Cloud Native Computing Foundation Technical Oversight Committee, 22 Jun 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: CNCF TOC Meeting 2021-06-22

Description

CNCF TOC Meeting 2021-06-22

A

Hello, excellent, we are recording, so let's get started all right, so we'll start with the reapplying projects, and I think yarp was the first on the list.

A

um So I, as I understand this applies to yarp and it also applies to another project on the list.

A

I think this is library components rather than standalone, and please correct me if I'm wrong in understanding that and I'm not sure whether we have projects that include client libraries, but I'm not sure we have any projects that just are library code and I feel, like that's a you know,.

A

Something we should consider whether we want to.

A

I think, first of all does anyone disagree with my belief that yarp is library code rather than a standalone project? Hey justin we've got justin as well. Sorry, I'm late. Sorry, we were just starting to talk about yarp and I was saying that I believe that yarp and also another project which I will will come to at some point. But another project on this list is library code rather than a kind of standalone thing which might be okay, but I'm not. I couldn't think of anything that is library code.

A

Might we might have client libraries as part of a project, but I can think of anything else. That just is you know and whether that is something we want to consider in terms of whether something is appropriate for cncf or not.

B

I mean the first question: is why wouldn't it.

A

A

I suppose that my thought process on this kind of came through. Well, you know. Does this particular solution? That's um you know. C-Sharp sharp. You know which is fine. That's it's perfectly valid for people to write applications in csharp.net, but do we want to.

A

C

I actually I I just thought grpc is a library.

A

Is it yeah yeah.

C

A

B

C

I actually I was actually agreeing with you and then I was trying to think of reported, but I think grpc is strictly a library. Isn't we didn't bring in a spec project? It's it's a multiple libraries for different languages, but it is.

D

C

Library and that's how you consume it and that it's it's unusual, I think, because cloud native projects are often consumed as services or shared services, but actually it's not totally out of.

E

So one another yard stickers um for grpc. We have so many different projects that use grpc right uh here. Yarp is coming in and there is no consumer of vr as such. So that's the distinction between yarp and grpc. I guess.

F

I don't think we really use that yardstick for sandbox as much since it's supposed to be for intubation and innovation.

A

True, but we have to believe that it is a good fit for the cncf, and I.

F

D

A

I just when I- and I appreciate that grpc maybe already is the precedent, but I think I was. I was worried that we're gonna have. It opens the doors to a lot of.

A

You know library code that.

A

Fulfills some small part of the cloud native, you know functionality and whether or not we really want to be drawing a line outside of that I don't I don't know. I I think I'm partly struck by just the sheer number of projects as well and and wondering whether we're.

A

uh Making things easier or making things more confusing by including yet another class of project.

A

No pun intended on the name.

C

A

G

C

Kind of justification for why it's not in dot net foundation and is kind of weak.

H

C

Because it's it's, I mean, there's the comment and the link to issues which is just kind of like we want to make it more. We want to bridge.net and cloud native, and it's like that's great, but um it still feels like one component out of many is in would be in cncf, and the everything else in the net ecosystem is in the net foundation, and it just feels kind of weird.

F

I

That seems like a reasonable.

F

E

Well, the other way to think about is like how many who will be using this and how many of them will be using it or like how is it going to spread right um to be useful in the cncf ecosystem, and do we see projects depending on this?

E

um That will essentially go into any of our usual deployment models, and things like that right? That would be the fit that we look for, and I don't see it yet.

A

So perhaps, uh as chris is suggesting, the um the net foundation might be a more suitable.

A

Foundation for this.

A

Why can't I see see the? uh Why can't I go back, it's basically how it's like I'm new to computers. What.

G

A

G

Just sometimes.

A

Why can't I see the uh maybe that's what I need. That's what I need, except ah there we go and yet hold another tab right.

C

Also not clear to me who the maintainers of this project are and how what its governance process processes.

A

It does not feel to me like it is a very natural fit.

A

C

I mean it's very much a microsoft project with a microsoft cla right now and then I'm not sure it's got yeah non-microsoft usage.

E

If we don't say no to this, what else will we say no to right, yeah.

B

A

Do we want to hold a vote? Shall we does anybody feel that they would be voting? Yes,.

J

E

We should just do a wordless and say no. We.

A

Probably should yeah okay, I.

