Kubernetes API Expression Working Group, 26 May 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: WG API Expression BI-Weekly Meeting for 20200526

Description

WG API Expression BI-Weekly Meeting for 20200526

A

Here we go so welcome to the coop API expression meeting it's 26 of May and we got no nothing on the fixed agenda. So has anybody anything to end here that should be discussed.

A

Doesn't sound like it until you're muted, if you're talking.

B

I'm, not okay,.

A

Then, let's go through our work items.

A

Slightest oversight transition.

B

Not much I can see that durian is not here. She really is trying to finish it, but it's also I think it's a changing job, so I don't know how much is going to be available after that, so we need to plan on that and yeah, but I think is trying to finish that, but it's also not making a lot of progress these days. So maybe we have to finish it ourselves.

A

A

The community I think it's also part of the discussion if we should keep apply or not with the first conflicts and, of course,.

B

B

Alright doesn't make any sense for also a certain place are.

B

You not intend to make a conflict option default to true the first conflict defaults to true I.

B

Think people will have to change the OCI CD system to make the to make the necessary change if they want to force it in they'll, see a CD they'll have to change it.

C

Yeah, are we gonna talk about that today? I.

B

Mean we can do you have something to say? Oh I,.

C

I know I saw that you opened an issue on it and I responded, but I didn't see much since then. I.

B

Have one free spoon, yeah I shouldn't probably respond to that yeah.

C

I think it's actually pretty important that we don't add a new option without deprecating the client side apply I, think that will be super confusing for everybody.

B

Yeah I, don't yeah, it seems like you're saying if we duplicated client side of light and we couldn't make a different come in.

C

Know what I just said is that I don't want to convince, because that's super confusing sure.

B

Yeah yeah I. The goal of the issue was not to have a discussion between you and I, because I think we've had this discussion yeah, but I feel like I've seen people's asked me to open this issue. Mics bright so was one of them because he was disagreeing with how it's done. I thought just instead about Rogers degree about the strategy. I've heard people disagree about the strategy. I don't know if I agree with the strategy myself. But if we've had this discussion together before my goal was to have opinions from other people, yeah.

C

Okay, maybe we could yeah so those opinions, because I don't like the partial state that we're gonna, oh yeah, I. Don't thank you guys, I'd, like some certainty,.

C

Yeah that'd be good III, don't think like great.

B

But you won't like what I want, like you say, sighs thanks! Oh thank you. What would you? What do you how'd? You beat.

C

Your way so I think we should figure out who who the stakeholders are and honestly I'm, not sure it's any of the people that you listed I. Think uh the people who should get a say in this are basically six CLI and.

C

Two of us are extent API machinery, folks.

B

Sure yeah I mean the videophone name also I had to seek CLI and the world really current bland, so I think that's fine. Okay,.

C

Maybe I'll go find the issue and just say that, and if people wanna say that we we think the stakeholders are sync COI and they haven't heard objections from them. So we're going to assume that we're not adding a second command we're a little over the existing command yeah that if people disagree with that, then like, maybe we should give them a few weeks to argue or something like that: yeah yeah. It could.

B

Be pasted the shoe in the and I'm gonna cut these to do these Jackson one thanks.

A

A

The next one would have already fall, took associative keys, I think Chinese, New Year, either yeah.

B

I can see Jenny I, don't think I, don't think I don't know if that was any progress. But if there was another well bit.

B

A

Till example for filtering policy yeah.

B

I've read these documents at the bottom: I need to.

B

The goal is to look at some fields and agree or disagree on what should be done if we remove them and so I've pasted an object and I'm going to comment on this field and say what I think should happen if we remove them and then I want people to also comment and give the opinion. I suggestion has an opinion.

B

Yes and Justin is saying that in general I think and we talked about it, I think his opinion is that, whichever path you take, if you apply an object yeah, you should end up with the same result whether you've applied different things before or not. So, if I apply an object directly, oh, if I apply something and then a politely subject, I should end up in the same state and because we're doing the dangling thing, that's not happening right now, but I think the reasoning doesn't work because of gflops anyway, yeah.

C

So I think I think his assertion is correct if we ignore the effects of other actors on the systems of the system.

B

Yes right, that's that's a good! That's a good Bank! Yes! So.

C

Other people are doing stuff I, don't think because they might be owning field sleep. So you have to ignore that, but if yeah, that might even be a good way of testing this, if we assume that there's a single applier, then you should be able to have a test that has like five different versions of a yellow file and applying them. One at a time should give you. The same result is just applying the last one directly.

