Cloud Native Computing Foundation TAG App Delivery / Bi-Weekly, 16 Aug 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: 2023-08-16 - TAG General Meeting

Description

TAG web site: https://tag-app-delivery.cncf.io/
TAG Slack channel: https://cloud-native.slack.com/archives/CL3SL0CP5
TAG git repo: https://github.com/cncf/tag-app-delivery
TAG meeting notes: https://docs.google.com/document/d/1OykvqvhSG4AxEdmDMXilrupsX2n1qCSJUWwTc3I7AOs/edit

A

B

Yeah yeah, we could do that too. It was a.

C

B

It was a pause yeah. We need to get everybody up to speed on or figure out a process for some of these things so that we can delegate the zoom meetings and I'm sure you guys can probably do this YouTube videos better than me, and so we gotta figure that out, but.

C

Youtube does have some.

B

Interesting like pieces, where you can trim things out and.

D

Now some of it is like AI driven some of it because, like the chapters that I think those are very useful for like meetings like this um but yeah, so you can do that manually or you can like have ai. Do it.

B

Oh okay, yeah I, don't know how to have ai. Do it so yeah I need your help. Let's figure out how to get everyone on there appropriately yeah. It would be nice like today, with KCl presenting like Google, put a tag in there, so people could jump to that. If that's all they wanna yeah. Do you.

B

Hey everybody we'll get started in like a minute or two.

B

C

This is great for a summer meeting.

B

Some stuff might have to get I, don't know because see how much we get to.

D

Are you sure this is Abby's email address.

B

What are you looking at? Maybe don't.

D

Say yeah I'm looking at the meeting notes and then.

B

Yeah she put that in actually, okay, okay.

D

B

I'm assuming she wrote in that whole uh little section, because I.

D

Didn't know that, yes,.

B

Alexander I'll paste it in here where we're all looking right now.

B

B

All right, so maybe uh let's go ahead and get started. um Yeah it's 1003 got. We got a pretty good Quorum here. Thank you, everybody for joining. um We do like to so our agenda. For today we have intros, then a few status updates.

B

um Then we have a project presentation from the KCl team who recently submitted KCl to the sandbox, um and then we have a couple folks that want to share ideas and uh and issues that they've opened. So let's Jump Right In the the agenda is in the meeting chat there. um So first is introductions.

B

Thomas. Do you want to introduce Leanne.

E

Yes, so um I'm happy to announce that Leah has been um approved as a take lead of positions have to get delivery so give everyone um yes and I hope she will support us very in a very nice way. She did it until now, so he she joined us for kubecon um drove a lot of um initiatives through through group con, and it was really nice to work with her and I think it's she's she's a pretty good enrichment for the tech. So yes, nice to work with you.

D

Thanks Thomas do I need to say something as well. If you want uh well I'm, just gonna, say thanks Thomas for pushing for me to become Tech, lead and uh yeah I I'm gonna stop the to doing anything. Now, because now I've already achieved everything, um I will still like Drive the website stuff forward, because I really want us to communicate better to the outside and give people from the outside better uh more ways in to the group and uh for cute country cargo, because I'm not going um to be there in person.

D

I want to think about ways that we can like have a cool, remote experience as well, maybe slightly better than we had in Amsterdam. That was I, don't know! If anyone watched the eight hour meeting we had but yeah. So these are the two things that I'm thinking about now for the next couple of months.

E

Thank you, and with this back to George,.

B

Yeah, that's awesome. Thanks Leanne, we we're really looking forward to your help there and everything you've done so far. um We like to open it up now for a couple minutes for newcomers. If, if you want to you don't have to but share, you know in less than 99 seconds share uh your interest in how you know. We won't the reason why we're asking so we can help you contribute.

B

um So if anybody wants to I'll pause for a few seconds here and let people jump in and introduce themselves, and um if you want to tell a app delivery story in less than 99 seconds, you can go for that too.

E

So I'm um Just for information, I'm a teacher. So if you don't want to speak, um I will select someone.

E

D

I think are you speaking because we can't hear you.

F

Hear me now: yes, yes, good, okay, perfect hi, everybody! My name is Alex I'm, a developer on the operator life cycle manager, project- and you know, I'm really interested in uh while I'm attending this meeting, because I'm really interested in seeing how we can all land on a maybe not a single but at least a standard for how we package and ship software.