J

A

J

Well- and it's not just um the microsoft comment either, I think it's just as it's not a good fit for what we see being a viable project.

A

Yeah, it feels I think this this whole question of you know how it fits with other cloud native. You know if there's nobody else already using it. I mean the.

A

I do think there is a question mark over the library projects, dot.

I

Net feels like.

A

A more sensible home for it all right.

A

Is that enough? Okay, uh what was the next one on the list of um repeat see please pixie.

A

So pixie have definitely when pixie came up last time. I was concerned about them having uh linking their uh github to their self-hosted. Sorry, their their own hosted version of pixie and they've sort of they've made some improvements there. They they've it's not like on the first page of the um github, but there is still and- and I kind of understand why they have it right.

A

I think it's quite hard to set up your own hosted version of pixie, so they're kind of saying you know for an onboarding experience come to our version that we already host. It is not clear to me who is operating that, and you know it is still.

A

Not unsurprisingly, it is still linked to quite um heavily from the kind of getting started.

A

Documentation um do we have any other examples of something where the project hosts a version of the solution like this.

K

There's like a demo backstage server, but that's a little a little different right.

A

Well, it's it artifact.

I

A

Or not, you have to sign up.

I

A

K

Artifact hub is the only.

A

Yeah, an artifact hub is operated by cntf right.

I

A

Yeah and could we operate this.

A

Hosted version of pixie.

K

It's a good question potentially did I have to figure out what the costs are, but having other vendor, I'm assuming they're done in a way where other vendors could go operate this on their on their own. If they wanted to.

A

Presumably yeah.

A

Yeah I mean they, they have got like the full. The repo has, apparently everything that you need to yeah.

I

A

Run your own version.

L

Does anybody run it already other end users? They do it.

K

uh I know there's an aws collaboration, I'm sure new relic who bought pixie is doing something with it. I'm trying to go.

A

I assume you relic are essentially hosting the.

K

Yeah, probably they paid for it yeah and I think aws said something that they plan on uh at least contributing to the project yeah, which.

A

Is cool and I I I would really like to find a way for this to come. What I'm worried about is the precedent of oh, you know you can operate your own website and essentially have the cncf funnel you, people who sign up for your.

K

That we would not allow that is.

A

K

That's against our policies. You can't have like a sign up for them or or harvest harvest emails. That way.

A

K

If it's a demo server like a backstage thing, that's completely cool all maintainers have access to that.

K

My assumption is here: they probably just do like a demo thing running on the cncf cluster right.

D

K

Probably um what I would do and I'm guessing new relic aws others- will offer probably hosted sas based offerings for for this.

A

Yeah yeah and probably like other things, if there's like a hosted version of something their project page, can you know so long as it's prepared to list all of them? It can list them yep.

K

Alphabetical order, just like envoy and a bunch of other projects that that point.

E

What what did we ask them last time uh and they came, we had to revisit.

K

Oh, let me go find it. We.

C

Asked about whether it was all the reapers being open sources.

K

Which they said.

C

C

K

And I think they were forming some type of governance structure.

F

F

A

L

A

I'll I'll show you the thing I mean you know it's not. I. I completely understand why this is here, but you know you go through quick starts and um you know it. It tells you you can. Oh, I can't even find the version. Oh here we go, you know, there's this pixie community cloud which saves you from deploying it and which is.

C

On with pixie.ai, or maybe pixielabs.ai or yeah, which are not.

D

C

In the domains that they're going to transfer to cncf.

A

C

So then, you end up.

A

Here, with a login.

C

And so the question is: is with pixi and pixie labs? Are those domain what's going to happen to those domains in the name, because if they they get a hoster a pixie service with a with pixie and his name and other people? Don't then that's a problem.

C

I mean if it's just they manage service for this and it's all open source. That's fine, but they can't presumably do under the trademark of.

K

No it would, it would have to be, you know, new relic blah, you know with pixie or four pixy type type setup.

E

Yeah, so the only other thing I can think of is a rename pixel to something else, and we we get to host only the sources, we don't care about uh the website or whatever service they are running.

C

Well, I think we need to just get them to clarify what their plan is going forward with this and what what the brand, what the branding of this would be and that it will be a pixie service, not the pixie service.