C

B

Is not what we're doing today, though? Yes,.

C

That is very surprising. You.

B

B

What you're saying is that if the HPA has changed the field and the plane without the replicas field mean that we should leave it alone. But if nobody else has applied the lab, because I'll change the replicas, then we should only set it.

C

Hold on is there a default on because it's because defaults matter right is there default on that field?

C

Yes to London, so we'd have to either have the rule that defaults and fields don't follow this logic, or that when you remove a default in field, it gets reset to its default, which is what happens today with clients that apply, and that is also about, because, if you start off with a thousand replicas change it to or delete that field from your manifest and then win the race with HPA, you don't want your service to actually get turned down. In the meantime, you.

B

C

B

If you'll be on the ecto, so yeah at this point you may be the only actor on the aged really hasn't triggered yet. So there is only one actor at this time until the HPA triggers right exactly.

D

B

At this time, you're the only actor if you're saying that, when there is only one actor, whether you've applied with a thousand Valona, doesn't make a difference, then it means we have to reset. You are correct: I'm.

C

Saying that that is not the right answer, so his rule doesn't apply to you. I'm not defaulted fields, yeah, okay, okay, yeah, I, think his rule might apply to non defaulted. So that's optional or like element just like list item item fields, yeah.

C

C

I'll try and find this document and state that there I'm already in the middle of commenting on the other issue.

B

I'm going to make the link thank them.

A

Okay, next one is status by thing: I updated the description in the SMD PR.

A

And I started with I put new change from SMD into the kubernetes repo and use the versions correctly.

A

So if anybody wants to review the stuff in colonies before I go on and do it for every field type for a resource type again, I'd be glad for that.

C

Review and these.

A

Right on the assembly- yes, and probably also on the implementation, as it would be annoying to continue again to all the coins and then it was for nothing. So the PR is big, but what you can do is edit slow. You can go to the pod strategy, which is the way it's currently implemented and see. Then you way the stretch is built with the reset feels. So that's an ex had to keep those things together. It might be ugly or wrong in your opinion, let's go to two-year, but it will.

B

Need these merger.

A

And it allows to provide dynamic bank functions that get call it run time for collecting feature gates and individual fields. I only did one so far because there's a lot for pods, so it might be feasible to make a generic function for pots because they are used every of those. This drop disable pod fields is used a lot, so that would be fun to build, but in general that's the way. I came up with defining the reset function. The reset feels and the dynamic receptacles in one place.

B

B

Don't hate the idea of the having everything in one place.

A

Yeah I, don't know if it's more readable than before, like having one call at want everything together and expose an interface, but I can understand if it's weird to have function. Definitions in a in a function call I'm not sure if I like it myself, but it keeps stuff together. Yeah.

B

Yeah I can look at it later. Yep.

A

And this stuff.

B

I sent it to you Daniel on last week, someday if you can have a look while trying to have this this deep removal of fields, homicide.

B

Yeah I asked some questions than this: one didn't I yeah and we tried to address the documentation and named me and.

D

C

I mean I asked for the documentation, because I really did not understand what this was supposed to be doing. I didn't understand the difference between this and a regular one. Yeah.

B

We struggle to understand why you didn't understand, but, uh okay.

C

Yeah, if you've had an stuff I'll go look again like basically are we sure we need both, like maybe they're, a regular one, just isn't doing the right thing or no.

B

Because sometimes we want to remove exactly the said that he specified yes, we want to do a removal like if you specify the would like a is bad. We want to remove everything. He knows that.

A

So this one behaves like deleting a folder and there's folders in the folder. You delete the forms inside there. The previous one delete deletes, be it like. You address a direct field down until its name and that you have.

B

Examples of that in the test now Kevin, don't you yeah.

C

That's gonna help alright.

C

Sorry I'm, like pretty swamped so yeah.

A

I'll try and look at it. Yeah take your time. If you have questions you can.

A

Okay versus happened there yeah I.

B

Think we should move that to the backlog, because we are definitely not working on that.

A

A

A

And I: oh yeah, we have this, but that's sometime, much related right now and I. Think Joe is here. Do you want to talk about the schema validation.

E

Yeah I haven't been able to do much since last time we talked since I got pulled away on other things, the so it's pretty much what we described last time: okay, how it's coping getting this doc together, so I'll try and get that I'm going away this week.