F

um One of my biggest fears is seeing about like there's a bunch of great projects that enable you to extend kubernetes or install packages. uh uh You know bring capabilities onto a cluster and what I don't want to see is the ecosystem fragmented in terms of how like what stuff can install certain things? And so what I'm doing here is uh coming in to see how well what other people are doing in the space of packaging, stuff and I hope to just learn from you and see what you're doing?

F

And hopefully uh you know which I think the purpose of this meeting is is to drive a common way for doing. You know the packaging portion of it so that all our projects can bring these great projects to different clusters and uh and also not fragment the entire ecosystem, because that seems bad.

B

Thank you Alex. That's awesome, um I I want to let you know. You probably know, because some of the operator framework folks have come to the meetings, but there's a group that we formed called working group artifacts, which is, is very deep in just that domain, uh pursuing some Synergy standards in the internal formats of oci bundles, but they want to think bigger than that, but particularly oci, bundles. So perfect,.

F

C

I'll I'll be attending.

F

That one too I just saw this one as well, and it's like okay, let's just jump on all so it doesn't get involved in the community. Thank.

B

B

B

All right, I will I'll stop there, it's hard to stay silent for so long, but thank you for bearing with me um all right. So, let's go to the uh the next one, which is status updates. uh First, one on the list is uh just the annual project reviews um I, think Karena angel had taken one uh I know Thomas. You said you were going to be looking at one Colin you had mentioned. Does anybody want to share how that's going or any feedback.

E

um To be honest, I'm current I was currently trying to find the second one who is reviewing with me. um I think I found some I found one, and we will start doing this now. So, let's see how this works so currently, the hardest thing is to find out how the pro how the process works so I found one document where this is described pretty. Well, um let's try to find our products how this works for us.

E

C

Yesterday, that's what I was curious on too um going at this cold first time. uh I definitely want to spend the time and do it, but if anybody else has had some experience doing it in the past or Thomas, if you want to chat about it and share some notes, um I think that'll be helpful, or at least for me.

E

I think that UCS also created annual reviews um um issues in their report.

B

Yep also the past couple meetings we've been going over this, so you could take a look at the notes from 719. In particular, hey wait. A second 719 yeah, mostly 7 19. We had a big discussion.

B

It's really like it's a very Bottoms Up I mean a little top down because we have Emily's support. She's. Very, very much wants this to happen and it's the first. You know concrete experiment delegating work to the tags. So you know, if you jump in and you see- oh I, don't understand something you know chime in on the issue where they're discussing it in the talk or write to Emily or me or Thomas. You know uh individually. You know they. There should be an iterative process and, let's you know, improve it.

B

If you see need um I haven't jumped in because I'm a little too much going on. But if you guys want to work together, it'll be amazing.

E

The few things also until now is that that the process or what to do is pretty straightforward and it's it's not too much time as far as I've seen, and it's only about getting uh finding some um so analyzing a bit trying to find where the project goes and in the end, writing something. So it should not be a big problem.

B

Cool, maybe this will help us build awareness of each of all. The different projects too, like I mentioned in the chat earlier today like find ways to bring the projects a little more together, rather than you know, 50 silos, that's a that's a non-goal, but just something that you know would be nice, okay, so moving to the next thing in status updates uh the working groups.

B

um So let's go in order. I guess here! uh Actually, no, let's go in reverse order, so artifacts uh make a good progress. We had a meeting on Friday Alex is not here today in our but I didn't ask him to come. So that's not him on him. uh The group has a document that they've started and they're tracking for different kinds of package managers and different kinds of Registries uh the protocols that they use for communicating with each other like to get a package and the internal formats that they use.

B

You know the differences I mean they're, even going to like things like npm and Python and stuff, but um oci and images too.

B

um We do have the pending issues, so I'll call that out again, like I do every week. It seems for the leads for that which they're already functioning is the leads Alex Salam and RAM Kumar chinchani. So please take a look at those issues and uh plus one them, so we can get them merged and and make progress there uh platform. So on platforms, I see Abby wrote in a lot of notes: I, don't I, don't know that! There's anyone else here.

B

Besides me and Colin I guess that are particularly deep in that group or Leanne too sorry, Leanne, um but but the platform maturity model is is coming along. There was we had a great meeting last Thursday uh where we we finalized. The five aspects and everybody is, is I mean we've got 20 30 experts that have been looking at these um and everybody's pretty happy with where we're at so yeah. Did anyone paste it? Let me paste in the the link to the current o2o version uh that Abby's been sharing um so we've.