K

The the trademark they're for sure going to transfer over that I have clarity yeah, it's the question of, like you know their whether this demo service runs on the cncf cluster or new, relic or aws offer their own like hosted things which is fine as long as they name it. You know appropriately right.

B

One thing I don't know if y'all have noticed, I was just flipping through this, the docs when you start on the dock, that is linked from the application. You are on a dev site if you click on the link to the the hosted. If you click on the pixie community cloud, you move over to ai site.

B

Yes, the documentation looks exactly the same, but it's all under this other domain.

C

Yeah and that domain is not on the list, that's been transferred, whereas the other ones right, x, dot, dev and pixie.io are listed, has been transferred, but I'm not not there, but.

B

All the rest of the docs are under that dot ai and they look exactly the same. The table contents doesn't change everything else, so it's pretty sneaky.

C

Well, I mean, I don't say sneaky, I think it's because this was this service that was acquired.

D

C

Being open sourced, so I think it's okay to have transitional issues, but we need a clear delineation of what's happening. I think.

K

Yeah we could ask them about.

D

K

I think pixie ai was the name of their company. I believe.

I

K

Was purchased generally, we would ask for that domain to be transferred over as part of the process, but if they're not using it in any fashion, then we don't care as long as it's not confused as long as it's not confusing to end the end users, they could that it means no longer being used. We don't care, but the justin points. I view this as like a transition thing where you know as part of onboarding we kind of deal with this whole um thing.

K

There's you know, we've run into other projects that we've had to explain to them. How, like here's, you have to list all the services that you know people are building upon in alphabetical order and a fair fairway and so on. So I don't see anything too crazy outlandish here. So.

A

So do you think we can go to a vote with the sort of proviso that they would need to resolve this community-hosted site and the use of with pixie.ai and the you know the sign ups here.

K

On the harvesting emails like you, just can't do that right. There would have to be some other. You know if, if new relic runs their own service, they could go for to their hearts content right, but it can't be.

K

um You know something that they run with, that name right.

A

So should we have a vote on the assumption that those things are resolved as part of transition and because they have to be because otherwise it couldn't be in the scenes. Yet yeah.

E

A

A

So I think the next uh we can probably talk about them together, there's mescheri and smp, which was service, mesh performance, very cool.

A

um So when we discussed this last time we were um talking about how um you know we have the service mesh interface already we have service mesh performance and then meshery, which kind of implements some of these things, and uh we went to sig network where we have the interesting scenario that um sig network is essentially being run by lee and he's also involved in all these projects.

A

So I guess the good thing about that is, he knows what's going on, um but he did go out and have a conversation with lots of maintainers on the service, mesh interface project and mesherie and smp. I think there are a couple of things that we've learned from that one is that smi is potentially an at-risk project. I would say it's not going from from what I hear. There are some issues there in terms of its support from its own maintainers.

A

um I think there is some potential for smi and snp to be kind of better aligned.

A

There is interest from people outside of the existing maintainers and some other companies who want to get involved in smp and want to see it in the cncf before they do so um and then mescheri.

A

The project, I would say, is very clearly saying we are much bigger than both smi and snp, and we want to have a life independent of those two spec projects which I think is not unreasonable.

A

So my own feeling is that yeah we should treat measuries sort of you know separately.

A

Smp seems like it's one of these areas of experimentation and collaboration and that we should be signaling to smp and smi that we want to see. You know if we're going to have like a definition of what service meshes are, let's try and make.

A

I don't think we're going to see one of them we're not going to see both of them incubate without a very clear agreement between the two did that make sense.

E

So do we want to vote for measuring separately right now.

A

Yeah, I think we should be voting on both measuring and smp as separate sandbox projects and then, if they pass, I think we should be signaling through sig network that we want to see smp and smi.

A

You know not diverge at the very least, not diverge.

A

Does anybody else have any other sort of points they want to make about either of those two projects.

A

Okay, should we do votes for? Let's do measuring.

A

A

Erin, I think you've sent me a vote directly.

G

Oh aaron, if you send it over in the main, we will pass there we go. Thank you great.

A

And let's do votes for smp as a separate project. Oh and I've sent it to everyone.

A

Let's try that.

A

A

All right so now we can go back to the top of the spreadsheet and the new applications, so porter lb um and noting that we already mentioned metal lb. I think there's there's a a couple. Maybe three, maybe even more uh load balancer projects all throwing their hat in the ring at the moment.