A

Okay, Mike and Tim are not here this week so and last week, I think not much happen on it. So thanks Qatar and.

D

A

Converts your structural schema to be schema, I'm, not sure. If anything went on there.

B

A

Here we at the s suppose I deploy, doesn't honor force discussion. We have to last yeah.

B

I can't remember I, think I think it should be easy to fix. I mean I, know Daniels, not a big fan of this flag. We would oh yeah I, remember saying that I hate this flag, but the plan is that have it now so I think we have to do. We have to implement it wait? Why?

B

Because we have to like, we need the deprecation policy if we want to remove it or not do it. This is the this. Is the delete, recreate yeah.

A

Another comment: another part where we discussed recently.

C

C

There's one we can make force incompatible with server-side apply.

B

I, don't see how it's related to some sort like people who use the flag the flag before. Why would they not want to use it with sauce and blood? What's different.

D

C

C

So this honestly force is not a good idea. I know things are recreating. Them can have really unexpected consequences right, like if you add, for if you apply a namespace and add force you'll find us. The server deletes. Yeah server deletes all your stuff, um so I I think it's too dangerous for yeah I think we should have named it like delete and recreate, or something like that that at least people know what they're getting I don't think force is a good description of what it actually does. It's.

B

C

So I, you can argue that Q control does it this way for all commands and we shouldn't be special.

C

Which I guess maybe is it a global flagger? Is that a flag, that's added to every sub sub command? It's just fun, I think it's just for apply yeah! It's it's not.

B

C

For so small, like so yeah I think edit just does the update, have a force option. I think update has a force option. Maybe if it's a global flag, then we maybe we have to keep it. But if it is a per command flag, then we should definitely.

C

Like market deprecated, beaver for clients, I'd.

C

Make it incompatible if you try to use server-side and add a delete and recreate or a delete? First, maybe that's what it should be called I mean.

B

Do we really want to have a new flag just for server-side of like wheezing the same command? That does the same thing. That's.

C

What, if it's, if it's a global flag, then it should be fixed to the globally if it's only for a server site. If it's only for applied and yeah I think we should change it so yeah it seems to be Noble, we should I guess we should change it either way is what I'm saying the question is: is?

C

Is it our problem or is it cute? Controls problem looks like it might be: cute criminals, problem, yeah, okay, yeah, okay slick. If it's a global flag and we can't change it and we should I guess we should implement it.

B

If somebody's using these flares bottled, LCI CD felt a little good luck to them, they're gonna roll we're gonna break it for these people again.

C

It's it's really specially confusing for a server-side apply, though, because we have the kind of forced conflict, yeah sure so I wonder if we should specifically prevent it for server-side apply just to keep using so using force when they mean forced conflicts.

C

You know using you know we could add in parallel, but because I guess if well maybe people don't need a way to do this, but if, if we think people need a way to do this,.

C

We could add a sibling like like an alias flag, a global delete, first or delete and recreate flag. So people could call that with apply instead of force.

C

Alternatively, we can call.

C

Force immutable field sources, we have an apply specific command, I, don't.

D

Know, do you really want.

C

People doing this like: is it really that much of a hassle to write you know, delete f the file name and then apply f the file.

B

Name well, that's different, because D f is gonna. Remove everything.

B

We go easy before we delete, so it's if you said, Faust, immutable and I think the initial goal is to say indeed, if you're trying to change an immutable feel you need to remove and then I hear. But the prime is that I think, even if you get like 500 or like whatever they are it's going to delete tonight, okay, which is table creation.

B

So it's pretty bad happened.

C

60 or I- maybe it seems like this- is.

C

Yeah I: do we have to solve this problem? To get server side apply on? What's the minimum that we can hope we can do to make progress here.

B

Well, I think we have to do what some what coop kernel does if we want to change the behavior of coop kernel, because we're very bit wrong, and we should raise that. But it should be a separate also going on.

B

But I think for now the only option is to actually do what they asked us to do. I mean they're, not asking us explicitly, but I think that's the that's the behavior. We have to implement as bad as it is.

C

Yeah, it's really unfortunate. I think users are gonna.

A

So from the command description, so like its immediately being to leave in replace not only when there's an error is that wrong, because it only replaced on error.

A

The reference here says immediately removes resource from API.

C

Now you're, looking at the delete.

A

I'll wait to replace here, yeah.

B

We'd some replaced is the same as a blood-stained.

A

B

Yeah because the the flag is degloved once and then or years, every well.

