Open Feature Project, 16 Mar 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OpenFeature - Project Meeting, March 16th, 2023

Description

OpenFeature website: https://open-feature.github.io/

A

B

How's it going.

A

Quite well cannot complain.

B

Good, although I don't know if those are mutually exclusive,.

C

B

Can it can be going quite well and you can still have complaints, I.

A

Said no, no, it doesn't know it's it's perfect. It's everything is running smooth. It's fine tomato seeds are planted. So it's a good thing.

B

I'm assuming indoors, or is it just a lot nicer in Austria than it is here no.

A

No you're, normally you normally start planting your your tomato seeds, indoor.

B

A

Makes sense so now for three weeks or so so we have a certain size and and April or May they will get outdoor. So, okay.

B

Yeah, nice, Hey, Thomas, hello, hello, you'll, be at kubecon right, Thomas cool all right, looking forward to it.

D

Yeah will be great: when are you when.

E

Are you coming from to Amsterdam.

B

I'll, be there all week, so I'm gonna I'm, taking the red eye from uh on Sunday night, so I'll be there like Monday, morning-ish and then they're all week. Yeah, probably lots of opportunities to get together and then I'll I'll remind the the community at the meeting today. But we do have a slack Channel um for keep gone. So if you wanna, if you haven't already feel free to join that, we'll try to communicate like where we'll be at and the booth and all that kind of fun, stuff, yeah.

E

I'm already on it.

B

B

Little time to join real, quick I think we already we have a decent agenda already. Let me uh share what we have already and if you don't mind, just adding in your name as a participant, give it a minute or two I know: Todd is at least joining and he's got a few of the first topics, so we'll have to at least wait for him.

B

You can welcome hey.

F

B

All right, I, just centered it I'll send it again. um If you want to join the meeting notes, real quick, just try to add yourself as a participant and I'll follow those rules, and then myself right now.

B

Alrighty well um I guess we can skip the uh introducing newcomers. I recognize everyone on the call. Well, actually Simon I, guess you're! This is probably your first time so yeah. Please, please go ahead. Okay,.

A

I'm not sure what I should tell hi. My name is Simon I'm, a software engineer, Dyna trees and I'm helping a lot of our communities when it comes to tooling Etc and that's the main reason I'm here. So maybe you have seen me in the slack Channel already pointing out some Polo pull requests and suggestions and I'm looking forward working with you guys.

B

C

Right yeah thanks.

B

Simon good to have you here and yeah. He like, like you mentioned Simon's, got a quick topic: pretty cool stuff, he's working on so uh yeah. Definitely looking forward to that side that part of the conversation today uh and then Andrew I'm, not sure uh yeah, hello, I, think.

G

Hey guys you're new, hey good. How are you doing I'm not doing too bad yet sorry, so yeah I'm new hire over at um over at tabletics and devcycle, so uh Nick send me the invite and said you gotta check these guys out. It's gonna be great, so uh yeah yeah pumped to be here and check out. What's up, what's going on all.

B

Right, perfect: well, welcome good to have you I'm gonna Center, it again in the the slack trap, but those are our meeting notes and feel free to add anything to the agenda uh and then, if you're willing to add yourself to the participant list. So we have the the history there and I guess we can go ahead and get started.

B

uh Looks like first up. We have Todd.

F

Yeah so um I think I'll be fairly short and sweet, but we, like I, talked about last.

D

F

We've obviously had stuff and works for the client side for a long time. um We did recently merge two of the three uh open feature: enhancement proposals that um that I I characterized as like critical for the uh for the web and client support in general, um and also we have a draft PR, that's been getting. uh You know a Resurgence of attention lately. So with all that in mind, I figured it was probably about time to actually release an artifact. That kind of would exemplify some of this stuff.

F

So we did publish the web SDK um it's being published right now from the JS SDK. uh So look in the JS SDK. It's now, basically like a mono repo um uh and I I realized Mike, I'm I'm supposed to be the Scribe for this meeting. But while I'm talking, if you wouldn't mind uh filling in for me, I hope, that's not too much of a problem. Thanks um yeah so feel free to go check out. The web, SDK uh I think there's only been two versions versions of it. It published it is experimental.