A

A

Porter name collision seems to be a bit of an issue.

C

I mean I'm sorry, I have an issue with them, not filling in the form.

C

I I know what you mean, it's mostly blank, I'm sorry, that's.

A

C

We made the process really minimal, but if you can't, if you can't do that, then I'm sorry.

A

All right do we want to vote for that, or do we want to just bounce it back from, say a you would have to change your name and b. You would need to complete the form.

B

I think that, yes,.

A

Yeah, because if they did both of those things, we might in future accept them.

B

J

J

But given how many different load balancers there are, are we going to have a stronger differentiation at the sandbox level than we would normally or not.

L

Like they haven't, submitted the difference uh between them and other uh a lot of bouncers for better metal like metal.

A

L

A

Guess I guess what we're saying is we can't you know? We can't really consider the application very thoroughly because they haven't really filled in the form and it's always possible that there will be more competition and we might not accept them in the future, because you know there's no there's no guarantee and there might be too many load balances.

L

So we have this, uh we have this field explanation of alignment and overlap with existing cncf projects and it's optional. Should we make the requirement, I'm not sure why it's optional on the forum, the description is an option. Is it the description is not, but.

A

How did they get that.

L

M

Yeah we should, we should ask them to fill in the description part and as well as a explanation of the overlap. I think, although they are optional, but I think they are very useful information for qc to make decisions. In my opinion,.

L

I think so too, we have so many projects and it's important to understand the overlap with the existing projects to see how it differentiates.

A

I agree yeah. Can we make that alignment field into a non-optional.

G

I will check in on that.

A

All right so should we move on to cubevela.

H

Just have one question on the previous point: are we sure we didn't override the content after the submission? Because I think this is it the one.

A

Oh, you think maybe somebody has inadvertently.

H

Changed didn't then do or something.

C

H

C

There is a deleted. Okay is that.

H

Because if you tap space on any so it just overwrites the whole thing. Oh.

G

That's interesting.

C

C

uh But it appears to just say: porter lb is an open source load, balancer implementation designed for bare metal keeping at these clusters.

A

How did you find that.

C

All right, right, click and there's a show edit history, new yeah new feature in google docs.

D

C

Created just in time for this meeting.

K

Nice all anonymous users deleted it. It.

D

Doesn't have a way of restoring it, as I can say 11 days ago, so they did.

H

Yeah, I was just testing now and actually you can just mistype something and just override them.

A

K

Still think it doesn't change the naming issue right so.

A

Yeah, so they would have to change their name.

A

I mean we. We could vote on whether we want to accept them on the assumption that they do agree to change their.

K

Name up to you.

A

It's it, you know, it's got, it's got some interest. I think there is the the fact that it comes from a cube's, the cubesphere distro.

F

Does anyone know the major differences between porter, lb and metal lb and any experience I see both of them have applied.

A

I've I've, I have let's say, come across not used myself, but I have been sort of adjacent to some usage of metal lb, um which I think has some some good points and some I have heard people concerned about scale, but you know that is probably indicative of the fact that they've been using it at some scale. So it is real yeah.

A

Oh dims you're, sending us a link to the comparison of the two is that.

E

uh There's not much there, but uh you know how is it packaged and how is it uploaded that kind of stuff.

L

Should you request project owners to submit the the difference or explain the difference.

E

But then we have to go tell both of them right, like.

H

A

Yeah- and we have usually you know, we have that idea of the sandbox being potentially uh space.

E

A

A

I gotta say something else I didn't love was the fact that um they put on their readme. They put the cncf logo because it's been accepted to the cloud native landscape.

A

I haven't seen that done before. It feels a little bit cheeky.

M

C

I'm also a little bit unsure about, but because it says it's a sub project of cube's fear is it I mean it. It cannot be a subproject of something, that's not in the ccf once it's in the cncf and how independent is it and is cubesphere? Potentially it wasn't enjoying the cncf.

C

Because it seems like if it's a subproject, then it's saying it's not independent right and it.

E

C

E

It's a consulting company justin, who is uh cube sphere.

C

Does it I thought I thought it was, uh I think.

A

It's a distro, I mean it might.

C

Be something it is just. It is a destroyer, but.