B

C

Does it so is it about reef? Is it about recreating or is it about bypassing Grantsville deletion.

B

Boast no, it's yeah! It's about doing the DD to like.

C

B

Various races, even because it won't be able to really create the object, otherwise, no right.

C

But on the I mean it kind of means that it means different things in different places.

C

And if that's the case, then we can decide what you mean for server side apply.

A

Also for applied we using force with so that apply would mean we the object and we trade, which would lose all managed fields and that's tendons.

A

It would be necessary because you would have to do in anyways, I suppose, okay.

C

C

What's the actual problem, what's the actual problem we're trying to solve here.

B

Somebody can you need to care saying that sauce doesn't work flows out of a side by and well wondering if we need to implement it or not,.

B

And my suspicion is that we probably should and Chloe today.

C

Do we all agree that this is very yes.

B

C

B

Again, I think it's a case. We have to argue with this yeah. Thank you.

C

Okay and right now it doesn't work because we didn't implement it.

C

You know what I personally I I think this kind of boat, I.

C

We have to do something eventually, I, don't know if we have to solve this right now, I'm much more worried about the semantics around feel fields which which fields get deleted and which could reset when you remove a field for your yes, I think that is a much more important problem.

C

This is like it's broken right now, but at least it didn't destroy the person's cluster or anything like that.

C

C

B

Agree: it's not very important.

C

So maybe can we like put this on the backlog to revisit and like a month or something III I, just it doesn't feel like. We should be spending a bunch of time on this right now. No.

B

Should we, at least being the CNI maintain, knows maybe to ask follow pinyin yeah.

C

You're all right up equipment. Can you paste that into that job? All I was doing so all right, a really quick yeah. That makes sense all right. I didn't mean this to be the meeting where I take a bunch of work.

A

Feet are up by three 1/16 Clanker, that's one where we roll a discussion with Jordan.

B

I think yes, poor kids but yeah. Let's part of it.

B

Yes, yeah I mean.

B

Well well, working on this I mean I've made bogus' the bit. My PR is ready. Your PR is mostly ready as far as I understand.

A

It's a I, don't know if, if you come to this question, but basically we're dropping managers on any conversion error to a manager version No.

C

Say that again.

A

So currently we we drop managers. If a version is not available for conversion. Okay, and now we drop, we would drop the manager on any conversion error.

C

Any conversion error like what's that an example: conversion error, the.

A

Object is it's not convertible order version is not there and one reason was that see: are these curly tones return.

C

A

C

Managers are just the managers for the waters and only only the one, only the manager, only the manager that had the error, yeah and therefore.

A

Can't be compared compared.

C

Okay, that doesn't sound too bad I. Guess like what are you gonna do yeah right.

B

We used to say.

C

I mean that's, I mean what happens if you have registered a web hook, which requires a conversion between those two versions.

B

The web hook faded, it's gonna drop them in a job I.

C

Mean if the system is misconfigure to this extent, should the request really be succeeding? You.

B

Expectantly quest is gonna fade anyway, but we believe it's not up to us to decide.

C

Okay, if the request is not gonna succeed that it doesn't really matter what we do right right.

B

C

B

If you could have succeeded, then we shouldn't be saying: no, please don't go it shouldn't, throw it because of us.

C

It's kind of an academic discussion right.

C

Why is this? Why is this case being brought up like? Is this actually happening?

C

A

A

Four, when we had the error that people applied when resetting the ministries, we set this empty object in its set, and that fails and exactly this place because we said a minute manager which doesn't have a version therefore become converted and previously that worked for internal types, but C or D stones return an error we we can check here so.

C

So like there shouldn't be objects that are not convertible in the system anymore.

B

A

Here we check if, if it's a missing version, error or in series, don't return anything that can be checked here, which is a different problem and that's where we came to the discussion. If what Jordan I asked, if we shouldn't remove any manager that can't be converted and.

A

What's the longest touch discussion, I had tried to recall it.

C

Okay, I, we pushed too many things under my stack and I can't observe this problem.

C

It's I I, don't understand.

B

We had a one-hour discussion with Jordan on Thursday about these and we came up with a plan and documenting the plan because we documented the in in the meat chat rather than an issue. Para I'm sure I try to remember what the discussion was and I also, if this is connected to another peer that I have well-well, not failing and failure, yeah I know and I. Actually not you have might be off anyone.