B

Now, what you'll we've now divided those five aspects into five GitHub issues and and people are we put a little framework in for what we want people to do, but people are are creating basically a couple paragraphs about the each of those five aspects um check out the issues: I guess in our uh in our GitHub repo for the for the tag um or join the Ouija platforms, Channel and slack. uh There's a lot of conversation going on right now about each of those.

B

Anyone else Colin Leanne anything else to chime in on platforms.

D

There's a meeting tomorrow right with another thief deep, it says investment, deep dive, I'm, not quite sure what this is.

B

Oh so yeah, so that I didn't yeah. So that's for the individual aspects, if you're interested in in joining one of those, um so the person that's working on investment, which is about budget and funding and the different stages. A company can experiment. Organizations expect to go through as they adopt a platform mindset for that investment and funding and team composition and stuff.

B

um Let's see any other things, operators I know that Victor Liu is he on the call. Today, yes, hey Victor um I know he started an issue, so I'll go ahead and just mention it here about. You know iterations on the the operator paper number 439, so I guess take a look at that paste it in let's Victor. If you want to talk more deeply about, let's hold on that I think till the ideas section. So we can get to KCl.

B

If that's okay um and the last status update is kubecon, Chicago I need to update 408, but we did request a project meeting and a project Booth. uh So hopefully the project meeting will be I, don't know, however many hours they give us two to four hours.

B

I guess on that McCormick Place at the location of the conference on Monday I, know, there's a lot of events going on um that day, but here's another one, uh yeah I mean it'll, be a place where we'll, hopefully you know, depending on everybody's availability, bring together Luigi a working group leads to share. You know what they're working on and and recruit more contributors or get feedback. We try to bring some project folks in.

B

uh We ask them to talk about their project for five to seven minutes, and um you know just to further Synergy and standards, and talking about you know how these things work together, so yeah. So that's in progress uh and I guess I'll pause there. Our next thing is to have KCl present. So is there any other status updates that that we missed or that we should add here.

B

Okay, so um which was a way to mark this point in the meeting. Let me ask pennfei is that uh if you could take over um and present to us, KCl and we'll listen and uh yeah, ask questions eventually.

G

uh Okay, could you see me, uh let me shows uh slides, wait.

A

G

uh Could you see my slide?

G

uh Yes, okay, let's begin uh hello, everyone, my name is the engineer and on group today, I will share the whole KCl project.

G

um Next I will introduce this aspects: mingle, including background Concepts architecture and evaluation of the KCl project. Now, let's have an overview above the background.

G

uh Also kubernetes has become the standard of container or Gate Street. It still has some limitations. Currently, kubernetes is regarded as a platform platform which places a high burden on Developers.

G

Besides, there may be lack of a lightweight declarative configuration compensation method and the client side developers may need to learn a lot of knowledge and write a lot of yaml tool delivery, their applications and almost every developers. Application is generally composed of third-party software, such as database and the networks with a large demand for dynamic configuration and the customized deployment. This also brings High challenges for the stability and the efficiency of configuration Management.

G

In this context, we have developed KCl to reduce the burden on developers and improve the efficiency mainly include mainly through the dynamic configuration abstraction and the automation.

G

uh Kcl expects to solve about issues by hiding from and platform details by defining API abstractions, to reduce the burden and the manager large-scale configuration across teams without side effects through the configuration language and provides the ability to write, combine and Abstract configuration blocks, such as struct definition, constraints and logic.

G

Besides, KCl can be used and enhancement for configuration tools such as health, customized and kpt, through its Dynamic configuration abilities.

G

Next, let me give you some Concepts and cases uh the KCl project as a configuration language, the most important feature provided by kcl2 application and the platform developers and sres is a dynamic configuration management by abstracting code and using templates, validating and mutating resources using kcl's, Gamers and functions.

G

The platform can provide developers with an application, Centric and easy to understand interface.

G

In addition, Wednesday's interface cannot meet the configuration requirements, KCl or loss platform users to quickly expand and Define their own models, and these models can be reduced through oci registries.

G

In addition, tools, groups.

B

And figure deposit for a second um through oci registries. Could you just describe what you meant there.

