start working with you know, I would like to just you know pick them on one on one and say I.

D

Think aws, uh you know, quinn hanan um was the one that provided data last time.

C

The way I see it happening is somebody's got to you know, push this through right, so I can connect. I can be the connecting guy and collect the data and just put it here, and then we can review it. You know every so often in our um six and then I push through for um my goal is to actually have these things defined at the end of it um relatively quickly.

A

B

Yeah, because martin's trying to be uh try tried to do the resource and the timing thing and at the moment he's the only voice in the wind so yeah by by having this, we can then break out and sort of say. Okay, once we've identified the the companies that are involved, it's those people that would be directly involved in defining it.

C

Yeah we could even do you know, uh you know, connect people together and show their groups smaller groups or whatever, and then figure it out too.

B

Yes and try and make.

C

Yeah, we don't have to be involved in defining everything. So, for example, the three of us might be experts in pageview and we get a hashtag uh are interested in pageview. We could. Actually you know. Martin can go work on timing with a couple other folks and stuff: okay, okay, cool.

A

So if you can get this.

C

uh Help with the pointer contacts and then I can take it from there push through and then.

A

We'll develop this.

D

Yeah, um uh I think some of us who are active at least we. We will make sure that we provide uh complete data and then uh and then we can get started uh from there. Okay, then, if others, you know feel they want to participate. I think at that point you know they can participate. Okay,.

C

Sounds good if you're going to fill out uh the the rest of the stuff? Also, um you know feel free to do that. So this is the pattern that I used to just go, create a copy, this table or whatever and then start filling out yours with the cisco, slash browser, slash whatever your event name is: got it cool, okay, all right I'll! Stop! Sharing! Now, if there are other topics will be chat about that.

B

D

Have anything yet.

B

Yeah, probably a quick question for you ram, um so is it cjo who worked on the the back end htlp collector for the top net.

C

No, um uh the back end of the receiver is what you're talking about right, yep yeah. So that's.

B

Okay, um because I want to connect with them to find out how they're going to handle part c spans, um because that's that that's the primary driver for nested attributes that we're going to need, you know we're going to need them for spans as well. So so.

C

The thing that they did was for the internalization service right so that basically flattens out everything, there's no nested stuff or whatever give me one quick. Second um uh yeah you guys can continue talking I'll I'll, give you that name uh confirm the name a little bit.

B

Yeah yeah, so I I just want to see what they're thinking for there because based on the push, if and with the being in, like dot net by default. Now, if, if it's going to be flat, names all the way through and that's the expectation uh anyway, I I.

A

B

Want to find out.

A

C

It's it's tom and secret um yeah, okay, I'll I'll, bring you there aliases. Also on teams, yeah cool.

A

B

uh And that they come to the internal scene, don't they.

C

Not unless you invite them.

A

C

They don't show up all the time.

D

So um um not I'm sorry, I keep uh taking the wrong name um nev. I think, uh if it helps we can prepare a list of topics that you could take it up with riley yep. So we have one is the nested attribute. Second, is the session ephemeral attributes at the resource.

D

Are there any other topics that.

B

Not that I think I took my head.

A

Let me just check his calendar. He may actually be.

B

Busy during these songs- and I know he does cover a lot of.

D

Yeah at this point, those are the two top topics that we are we're blocked on.

B

Yeah tuesday he's blocked out um solid from uh eight o'clock onwards, uh every tuesday and wednesday that I'm seeing, um and that was an out of office between this time. So.

B

Yeah, uh it's probably not going to be able to make this video.

B

But yeah we I'll stick with him and we'll you might be might have someone else. You must suggest. Okay.

D

C

Yeah, I just think about already about it. Also, let's uh you know I'll I'll, give it a little push now, so you have a little bit more context when you.

A

Open it just feel.

C

Free to yeah, I'm sure he'll make time at least to understand where they're coming from and give us thoughts and stuff.

B

Yeah, I I was going to come in the office after this meeting, so right as long as he's there, I can go yeah place to face today.

C

Just just bump into him in the all day and.

B

Just opened up the sliding wall and said: hey.

C

Yeah exactly he sits right behind us by the way we we have it closed off, though.

A

D

um That's all, then, we can meet tomorrow.

A

Thanks guys, bye cheers.

A