M

It it is a company added is sure you can think of it like a rancher in china, something like that. Yeah.

A

That's fair enough.

E

The only other thing that I can add is um I I haven't. I didn't run across the name in any of um kubernetes related mailing lists, or you know, repositories or anywhere else. So it's brand new to me.

A

I guess also, I mean there's a there's a language thing. I mean it's a sub project of cubesphere, but if it's, if they had phrased that, as I don't know, porter lb is a project created by cubesphere. That would be.

A

That would be okay and we would assume that means well. They they were the founding makers, but they decided to contribute it that that would just be a transitional thing. Wouldn't it.

E

So do we vote now with the caveat that we'll work through the separation concerns later.

C

I mean the owners are a subset of the cube sphere owners.

A

But I think that's okay for sandbox, I mean yeah.

C

Yeah, if it's intended to become independent.

E

And it's I I think I saw um separate uh installers for standard openshift, kubernetes and k3s, and things like that.

H

It's writing a main page yeah.

A

Yeah see yeah kubernetes, k3s and cube sphere right, yeah.

E

So from that point of view, I think it's it's okay,.

A

uh What's it called for water subject to name.

A

A

Aaron neil's came to me.

J

Again, don't know why the chat keeps doing that she's, really money, liz.

A

It's not like erin and me are having like some really exciting back channel.

J

No, I click chat and then it just I don't know it suddenly selects you so.

A

J

Okay and then it defaults mine to sending back.

A

To you and yeah.

A

A

A

Do we need eight positive votes, or do we need a simple or two-thirds majority of a quora vote.

G

My understanding is that we need eight out of 11 for the voting um okay to count.

G

So, um given, as I do not see a passing vote in here, suggest they change the name and reapply okay, okay,.

G

A

I think I saw somewhere that this came out of crossplane.

M

Yeah my team is involved in this project and this is actually originally a crosstalk sub project and then moved to move of the independent project. Let's see like that,.

H

The last commit song looks pretty familiar the author.

A

C

um But so this is being donated independently of the om's bag. Is that correct.

M

uh Everything will be donated together. uh Basically, own spec is already coupled with this project, like the api implementation.

A

Yeah, it does say in the note somewhere where, where.

M

The first column.

A

M

A

um The om spec itself is now driven by cubelet and considered as part of this donation.

M

A

Yeah I hadn't come across the open application model before either, but.

C

I have um but yeah I I didn't I I didn't realize. I think I hadn't come really come across cube furler as part of it. So I kind of the other way around for me.

E

I ended up running into this uh when I was looking at like dagger.io and shipper.iu, so it's in the same category right uh hurry.

M

Yeah same category, basically like so a platform service on top of kubernetes, as well as the hybrid.

M

A

All right shall we anyone got anything else to add, or should we do.

E

Votes, uh how do you be like the people on this team.

C

With a plan b to rename the org to cuvela, then rather than om dev,.

M

uh That is acceptable. uh I.

C

Mean I'm not asking it's just. It seems weird to accept it right like it's it. It seems to if we accept it as kybala, it seems more like better that the organs give valuation we accept it. As am then it's fine to be called. How I am I I I don't have any preference, but it's like it seems, it seems just seems a little bit inconsistent.

D

I'm talking confusing.

C

Potentially, between which of them is the dominant part of it.

K

Yeah does that everything.

A

Om has a life on its own is. Is there a a a potential future where om is a spec and cube valor? Is you know just one implementation or something.

M

So they are actually coupled right now, that's why uh they are coupled together installation, basically like api implementation. So there's no other implementations. I don't think there's motivation to promote any other implementation based on that yeah. So the the original the current plan is that the whole relation moved to uh since actually it's including every.

D

M

To make kubernetes to work, including you can see, there are mission controllers, dashboard, it's command line tool, so that is the original plan.

E

So how much rework do you have to do to uh I'm sure you have oem dev in the imports or something right so that that's going to be your issue.

M

Yeah, so renaming will require some work. That's actually the reason we didn't do that before, because it will break some dependencies, but if we feel it's necessary, I'm totally. Okay. From my perspective, just some technical issues to solve, there's, no political or any non-technical blocker. For that.

A

Chris asks a good question, which is: does that mean that everything under om dev is.

K

And harry said.

A

K

And he said yes.

A

G

Did you move to a vote.