G

uh Yes, uh uh we we have provided a package manager, tour called KPM and we can use KPM tool to write and publish poor uh for KCl packages to the oci registry, including Docker, harp and GitHub container registry. We and we can use the KVM to logging different oci registries.

B

Got it got it, so, are you bundling the manifests into an oci package? The raw manifests, the KCl files.

G

Yes, the packages mingle, including KCl uh KCl, fails and its dependencies.

B

Perfect yeah that reminds me of the work that the helm team is doing and it and it's been com and actually like image package from from Carvel yeah cool just wanted to make sure I understood. Okay,.

G

Thank you. Thank you.

G

uh Next, uh in addition to the about standard uh KCl form, we also support the current form to support the separation of the configuration data analogic and it meets the krm function, specification and supports KCl code storage in multi-sources such as oci and the git through the standard. Spec caselo not only supports customizing models and generating resources, generation resources, but also supports to mutate or validate existing time resources in a hook method to solve the third-party application.

G

Customization pinpoint yes, and whether using the standard standard, long KCl or the krm K7 form with support integration with multiple CI CD and the github's tools. That means the git robots can complete automatic configuration modifications through the KCl automation API, which can improve through the application delivery, efficiency.

G

uh Okay, the next, uh in addition to the about automation, Integrations uh in how to meet the needs of different developers. We also provide these Integrations.

G

Firstly, we provide a multi-language sdks to help users operate on kco in different languages and integrate integrate KCl into their own applications, and we provide the package manager tool to support multiple OCA Registries, such as Docker and GitHub container registry, and we are working hard to support the g-frog and artifact harm and we also support the migration of other schema ecosystems such as go Json schema, protobuf, open, API and terraform.

G

On the client side, we provide the k-cell plugins for existing configuration tools such as Google plugin, Helm, qcl plugin and the customized KCl plugin, and we provide the case operator two at the wrong time to meet different configuration management requirements. Instead of developing a cluster web hooks more and more uh besides. Kcl can be used as a DCM language to deliver deliver applications with tools such as cooling, stank and convela.

G

uh Besides, the way have provided uh many building models, such as promises, istio and om module. Okay,.

B

Just on this, one is I just want to see if we're, if I'm getting this right so KCl on the client side does it render let's say: you're targeting kubernetes you're, rendering down to kubernetes crds or resources and then you're calling Cube cuddle. Is that kind of what ends up happening like what's your final form.

G

uh uh Yes, mainly a programming language, and uh we support uh many of CD tools in the cncf community, such as Coupe cattle. That means they render the KCl fails to the cam resources and we can call the cattle to apply the yamo manifests. Besides the way, how provided the uh provided, the plugins for the existing configuration tools such as such as help and customize? That means you can use Helm KCl template command to render the KCl fields to the kubernetes resource resource manifest and apply the to the cluster and kind of slide.

B

So you could have like a pipeline that goes through KCl and then goes through Helm to do the apply.

G

B

Cool cool. Thank you. um One other question: you mentioned an operator so does the operator do that kind of stuff in cluster? Does it render the KCl to the krms in cluster.

G

Yes, in fact, we use a KCl operator to generate manifest at the wrong time. uh Besides, we can use the case operator to mutate or validate the existing camera resources and cluster. That means for most for most configuration management and cluster. We don't need to developer the new web hooker for the kubernetes cluster. We just use the case operator to do these things. For example, we can use the case operator to inject a set car to existing kubernetes resources.

B

Got it um is that required just curious? Do you have to have the operator, or can you just use the client-side pieces.

G

uh uh For the client side, we can use the key cell to render the yaml Manifest, but for some configuration fields we may can women, we may get the real value and clean the wrong time. Four days conversation we we recommend the use the case operator to get the real value of the yaml field to inject the value to existing kubernetes resources. Dynamically.

B

Oh yeah good answer: that's interesting! Yeah because Helm's got like look up in Helm actually from the client side. Does the lookup, but that means you have to be connected. When you run Helm apply or Helm install so I.

C

Guess that's a nice difference.

B

Here is uh that lookup would happen later. Does that make sense.

G

B

Okay, thank you. You could keep going.