F

The version number is like zero one, zero Dash experimental.

F

um So it's pretty obvious to anybody who downloads it that it's experimental um but feel free to to build on that if you'd like to try it out, um and uh it should help us validate some of the things we believe we need in in the in the official web SDK. So yeah um there's no tests in the repo. Yet for it, um unlike all of our published stuff, uh but I do have issues for integration tests and unit tests.

F

um I didn't think it was worth putting too too much time into testing with something that was so experimental, um but but sooner than later, as as we stabilize, obviously those will come um so and it's kind of related to that. uh The playground I have a big branch in the playground that I I is all pushed up. uh That I did not link here, but I will um in the notes that basically implements some like demo quality vendor providers uh so I.

F

Basically, all on I have three of the five that currently existed in the playground. uh Demo, there and again they're demo quality they're, not released or published, or anything it's just showing how you might build, how vendors might build providers with the web. Sdk I'll link that Branch uh um in this meeting, so you guys can take a look at that also might be um educational, uh uh so those are the two main things I wanted to talk about in terms of the web SDK and then I also have two two links.

F

um We recently did merge changes to the governance Charter on on the technical committee.

F

um So now that you know, we've been kind of operating in a startup mode as I like to say for the for the first bit of the project, but now that we're kind of our bootstrap teams are reaching the end of their terms, I figured it was important to kind of have this outlined.

F

um Actually, let me pause there. Did anybody have any questions about the web SDK or any of the stuff? I just talked about cool.

B

D

B

Emphasize though, like I guess, the call to action there is uh is tested out so try to build. Like you know, if you're representing a vendor or a tool, you know try to build, you know, uh try to add support using that experimental SDK like now is a great time to get feedback, make spec changes whatever once that's basically approved, and then it becomes much more challenging so yeah. Please take a.

F

Look yeah, yeah and religious that the other thing they'll do like I said: I have this Branch with with the demo providers I was talking about um like I, said three of the five that are in the playground, so right now, I think I've done, launch Darkly, I, think lunch directly split and one more I can't remember which but I'm gonna do all five that are in the playground and I'll open a PR and kind of highlight those, and so, if you're, a vendor who's, if you're one of those vendors, you can go.

F

Look at your implementation.

F

um The implementation I've done and see if it kind of makes sense or you know what I might have done wrong, um but but anyway, yeah I hope that elucidates some feedback so I'll do that after this meeting yeah and then back to the technical committee stuff. So we did recently merged uh changes, defining the responsibilities of the technical committee and there is an open, PR, already Linked In the notes um for the election process. There's still a few questions there, so we may want to have that ironed out.

F

I'd definitely like to have it done by the end of this month, because the term ends for the first TC next month. I think it's April 18th, so we definitely want to have it done before then.

H

um Any questions.

F

H

Yeah um I think it's a it's. It's April 28th, so.

F

10 days, that's good! Thank you.

H

Yeah, no, that's all right. Sorry, it's an easy to remember date. For me: um yeah I think only one one question that was still open there, which I saw you put in about an hour ago uh in the draft I put in 25, but yeah you're right. There are only three people, so it would be impossible, uh so uh yeah I think we can change that to to 33.

H

I. Think one thing that's open for discussion um is: if, with this coming election, we want to just add people and to what number we want to get or if we want to stick with three people in the TC.

H

um Let's say hotel doesn't uh from what I've seen doesn't really have, for example, an expiration on the term uh when it comes to to the to the TC.

H

So we could follow in those lines honestly um and then simply when someone is no longer able to let's say fulfill the duties or you know decides they want to step down, then uh we can open up the conversation to lowering the number or having like an ad hoc election to then replace the person that leaves.

I

um So usually, TC 30C is not limited by the the amount of people I mean for the governance board. It makes sense because it shouldn't a grow too big and you want to have an uneven number to to get into positions while still allowing enough interest reports to be represented. There did you, the TC is usually seen as supporting the GC on technical questions, and it's also related to project size and project momentum.