B

And I suspect we have more things to change again. Well, anyways, it's a massive mess and we're trying to, but we'll be making a lot of progress. If that makes you feel better, but right now it's table because some people, this is breaking some conversions and it's breaking people when they're trying to do an update and if the manage fields update fades, it's actually breaking some updates. Yeah, that's pretty good! That's.

C

Definitely, that's definitely that's definitely problem.

C

Only it only removes the problematic entry in the manage fields list.

B

Well, yeah, but uh yeah: okay, we can. There are also two stages. There is the stage one with. Let me take the field managers that are coming from the object and from the EDD object, and then there is the we are actually trying to use it to convert the object, and these two things don't happen at the same time, if we have a program reading the manage fields from the object, we're going to look at the ones from life and if the one from life we can't decode either we're going to use an empty one.

C

That's a diva's decoding the field set yeah problem, yeah yeah, but that's not converting the object problem. Yeah.

B

And converting the object we're going to remove them one by one if they.

D

C

You miss configure the system such that some conversion doesn't work and then you've like run the storage memory or something in touch all the objects in the system. Mistresses recall that manage fields yeah.

A

Only if the conversion doesn't all also fail when storing the object or retrieving it.

C

Yeah, we don't I, don't understand. What's an example thing that causes the conversion to fail.

B

Good question I mean probably, as we said, the ones who carry on seed the objects who have been very dated before so we know it's, it's varied object, so it should contour to anything we want, except of the oddities.

C

I'm not sure so so a web hook is actually like I'm worried about a transient failure from a web book causing a state loss in your object. Yep that doesn't seem good I think you should fail if, if it would have put conversion failed like like, if, if it was a failure, you're talking to the web book and not the web book, returned it a like error like this is never going to work. Uh-Huh.

B

Okay: okay? What about these, though? Okay so you're saying it should return a failure right now, the theology looking at the 1996, it's going to say that anything that happens during widely during the update of manage fields is going to be ignore you're, going to have a loss anyway, because I mean we're not gonna. Try to change this that you've just done so there's also lots anyway.

C

Again, that's still escalate to transient and put failure into state loss of the object, yeah yeah, which doesn't seem great. Does it.

B

You know, but you want to fail because of that. Yes,.

C

Iii transient well, pug failure should.

C

Should cause the client to need to retry, don't you think.

B

Yeah, how do you know it's transient? How do we know that I.

C

Mean we know some stuff from the return like did we get a valid response from the world? Look, this says it couldn't do it or did we get a like strange response from the web code like like connection dropped or timeout, or something like that yeah right like it, doesn't seem good if your your C or D conversion?

C

What book yes host temporarily like I, don't know it has a scheduling, problem or something like that and then all of a sudden, all of your objects that people touch are doing something weird, because they've dropped the flyer information that doesn't sound like a good, oh yeah,.

B

No I see what you're saying especially the applier, because we could lose fields because of that I. Don't think so much yeah.

B

But do we want to fade the request? I mean this? Is this would not have happened before so it's kind of a new thing. Well, we actually called what I hope we wouldn't have called the web hook before and we're actually failing the request. Well, we don't have fade, though equestria phone.

A

I think for applying, we should fail for updates. You.

B

Have a lot my woman, waster I.

A

Mean if we don't with my change, I I.

C

Mean if there's a if there's a web book that is broken in the system, some operations are gonna, fail, great, like things that are written on the wrong version or something like that. It's gonna fail, so I think be sure. Some update requests continue to function is winning battles but losing the war. I think the sooner of the system administrator is notified that they have a broken left hook that they need to fix the battery.

C

So I I, don't I, don't know.

B

Yeah, okay: we can yeah okay, we.

C

Can think about it, so yeah basically I'm worried about escalating transient transient problems into permanent data loss that doesn't seem good. Oh yeah,.

C

A

That doesn't seem that doesn't seem right. That also needs it for apply. You wouldn't want to change a mage, because we would want to fail if it's not a missing version.

D

C

We probably need to figure out what what possible errors we can have here like classify the errors that we can get and decide whether it's potentially transient and we should fail the whole operation or, if it's permanent, and we should just go ahead and and deal with it.

A

I'm done where you want to put that into your issue in the discussion or which way to watch it.

B

A

Those things are pretty much the same. Yours.

B

Okay, so just one question, then you're saying that.

B

Okay: let's, let's assume that it's not a conversion error and it's a trench on there and we fail. Should we keep going with the updates or should we fail? You update I.