G

uh Thank you uh uh next I, uh let me show you the architecture and the components of the whole KCl. In addition to the language itself, KCl also provides many additional uh help. Tools, such as fermenting testing, linked document and ID extensions to help users to understand and check their configuration or policy, and the the next slide shows the specific components of kcao, which mainly includes three parts: the language core torchings, including including ID extensions and the language tools and Cloud native Integrations, including KSL operator and cattle, mentioned about for different users.

G

They can choose the tools they like to use or integrate TCL into their own application.

G

uh The this figure in this page shows the main components of the language. Core KCl is a standard, spec driven languages and mainly includes four core elements: config schema, Lambda and Rule. It also provides standing types and plugins to help users achieve more stable and faster configuration management, abilities.

G

In addition, KCl views I.

B

Just asked one question: so the in that last one the uh the llvm stuff at the end it renders the configuration file right. The compilation result is, is some sort of yaml or Json or something yes right.

B

What does it mean like multi-targets was native like is there code? Is there a binary.

G

uh The uh uh I just explained uh four different targets: the KCl, the KCl Auto, the result of KCl is a yabo or Json uh just like Opa Regal. It's sport, multi targets such as wasam and Native, but the Opera, but Opa Regal always returns. The query results four, four different multi-targets.

G

That means we can choose different targets for different different for different scenarios. For example, uh uh if we have some wrong time, we can build a KCl code into a WhatsApp module and the user was some wrong time to execute the water module and it will output the sort of Json or yaml data.

B

Oh, so, you yield a binary which, when executed, emits the rendered yaml.

G

B

Okay, interesting, thank you.

G

uh Okay, uh this this figure means that the KCl views configurations and they are relationships and the dag, allowing users to query and modify configuration uh like they would in a graph database which can greatly improve the automation, efficiency uh and last. Let me show you some comparations and the project status of KCl and unlike existing configuration languages or management tools in the community, kco focus on the modeling feature.

G

It's a Unicode schema keywords and the grandiose static types are different from other languages or tools, and this approach brings high performance to kcao, which can be used in any scale. Configuration scenarios.

G

And whether it is a simple configuration or a complex kubernetes configuration scenarios in our Benchmark KCl is almost faster than other languages.

G

uh However, all of this, just as a beginning in 2023, we will still focus on the KCl languages, torchings and cognitive Integrations, and to do more work in 2024. We hope then KCl can reach the version 1 to help more developers to manage their configurations and policies.

G

uh Because the KCl project has has just been open source for one year. We have still released many versions during this year and worked with many contributors and maintenance from all the the world to build the KCl Community. Here we hope to receive further assistance from the cncf community uh for adopters. Kcl is the currently mainly trusted by Engineers from many companies such as on group, Huawei and uzen, and on group we have over 700 developers involved in 1 million. Tcl calls indicating that KCl has reached the production level in Newton.

G

They use the KCl open cruise to application abstraction to support the middleware configuration Management in over 100 clusters and in Huawei they use KCl to solve the problems, such as the difficulty in obstructing a terraform HCL and the length of scalability can be solved to meet the its SAS configuration UI definition and they write almost 2000 KCl mode codes and the three k cell plugins.

G

The adopters uh has has more and more, we, we just hope we can get more adopters in 2023 and the next year.

G

uh Finally, there are some.

B

Community I have a question on on that. One actually is Huawei like that's pretty interesting they're generating terraform HCL their output format like just to ask how hard would it be to use the same KCl top level schemas to they want to move from terraform and use krm directly. Kubernetes are cross-plane or something how hard would that is that a pretty reasonable thing to do with KCl.

G

uh uh Fall uh for Huawei uh for the Huawei Company. They they just have only uh infra configurations and uh with with thousands of resources. uh So they just they just use HCL to abstract their hundreds of resources, but they quickly fine designed the HCL.

G

Lengths of the abstracting abilities, because HCL has no schema structures like KCl they, they just use the HCL variables to Define their configuration interface, but the the ignota list of variables instead of a structure. Thus, if thus, they find the KCl project and they very like the KCl schema structure, and this use the KCl a schema to render their configurations and input their platform spec and the platform use. The KCl render results and and output the terraform resources and and use the telephone resources to apply their infra.

B

Cool okay. Thank you.

G

uh Okay, uh that's all thank you.

G

Anyone has questions.

B

It's pretty interesting, I I feel like we have been seeing a number of projects that are putting a layer above the final API layer for managing your platform or infrastructure.