I

So while it might be overdoing it, but you could even argue that, like I guess, we should be a TC member for every major language uh that we are supporting in there. If this is what we wanted, nothing, it needs to be there like, for example, it might be maintained as a small language, languages and I.

D

I

We should just get have a good mix in there to not forget people who actually Implement providers, so it's I think more defining like what do. We want the mixture to look like when.

D

You can have.

I

All the SDK people in there, but it's for example, seen Thomas here as a pro somebody writing a provider with different interests or different opinion on certain topics uh that you would have as somebody writing the SDK, um so I think that's something we should.

I

um We should guarantee that, but I wouldn't limit it to the to the amount of people. You also want to have continuity in there, so it usually don't want to vote out people it's just very often for people to step like in their open source career that they're moving up like to becoming a TC member and like with every everything. That's like a career or an advancement step that the process should be clear there and uh I. Think that's the that's. The key part I haven't read it yet.

I

I'll have a look at it: yeah yeah, I.

C

Also think, in some cases.

I

It's just an appointment like somebody needs to be proposed by at least two other TC members and maintainers, and then it's just accepted by the GC that they join um and there's no like official voting. uh Almost so, but I didn't have a look at it. Yet uh thought.

F

Okay, well, I think. A lot of things you said are are really actually great points, um so maybe uh David or even you, if you want to look at the pr good, could add them, because what you just said, you had some concise input. That I think makes a lot of sense. So um yeah I personally agree with what you're saying I, don't think we need to limit it, um I think practically I I think one of the current members does want to step down.

F

So we probably do need some more people, um but we uh yeah I, don't see a reason to to limit it and I think. Maybe one of the things we can even add to the responsibilities of the TC chair is in the meetings that- and this is proposed in the recent changes that were adopted to the Charter, but that the chair would um maybe, in addition to those the share, probably should evaluate involvement of current TC members and you know, follow up with them or remove them from the TC.

F

If they're they're not participating anymore but I I guess we don't need to limit it.

H

Yeah like right now, it's basically it's it's kind of done by other TC members. So it's not a larger election or something to complex.

H

uh One thing: I just saw I thought I had actually removed was uh some other projects have that when, um when someone leaves, if there then is a majority of a company that automatically one of those people has to be removed, I would rather put in we hold an election to replace the person that left or we try to, because otherwise, you know, let's say they're they're three people right now one leaves and then the proportions don't fit anymore it it's a bit weird like.

H

Oh, you have to leave and then it doesn't work anymore, because then the other person has to leave as well in our case, but.

I

Yeah, but that I think that's a very theoretic case. We have represented Representatives like from different orgs. It would mean that there was already a majority like more than one from one org there before, which can happen early stage, and it would be like fully institution I would rather say.

I

If you remove somebody from the TC, the circuit should be replaced by somebody with a similar skill set, which you would naturally do e like a either being a maintainer of the same projects and also like having like the same role like being somebody providing working on a provider. Somebody from the maintain a cycle, um vendors and.

F

So on, but I think going back to your point that the TC is more of a like a supporting body, not like a voting body. We probably don't need to have an election to immediately resolve that we can just say: okay, we've lost the person. Now we need to seek to find somebody else.

F

um You know at a reasonable time frame, yeah.

I

Usually the decision like in many others, like speaking all the footage you see here, can be done by the GC members, together with the TC.

C

If it's already proposed.

I

In a majority by the TC, then it's just for the GC to pull and say, okay, and why wouldn't they? But it could ever obviously also be the other way around that the GC says: hey just coming up with something like Ruby is getting super successful nobody's representing Ruby in the TC.

I

We think it would be a good idea to have a ruby representative in the TC and then whoever is acting as the chair I, like the idea of having a TC chair like not in a role that he's the boss of the TC, but he's like more or less the one taking care of the organizational points of view and like acting as a single point of contact for the GC, because who do you otherwise usually reach out to, and this might even be a role that you're switching like from like quarter to quarter uh or on a quarterly basis from one to the other.

I

It was like acting TC chair um just at it, because it's mostly organizational burden um that or organizational responsibilities and no real decision. Okay,.