C

Mean you, wouldn't you would no longer have accurate Manesh field sets right, yeah.

B

I mean we wouldn't updated, so, yes, it would be wrong.

B

It's gonna be very hard in the stack to decide if this is the kind of failure that we want to ignore all that I, don't know how we would do that.

C

We you're getting an actual error right. We might be able to classify errors.

C

There's there's some fancy like errors, dot, wrap functions yeah, but as and is we might be able to classify it that way. Yeah, but.

C

B

I mean right now. My concern is that, right now we have people were like saying: let's not use 118, because it's vodka like it's, actually actively encouraging people to stop using qualities 118, because it's welcome.

A

C

What what condition, what condition are they in that's actually causing it? That's something that I'm trying to understand. Yeah.

B

I think we have a bug somewhere in the endo conversion facility, like while trying to convert Saudis I'm, not even sure why we need to investigate about the days of burning the conversion they end up in an in an L. And it's.

C

Can we reproduce that.

A

That that mostly went wrong is that CRT conversion doesn't return a typed error for conversion errors, a full version error. So we can't know if it's a version every one we just fail at.

B

To convert in this case, she's the yeah, but.

C

Let's actually solve the so what you're doing here is like solved. The meta error, like all possible errors, but I, don't think that that's appropriate in this case I think.

C

At first I thought like this: is a system configuration problem, but now it sounds like this is actually a bug in another part of our code like we should definitely not be throwing away data because we haven't fixed a bug in another part of our code. That doesn't make sense to me, even if the C or D error is hard to fix, I think we should investigate that and we can I think we probably find some people to tell us X. Actually, Joe is Joe's on the college.

C

I have with an inclination to fix like that. I don't know.

E

Sorry I missed the details on this one. It was, it sounds like there's some sort of CRT.

C

Conversion, what a book called failure.

E

Yeah, that's interesting. I'll definitely take a look at it. Yeah is it. Is it one the cursors on here the ID 106? That's.

A

E

Yeah this one read up on it. Okay, let me let me have a look at it. I'm bringing up my screen now.

C

A

A

So below the open fast one is one of the reports where they have this weird conversion: error, where they don't get.

B

A

At some point they get.

A

Jordan opened an issue for that.

A

So conversion fails because it's missing version and client and that's at least that's the error, but in that case Jordan couldn't reproduce it and and okay.

A

And we had a few like one wasn't calculated.

A

B

Mean I believe that if we have any L in our code, we should not break everything. So people, like my opinion, is I- think iris Joanna, it's possible that we have Peggy now occurred between having update or anything from pletely Balkan and having a missing managed field. Amazing update in the managed field, I think I'd rather have the missing managed field. Update like this is. This is bad right, like people yeah. This is bad. This is making me super. Uncomfortable and I'd rather have the missing manager update then than these.

B

If we fix the bug will end up in the same place, but the experience for people as we fix the bird is very different.

B

My main problem is that we're never gonna find these bugs. If we have the warning, though,.

C

Yeah, that seems very unfortunate. Yeah.

E

Yeah, the kind of what is, on my mind, is like how many different types of bugs do. We have like this, for there were a few or do we think. There's many I'd like to understand that a little better first before we make a call on like what what we'd like to do to fix it cuz if it turns out that there are very few and we can find them and address them individually, that might actually be better than to try to do something broader, which might end up masking problems. Yeah I feel.

B

Like people have shoes right here, we just come to in a couple of weeks: okay, okay, I've got bad.

E

Okay, I can I can spend some time on both of these and see if I can extrapolate from them to more broad cases as well.

C

Okay, thanks Joe I opened it to note tab I might comment on it. My game is full and I need to I need to drop off. This call before I absorb any more information.

A

We have four minutes here and once like Adonis, if you're ready to do, can you think of anything that would be a good starter issue right now? Oh it's.

B

About Stano issue, okay, I wasn't sure what the context was.

B

Okay, let me mmm that's not easy.

A

B

One is perfect, the.

B

Manually, updating managed fields should be impossible to sub resource. That's very good.

A

D

Sounds great I think I would have to work on it offline and observe a little bit, the context, etc. Maybe I'm gonna I'm gonna pay you in slack I. Actually.

B

D

Much contact context on yeah.

B

D

B

Let's talk on on the slack, maybe the working of expert, I expression, so that everybody can look, and we will give you a starting point and what you can these things yeah.

D

A

Alright, thank you recording thank you for being here and goodbye yeah.