B

um I guess most of what we've seen here in the group so far has been. You know, kubernetes crds. What renders down to that so things like cross planes uh xrm uh their compositional resource model. Essentially they try to add a you know, an or some sort of renderer over the top of krm. Also, um but it's not client-side. That's interesting um Abby's, not here but I, know her she's, also working on a project called craftix, which uh also it does some additional stuff over the lower level.

B

Apis um does q. Do this too I'm? Not that deep in queue but I know a lot of people are interested in it. Cue.

G

G

uh Yes uh uh personally, I like the Q programming language, uh uh but for uh for the Casio in other scenarios, the queue features and its performance could not meet our needs. This is because the queue uh uh the counts of Q is is its patching features. That means we.

G

We cannot use Q to write configuration to patch existing yaml manifests, for example, we have if we, if I, have a Helm chart, I can render the harm chart into a row yaml manifest, but we can't use Q to patch more fields to the Royal manifest such as image.

G

This is because the the call issue of Q programming language is two uh is to solve the configuration inheritance and the queue, uh and thus the queue proposed, its unification proposal and tools of the configuration inheritance, but in the kubernetes community, uh like Helm and customize, uh and the cool cattle patch, we can see it's very uh it's very normal, but for Q it's a, we can't do it and the second core issue of the Q programming language is to solve the type of the configuration validation.

G

uh But it is it to do this feature and it's wrong time. It, unlike KCl uh KCl, do do the type checking enter compile time. This means KCl has high performance than Q. uh Thus we, in this context we have developed the kco program in language instead of using Q directly.

B

Thank you, yeah that helps.

B

I'm gonna go on mute and uh other folks that wanna comment or ask questions.

B

Yeah ytt I see is another one that you've got there. That's from Carvel, which is one of our projects too so I mean overall I I will say that it seems that it fits in with our range of uh configuration management tools, uh some of the stuff you're doing, as is aligned with the artifacts work like bundling up the stuff and deploying it into a cluster um yeah.

B

This has been helpful insight. So thanks a lot and uh we'll post the notes to that issue. We'll we'll Point uh we'll put a link to your presentation here, so people can watch it from the technical oversight committee if they, if they want to or from the community um any other questions for us or feedback from people. Here.

G

uh Thank you. Just I have no uh questions.

B

Okay, well, thank you again thanks. That was great, very interesting ideas, uh yeah. So uh good luck and we'll keep up we'll keep in touch we'll keep working with you on this um next up in this section in the uh in our notes, so continuing down the list, there's a section at the end, which is for ideas, and please take a look. So please take a look, um but most of all today, Marcelo I think is here with us. Yes, um and he wanted to share an idea about documentation and Enterprise great examples. So Marcelo.

G

A

A

um All right, um um sadly, I I, don't have the time now to work on this project. But basically my idea was um I've been working as a consultant for many years now and I see.

A

Most of the companies have a very similar setup, at least on on the yeah on the main projects, uh for example, Prometheus and external DNS and self-management and and I was thinking like maybe creating a git repository with um very, very well explained um sort of examples driven by the community, which also uh all the project maintainers can contribute to um where we can basically Fork this to use in uh it can use it can be used in small companies and startups that can be only it can be also used, be used in Enterprise companies, um because, honestly, the the similarities yeah.

A

They are quite similar, I, don't know, I just wanted to uh sorry. I never shown this meeting before I didn't really.

C

A

It's supposed to be maybe next time I can come up with some examples of what do I have in mind. um But what do you think I mean I? Don't want to do it. Of course it will be a cloud agnostic and then there won't be any type of configuration um for one particular.

C

A

But if making some sorry no I, don't know how to say this and it will be definitely like open source and and we won't be a yeah I I, don't have to say anything English I'm, sorry.

B

Hold up yeah, no, no, making Champions.

A

A

um What do you guys think.

B

A

B

You wanna I. This feels kind of similar to the the Prototype stuff we've been talking about and like.

D

B

Or a page, what do you think so.

D

For context, Marcelo, what we're planning to do is to gather some uh platform prototypes and showcase them on our website and link to them, but we wouldn't be, uh we wouldn't be maintaining them. We wouldn't be doing anything really with them. We would just provide the platform for people to Showcase uh what they have like examples that they have so I think what you're saying uh could fit.

A

Under that, could you share the link of this sorry.