H

All right so I think I don't want to hog any more of the of the time today, but yeah I'll follow up and also add. Some of these maybe propose some of these changes in the pr as well.

B

All right, perfect, yeah.

F

B

I think I think it actually ties in pretty nicely to what Simon's going to be talking about in a second, so yeah good stuff, um so yeah I guess uh over to Simon. Unless there's any questions and just a reminder, I did link the meeting notes again too. If you're, if you're just joining, you can feel free to uh click on that follow along so over to you, Simon, okay,.

A

Thank you, hi everybody, so I will shortly talk about the centralized approach, I'm just suggesting for I'm sharing my screen. I hope everybody can see it uh about managing the community and managing the users on GitHub, because we see from time to time it's getting really hard to manage those manage the uh alignment to teams Etc and keep that in overview and therefore I opened a proposal for the Community Management, with perivalos perivalos.

A

For those who don't know, peribolos, there's a tool uh developed maintained by kubernetes and cuminators uses it for themselves to manage their whole organization. That's uh assignments Etc. So this is a tool which is actively used to handle such kind of loads and do this kind of work actually and uh the proposal is that we we do not follow 100 approach, which kubernetes is using because they are configuring. The permissions on a repository level and especially in the open feature world you most of the time, have two repositories like for the sdks.

A

You have the job you have one SDK and one contrib repository more likely and for that I suggested diff a little bit of a different approach uh for that I. Also have opened uh already a short POC, uh where we, where you will see some of the structure, how it is working and how it will be applied. It will be run with GitHub actions if I'm going too far into detail. Just stop me as a and the configuration is like following that.

A

You have a config folder and the first folder is more likely the organization you are managing, so you can put in there even a Sandbox organization or other organizations. If there are some to come, you have one organization file which describes the organization itself, who are admins, who are members of those organizations? You have the possibility to Define teams in here, so overall teams which are not in those working group teams. Let's say that way which do not assign to which are not assigned to repository I come to that later and the cool part is.

A

You can also Define your repository with a certain aspect, so you can easily uh provision and maintain those repository and then it comes to.

A

Then we have those teams or work groups like, for example, that.net SDK, where we Define, which repositories are within this work group and who are maintainers, who are approvers, who are admins and, for example, the same thing comes also for the organization we put that also in a work group where we have the author, the community, that our GitHub and the spec within one and with that we can easily add new members with a good history behind that and we can easily manage all the repositories and can assign them and generate groups.

A

Etc yeah I would highly like. So if you have some kind of feedback for that, I would highly recommend. Please write something to the to the issue: uh the more feedback we have, the better it is, and we will soon more likely take a look close at what I've implemented in the POC World. Maybe we have some some misconceptions or misread the markdown files or misunderstood something and figure that out and bring that into a shape. If that's okay for everyone.

F

This seems really interesting to me. I guess one specific question I have is: does this make code owner's files in individual repos obsolete or does it work with them or what do you? How does that work.

A

No, it's it's more like it's. Only it's only generating the GitHub permissions, it does not can generate the code owners, files, Etc, so code owners still exist and you should still maintain them. But but currently, for example, there is a maintainer group. I took sort of a look at it.

A

There's a maintainer group and sometimes to maintain a group only has some repositories right permissions, although it's the Global maintainers Group and it's currently quite inconsistent and with that we get a little bit of consistency and also a little bit of reproducibility and transparency into how permissions are assigned and who has access to what I.

I

Think the question would be: could it manage the code owners because just file today and one of the pro of some projects, a PR on exactly this- that code owners were not in line also the structure changed.

B

I

B

Worst case, we could use the code owners to reference that group that was created through here so I think there's ways to kind of to handle it, but I yeah and in this, in my opinion, is a great first step. At least we can make it very obvious he has access to what uh make it.

B

You know, even in line with what we just talked about too it's like if, if the TC changes- and we just open a pull request- and you know it's very transparent about who has access to what and and how that kind of all flows so um yeah and if I think we get later in that complexity too, with, like you know, maybe centralized code owner files or something too at a later point.