C

I mean I'm not worried about renaming, I just think it it might lead to confusion, it might be worth considering that's what was.

C

I know it's painful.

M

Right we discussed that before so there is a plan to name that, but some maintainers raised concerns about that, because um we have production environment that rely on that path which will make it. You know take some time, but I think we can raise the issue formally to discuss that. uh I mean in github to see if folks are okay to renaming the github report.

A

M

A

One one possibility would be to sort of consider the project as like, om and vayla kind of as if they were a combination, project or sorry, om and key fader.

I

I feel like the naming.

E

Is not the most.

A

Important thing all right.

E

Moviemobi docker docker is still going on right now.

E

K

Spire type of.

M

Yeah I can understand the organization name is essentially, for example, like clone native application system or thing like that, which will make sense to cover what's inside currently in that organization. That is actually ideal. Also, I would try to ask the team to raise the issue discuss what should be the right name for the organization github organization during the sandbox onboarding process,.

A

All right, the next one is k8 up or ktop, or I don't know how you're quite supposed to pronounce it.

E

They said ketchup.

A

A

That's very good.

A

I just wondered how much this is really a whole project and how much it's just a wrapper around. I think it uses something called restic to do. The actual backup like does a.

A

Is it enough to build an operator around something? Is that really a whole project? I don't know.

H

Yeah, I didn't have a full look, but actually we we are looking for a component like this, but that does both the backup and restore uh it's it's it's not as trivial as it sounds uh in fact,.

F

Yeah I'll plus one to that, I think uh valero is another uh solution in this space and they also use restic.

A

H

Yeah and in both cases, we have issues with uh with actual implementation of the persistent volumes and how much support they have for restoring from snapshots and things like this.

H

So a general solution like this, it's uh it's not too bad,.

A

Valero uh rings a bell to me: dude is it in the sandbox already? I can't remember.

J

I think so it might have graduated out of sandbox. It's definitely something we need, though, like in terms of our strategy of like managing data. At that level, we don't have really hardened solutions in our landscape. Around these type of things doesn't.

K

Mean we should accept this.

J

Based on that sorry go ahead. Chris.

K

It's not in cncf, it's I think it's vmware still has it, but dims could continue.

J

I thought they gave it back when joe was on the cncf. I guess I'm imagining it.

K

Probably another special.

J

E

uh If there is, uh you know, consensus for sure I'll ask I'll I'll check check with uh folks and get back.

A

There is a um chris's comment about having to change the license. They do actually say in their kind of column, p. We're currently using the bsd3 clause license, but we're open to changing the license to apache too. So I think that would be uh that that sounds.

C

K

Only eight.

C

Only 18 contributors to get signed off to check.

K

Yeah, so not not too painful, but still painful.

A

172 stars is not massive, but it's not.

A

B

E

So this project depends on arresting with a w, and I don't know if that is part of the.

K

Submission as a third-party dependency, can you be a little more clear what you mean here, tim.

E

um So there is a one I pasted link.

A

W-R-E-S-T-I-C, I'm interested and that's part of the same organization.

E

Yeah I was looking at the architecture and it mentioned that the operator is k-top and ketchup, whatever the rustic wrapper is wrest. So right.

A

So I guess there is a question over whether this donation includes the other one too, that rustic part the mooristic parts. Does it say already.

A

I guess we would want clarity on that. Yeah.

H

Yeah, I I talked to them for the cross plane uh review as well, and- uh and they mentioned they are- they were submitting uh this this project they working on giving more information as needed as well.

E

We need one more piece of information before we can vote. Yes,.

A

Which is whether or not this rustic thing is included, I mean the other. The other possibility is to vote subject on to that being included.

C

In general, we've asked for a confirmation about every I'm, not sure about. What's in and what's out, yeah.

A

Okay, all right: let's, let's go back and ask that then cube vip.

A

This is another load balancer by the way.

E

Yeah this one is being used a lot by a lot of people so yeah. I have.

A

Heard this one yeah.

A

385 stars, I think yeah.

E

Pretty popular with the crowd that does the cluster api stuff.

E

C

Not got a lot of contributors, I mean.

H

C

C

E

Very enough yeah yeah.

A

C

I mean it's mainly it's mainly it's mainly down.

E

Yeah and yeah from vmware, um you know he was working with he's the second one, um so he's the one working with them to get it working with cappy.