D

There is only a GitHub issue at the moment: there's nothing besides that, but I can add them to the agenda.

A

Then you can find in.

D

The docs later yeah.

F

A

C

Is something that.

D

Yeah we could probably fit under that I.

A

Think it's sort of like this I am reading it now yeah I mean it is within these lines right um something I I show in this. The call, because um I believe this. This would be a great help to um onboard people uh and make it easier to adopt. You know cncf projects um and yeah, because in the end, most of the time, I also find myself having to read a lot of documentation when all I want is a very simple example of how to install one particular project.

B

Yeah I think Marcelo, if you're on the right path and you meet people this, that's what I've heard you know if you're going to path and there's other people there too, then you're, probably on a good path. So I think what you're talking about is is great and we need people to help us out uh Leanne's trying to gather them up onto the site.

B

But if you like, create a repo we've been talking with other folks like uh what's his name, um salaboy Mauricio salatino he's been helping us with that kind of thing too Eloise rightbauer one of our chairs.

B

He had some ideas too, so uh we're definitely open to you know if you, if you create something- and one thing we want to have like, we don't expect people to maintain it for the next five years like if you have a sample, and we will that we won't own it, but we're kind of thinking we'll put it a link to it. On the page.

A

No, no, of course, um uh I can definitely submit uh PR to the tour repository, uh probably.

D

I think the only requirement we would have is that it would have to be based on open source projects right, like yeah totally, wouldn't be yeah, of course. Of course,.

A

H

A

Will come up, um I have to redact most of uh what I've been working on, but I can come up with the first example of what do I have in mind something to at least start working on. um So at least with the people who is in this ticket. Who seems to be interested in this and.

B

B

Yeah, thank you for for having the guts to come in and share that with everyone. So I appreciate that. Thank you.

B

um So we do have there's two more items in the please take a look category. uh I guess you know we we do have like five or six more minutes here. So if Victor I know you joined now, if you, if you prefer, we could just bump those and you guys could I didn't. You know we weren't necessarily planning for you to speak about them now, uh but we can give you more time to the next meeting or if somebody wants to jump in otherwise. I was just gonna say to everybody in this video.

B

Take a look at those uh chime in you know. Sometimes it takes a little while for.

H

Everybody to actually I probably don't need so many time just a couple minutes. If uh there's no other things, I can just go go ahead. um So if you look at the really there I think the it's about the timing right. So uh the if you look at the issue, I have created it's about one. The operator pattern was published and one the paper was published right.

H

So uh the artifact Hub, um which one is the one of the cncf's main artifact um uh Hub, where you store the operator uh artifacts, is only release due June of 2021 and then the paper was published, July 2021. That means, if you look at the stats on that website, most of the repositories and projects was created after that. That means a lot of the experience lesson learned and all that came after the paper was published. So that's why I think overall, it's necessary to now refresh that paper.

H

Then, specifically, the reason I was looking into this area is I I'm a database guy I'm, actually not a developer but interested in learning about the development operator as well. One reason it's hard for to run database on kubernetes.

H

Obviously this operator is security right and then, but really, there's not a good place to find You Know lesson learned and best practices security issue encountered when running database on on kubernetes, so using a user operator so I guess we probably need a survey to gather those information about how people are running kubernetes operators and designing operators.

H

So it can be secure right and another thing is just like what you saw um when you asked Penn Fey, why you use the operator right because there's actually quite a big debate right now about the operator and whether it's necessary, sometimes is in what's really the best use cases, and there are also um people like cross-plane, as you mentioned, uh Josh, and also kcp even mentioned that you know in a lot of cases, you probably do not want the operator so again that go back to the survey.

H

There's probably a need to create a survey by you know this group to really list some of the questions to you know why using operator, you know what you know why and why not right this way. We can get the information back and and probably update that white paper. There are also issues related to operator. One thing: I have found that being the database guy I know the scalability is a big problem.

H

uh I have found multiple places saying when you have too many operators which, by the way, store the actual uh metadata in LCD and that become a a bottleneck for running operators in in a cluster. So um and then, of course, there are also other technology. uh You know cutting glass things like you know, chat gbt in an edge Computing. A lot of you know running gpus with recruitment. So a lot of things that relate to how do you design operators?