I

Yeah I mean I, think it's great uh having it centralized. I didn't understand it just from the pr. So this is super helpful uh I think what would be great that people can modify it as a short Doc sec, how and how it should be modified. If you want to change something, I mean it should be straightforward, but honestly, when it just looked at the pr and not knowing variables, I was a bit overwhelmed with everything that you were committing.

I

Now that you explain it in two or three minutes: I have a much better idea and understanding about it. So I think that that everybody comes always into you and say: hey Simon I want to change something here or am I going to.

C

I

A

Yeah I can definitely update the readme and add a better explanation to that and.

I

I think they're. Quite there, they comment on candidates also manage code owners. Files just shows that there is the need and the pain point in the project to have this done more centrally.

A

But I think what what Mike said with using those teams in the code owner files rather than persons, will actually allow it to better handle that, because you can use the teams and Define them in their owner rather than individuals. And then you can use that in combination.

I

Which is the more maintainable way? Anyways? Yes,.

F

B

Cool uh yeah well done thanks for.

C

B

Quick overview and if anyone has any questions or feedback, you know either leave it here, reach out to Simon, put it on slack so um certainly uh interested in the feedback here, but I think it's a it's a nice initiative and it looks like you have the next topic too Simon. So.

A

I I started to working a little bit on the documentation and to increase it. Maybe the contributors from flag D saw my pull requests Etc. The first step we did is: actually we tried to make the content more visible from different kind of sections of the open feature community and one part, was adding their Community repository because the community files are quite important how we are structured, and now uh this is also. This is a different uh topic- browsing I'm.

A

Sorry, uh we added them now the community files directly out of the repository in the documentation. This allows people to wait, easy and navigate our community files gives a better overview and we still have it in one repository to maintain- and this way we we have a better overview of those people can easily read into that, and it's even easier to maintain when you see the structure in a more like web page way, rather than on uh GitHub markdown files in Reading just markdog files.

A

We will also continue with that slowly and try to integrate flag d as a proposal for now there will be soon a deploy preview, but we also will have the flag D documentation, then, within the open feature documentation just as some kind of sneak peek how we could handle other documentations within the documentation set for easier usage and visibility and overview for the users.

A

Any questions regarding that.

B

All right, cool thanks, Simon all right, um I have a couple quick ones, uh so, first up um Google summer of code, so we have one project proposal out there right now. um David and I have signed up as like co-mentors on that and the so far there's been a lot of interest, uh almost a shocking amount of interest, to be honest uh and like uh almost mini competition between students and stuff that we're trying to manage right now um so I did want to throw it out there.

B

We do technically have, until the end of the week to add additional proposals out there. So if anyone wants to be a mentor, there's definitely mentees I, think that are interested in uh in you know, working on these projects, so uh just throwing it out. There there's certainly no obligation, but it could be a pretty good opportunity to work with some pretty motivated students um and I just provide a link in the docs here.

C

F

I had a very brief thought on this: I don't know if we we agree, um but one thing we could have uh uh help with I suppose is implementing some of the sdks that aren't finished. So there's the rust and the um the rust and the uh Ruby Ruby SDK I think Ruby might be close to done, but you know even even rust. It's obviously a language. A lot of people are interested in we'd.

F

If we wanted to offer those up as summer code projects, we just need to make sure we had somebody comfortable with those language languages that.

B

Is the challenge yeah 100 so like and that's I guess a good point too. If there's something like that, that people feel comfortable being a mentor, um it's just is basically a matter of creating the issue somewhere and then creating that proposal. So it's it's pretty straightforward. The cncf was approved as like a an organization for Google summer of code, so like we will have projects here.

B

It's just a matter of like do the students or the the mentees want to work on the project, because I think they have to apply for it um and so there's a bit of a process there so um yeah. So just just throwing it out. There there's certainly no obligation, but if anyone's interested in you know kind of helping out um I think it's it's a pretty good opportunity for both the project and for for obviously those students as well.

B

um And the next one I'm going to defer, because I see that uh Justin's already added a section on this, so uh I will uh let him take that one over uh when it's when we get to it so for now I'll hand it over to Skye.