A

Yeah yeah, the other contributors are pretty small, but you know that's.

A

That's part of the whole reason for sandbox. Isn't.

L

G

Justin, I'm assuming that is in fact a plus one.

C

Whether I actually oh yes, yeah, it's just pass wrong with a chest. Shift key, keep the confirmation very.

A

Enthusiastic all right, the next one on the list is cube. Dl, uh which looks as though it's a kind of cube flow alternative, which is kind of intriguing um another alibaba project, very prolific.

G

Quick time check, though, we have roughly five minutes left in this call. Yeah we'll just have to get through what we can yeah.

A

Anyone uh I mean 165 contributors and sorry 165 stars and seven contributors is quite feels a little bit early.

H

Yeah I I just started having a look at it today, because they are very vocal about how the features they have that cook flow doesn't have. But then it's true that it doesn't have a lot of contribution. Yet so you.

H

F

I want to caution against again against using that as a metric for uh sandbox.

K

Widely deployed inside of ollie, it probably has some usage yeah.

M

So this project is also uh my team is partially uh for facebook. So uh the the overlap with puberty over kubiflo is the uh tensorflow job crd. uh So the cool video had its own crd for tensorflow.java, which is a similar but uh had differences with kubiflow, but for other parts they do not have overload. That means the kubiflow components can be used with qbdl, for example, it's workflow and meantime kubidio provided several other features which do not exist in could be flow. So this this is the relationship between these two projects.

M

It actually has explanation. It's really, for example, could deal provide tuning crds a lot of tuning theories besides chaining and serving, and they do not actually implement stirring. So you can use kuby keynative to do stirring, just like it could be floated so that so, basically like that.

A

Where, in the, why do you want to contribute question? It says we at alibaba have widely been using kubernetes to manage large scale, deep learning workloads for many years, but doesn't actually say, use cube, dl.

M

I think that means they are using pretty healthy. That's the only option they have.

A

All right: okay, okay,.

H

All right, it looks super promising, but.

H

As we mentioned before, but yeah, the the features are like ticking. All the boxes.

A

Do we take this one to about then.

A

um All right, we are almost up at time. Let's see whether the next one looks like a really not oh, the.

E

A

Is um crustlet.

E

Yeah, so the thing here is: go one um go to the ninth one: seven and nine go together, uh less oh right, chrysler and oh yeah and crater yeah yeah. So the question that we need to ask them is: can they work as a single project? I think, and not as two separate projects.

L

I wonder if they see creator as a sdk library for other rust projects, not just for crust. Lid.

A

I think there is a question out there over over libraries again.

E

Yeah yeah so start there and then we'll see where it goes right.

L

E

L

We can also ask a question what what other projects uh use crater besides crosslit? Is there an adoption right.

A

So does that mean krator is a dependency for christlet? Yes,.

C

That's read: it is, and and written by the same people and submitted at the same time. So it's not.

D

C

It's not, they haven't adopted someone else's thing, they've written their own thing, but they're submitting them separately, which is yeah a little bit unusual, because normally you would put them together and maybe spin them off later if it seemed worthwhile rather.

A

C

Them separately and then yeah later.

A

It's going to be awkward. If we, you know they can't really be decoupled. If like it would be awkward to have, let's say chrysler in incubation but crater in sandbox. That would seem strange.

A

um Yeah, so maybe going back to them and asking why why creator isn't part of christmas as a project? Okay, we're still at seven for qdl. So if anyone else wants to add a vote.

E

I

That, let's say erin sent it to.

E

Me they did answer that question saying that um in the kubernetes ecosystem there isn't a good library for um talking to kubernetes and that's where this project crater is useful. All right.

E

I I would still say they should come together as a single project and then we'll see where it goes later.

K

And by a single project you mean, like all, live in the same github.

E

Repository yeah same github, arg.

K

I

A

All right, I think we are up to time. uh I think, because we have several left. It means that next time we do a private meeting, we will go through what's left.

H

On the spreadsheet.

A

So, apologies to anyone we haven't got to two weeks time.

G

I think is it two weeks a week's time I will check to make sure there's nothing else on the calendars, and I will put that in the follow-up and also with a note on the spreadsheet. So thanks right have a great day.

F

M

Thank you, bye-bye.