H

Is it going to be able to leverage all those new technologies, so I propose whether asynchronously or synchronously maybe start a like a group to start creating a survey and that's going to take a lot of some time and then start getting information go to different places where people designing operators to gather those information, because when I I checked on the um the operator, uh the two Registries I, don't think. There's any information about this information. I just mentioned right, so we got those information back then uh probably download ropes few months down the road.

H

Then there's probably enough information to go into how to refresh that web paper. So that's my suggestion.

E

Yeah so I think that's my that's my point here. um So um I was one of the I was the third year after operator working group. When the thing has been created a few years ago, um it is a I think it's a good idea to refresh the white paper and elected you.

E

um We also thought about doing this. I think a half a year ago or a year ago, um but the traction for that is not very high. To be honest, um in my opinion, it would be really cool if someone would drive to say first. So if you want to step up and um do the do, this um I would support you in any case and I think the rest of the people who created the white paper up too um one thing I would not start with.

E

Is this very because um um when we created these white people, we also did some kind of survey. The the and the answers we got were not very much and I think we need a clear idea of what what we want to change in the white people or what we want to add in the right people before before asking questions.

E

What do you think about this.

H

Yeah um I I think um really from the past few months, actually learning in the in open source, Community there's so many smart people, so uh I'm, probably the least experienced, but I find that sometimes it's interesting to uh just bring up the question and raise the question. I mean I, definitely love to participate and uh if there's no one else, I can start by. You know driving that. You know the just collecting questions and uh it'll be also nice to. If you have the previous survey yeah.

H

If you can share that, um we can add to that and and see whether anything uh that is is new things that was not covered before um yeah yeah, just yeah. We just one way or the other. We can create a survey first and then uh go from.

E

There I think we could also schedule a meeting for the V2 um and also meet with Alex Jones, because I think he also wants wants a refresh of the white people and to find out where we, where we should take action and what we. What we want to achieve, though.

H

It's good thanks.

B

So Thomas I think what you're saying the next step Victor is to gather stakeholders or Thomas. Are you saying.

E

I think in this, in this case, we wouldn't have to create the working group, because we already did solved one um I think we should treat it more as a project in the tech and um yeah for, in my opinion, we should try to find out who is interested in in working on the white people. Again.

E

um Then, um then scheduled kind of a meeting where we talk about the things we want to. We want to change, and then we could do some work around it um try to find out what we need, a survey for and then diversary and then adopt the white paper.

B

So, for now that issue number four or the 39 yeah that might be the place for people to come and talk about these various things then go from there to a meeting.

E

I'm not sure there is also a.

E

There is also a operator working group in the in the cncf slack.

B

Oh, like a Ouija operators, Channel yeah, we.

E

Operate the white papers I'll take a delivery, operate the white people I'm, not sure, if anyone's in the, but we should um I think Victoria also part of it, and we could also ask there if someone has ideas want to want to collaborate on this and try try to get this forward.

H

That'd be awesome: yeah I'll follow up with your friend and Thomas.

B

Thank you. This is great um yeah. The survey that we did for platforms was also interesting, but I I kind of have learned from that. Is we needed to iterate on it a lot more and I think it should have been more concise and you know maybe we will for our next version, so yeah as long as we're working together, Thomas you'll, be there let's just iterate and we'll make the best decisions.

B

E

Should take care of when we are creating a survey. I should both have more than five questions, so we have to choose wisely.

D

I think Colin created a survey for the platforms uh working group ones. It.

B

Was a I, don't think it was Colin? No.

C

It was somebody else, I filled it out, though,.

D

B

We have been trying to, but not as much as we might if it was short and concise, and that's a lesson learned as we I think. One thing that we uh didn't do is get clearly understand what we want to learn beforehand and then you know create four or five questions that teach us that um so yeah. If.

C

We can do that for.

B

The next one I'll be looking out too. So, let's not rush it out. It's kind of on me like I, really wanted to get that survey out of the time of the paper, and so we got it in like two or three weeks. I got something out, but um we could do better. So, let's iterate, let's learn great.

H

To collect a list of previous surveys that could be helpful.

B

Yeah I could send you the link I'll find it uh for our for the platform stuff.

B

Okay, we're a minute over so to be respectful of everyone's time. Thank you all so much for joining we'll get the recording up shortly and uh see you on GitHub and slack and uh everywhere else, or maybe at Niagara Falls, because that's where I'm going right after this.

D

B

You all right, bye,.