D

Hi so I've been working on a proof of concept for extending open feature operator to support the deployment of flag D um in client-side context, so there's externally accessible, um I've added a link to the proof of concept and ofep in the dock and in the flag, D Channel. If anyone's interested, we're hosting a office house to discuss this further and kind of go over some of the decision points tomorrow.

D

um There is a link, in the dark hand, channel to register your available times.

B

Well, thanks again, um David.

H

Yes, all right, since you brought it up at the beginning, um just some updates for kubecon, so we officially do have an open feature. Kiosk. uh We actually have the full um the full schedule. Let's put it like that. So if anyone is interested in participating, I'll be posting a I, don't know if it will be an issue or something like that to kind of gather who would be available to to kind of get the kiosk uh throughout the conference. We also received a project meeting slot, so we have one hour.

H

This will be on April, 18th, I'll, add in the time I'm, not sure if it's 10, 30 or 9 30.

H

um yeah and then, as as we mentioned before, there's a slack channel uh specifically for kubecon. So if you want to join in there, if you're planning on participating, yeah, just reach out and I'll I'll be putting there, mostly the the uh the information for the kiosk and the project meeting.

H

uh That's pretty much it.

B

All right cool thanks, David uh Justin, yeah.

J

Hi, um okay, so I filed an ofep for evaluating uh and an inline evaluation provider which is evaluating the rules on the server that is executing the client code, not on a remote server.

J

um I had a question about how so the general sense was. That should not be an open feature uh and so I marked the thing as rejected, which is fine um and then I made it I put it in the contribute. Repo I have a PR for it, which I think still fits the intent of the rejection, which is this is not for the main spec, uh but it can live in contrib.

J

They didn't know. If there was. We should say something in that ofap like rejected, but you can find it in contrib or something else. I, don't know uh so interested in feedback on how to handle that peace.

B

Yeah I guess I can go first.

B

um I think that is kind of the right approach like this is more like a provider implementation in my opinion, which would be kind of like you know, a vendor or a you know, open source tool or whatever and I think it's perfectly acceptable to live in in the contribs repo as such, basically, and so that would be similar to you know any of the other tools out there as well, and it's just you know yet another way to kind of plug it in to open feature and and use it.

B

You know, if that's if that serves like you know or fulfills your use case essentially I think it works just fine so and linking to it, I think makes sense too. So if you want to open a a quick pull request, I think I merged your existing one already. So if you want to just add like an amendment there, that just says like you know, the work has been done here or something that would be nice. Yeah awesome, yeah.

F

I think of an old folk gets merged with a with a not approved State. Putting like a like reasoning. Why and, and subsequent links makes a lot of sense in this particular thing. Also I just want to um highlight, like I, think it's a totally valid use case and something that openfeature can even talk about. I think even does talk about in some of our blog posts.

F

I just don't think we're in a position to spec it right now in any way um in, in my opinion, uh but I still think it's very valuable thing to to see and to have and to even talk about.

J

Yeah, the intent was not for this to make it to the spec. It was to talk about it as a community about what do we think about? This idea? Are other people Beyond me just are interested in it? Yes or no.

B

J

C

Short answer is like.

B

This is a common pattern, actually so I think there's certainly interest I just don't know if there'd be interested in working on like a common, like kind of like in in process. uh You know evaluation, engine of sorts or something I think that's probably unlikely, um but you know I guess uh we can. We can leave it open. I think you did start the discussion. If anyone you know does have interest like you know like we can re-re-explore it. um You know if that turns out to be the case.

J

J

The next one is we've: we've talked a little bit about scoped providers, so, instead of having a giant single provider in, uh we have multiple that are per client.

J

um I wanted to just make space to have that conversation to make sure that if anyone had anything that they wanted to talk about with regards to that, so we can.

B

Yeah we may want to do the same thing that we're doing with some of those those flag. G proposals like maybe we carve out just like a dedicated like office hour thing for this, because I that I really think this is a good idea and in trying to just get people on the same page and push it through, maybe like the quickest way to make this happen.

B

um It's up to you. If you want to do it that way or not, but uh you know I think this makes testing easier. It solves some of the other problems that, like other people have mentioned, um you know so I I, really like the idea but I think there's you know we just need to get people on the same page, just work through the the remaining details and then you know spec it out.

B

In my opinion, I don't know if other people are like strongly against it, but I personally think it's a good idea and there certainly could be value there.

B

Maybe maybe give a quick pitch too, because I'm not sure if everyone else has an understanding and.

C

Then we can go.

J

Right, okay, so the let me make this a reasonable size for everyone else.

J

um Right can I just view this perfect, so the general idea here is today we have an open feature set provider and we just pass it in a provider in in the New World.

J

We have a notion of a default Global provider type and then you can additionally scope this to a specific client's name, and so you can see here. uh We we have a client's name of my Suite library and then the application author can say: hey when my sweet Library tries to evaluate stuff I actually just want to use this stuff on the on disk.

J

uh Don't don't go contact flight key for that.

D

That's that's kind of the.

F

Whole pitch yeah I mean I, think I think it makes a lot of sense to me and it's pretty appealing and um if you uh I put the issue in this PR that you're showing right now in the notes, but I think that um we've already seen some people kind of rub up against limitations around this.

F

uh um So so I think it's a kind of a good thing to to think about um I'd love to get is you know, especially if you've done, um if you contribute to the spec or an SDK I'd love to get more people's feedback on this, uh but it seems interesting to me and like it could solve some real real people's problems. Specifically, there's there's one um from a contributor uh in the Java SDK who works at I, can't remember exactly but who's running into this issue in their rather large Java code base.

F

So I think it offers us some valuable flexibility, yeah.

B

And the testing aspect would be pretty handy too, because we had to do some some pretty wild things to to get it working in some of our tasks with the kind of the global Singleton model that we have so um okay, uh yeah I think let's continue to push this one forward, Justin and see if we can get this thing.

B

uh You know if we want to do like a quick sync like in you know a call or something we could do that if we want to just keep pushing it through slack um that works too, but I think conceptually, it makes sense. It's just really kind of figuring out what you know, how it would work you know and how it would feel to to users without introducing like breaking changes, would be the other uh consideration there and then it looks like you have. One more item should.

J

Be pretty easy to answer people at my work are Wonder, so I I finally got to deliver the kubecon stickers uh that I picked up for open feature to the our main office um and I just got a slack message this morning, they're like hey, everyone wants more stickers. What's going on with that, how do I get more stickers um and so I think I. Think eBay would be very fine to pay for set stickers, but if you have contacts on how to get stickers that are open feature branded, that would be great.

B

H

um The cncf definitely like so at coupon there will be the stickers because the cncf supplies them for all for all the projects getting specifically stickers made I. Think that's more. Your question.

J

um My question is: my company is not going to keep counting you to my knowledge, uh but people in my organization would love stickers. How do they make that happen?.

H

I I don't have a good answer for that. I.

B

Mean I can grab a few from kubecon. We also have I have a few that we've made previously. That I could send you that that would sort of help uh solve that problem. We also have the uh oh what the figma files for these, if you want to basically just take them, export it and then like go to whatever Sticker Mule or something to to order some shiny ones or whatever. So that's all totally available. So whatever you'd prefer um I'm happy to ship you, some too I have a a couple.

B

You know 100 at this point.

I

Yeah I think that the the printing so I think we will find this sponsor for printing. That's not the big issue. It's rather a distribution topic like sending them all around them and bring them to kubecon or to sort the events would be fine and usually finding somebody takes Logistics of Distributing them widely. Is the challenge yeah.

B

Yeah but I can send a feeder Justin I I can trust on the distributor, eBay I guess this time, but yeah for one-off things. It does become a little tricky and I. Think, like the the short answer, is where either need to figure out like some kind of distribution, thing that doesn't cost a fortune and it's relatively easy to do or have people show up to like different events and we'll just have people that that bring them and can distribute.

I

Yeah or identify a couple of core people who are going to speak and we didn't send it to them, but they can better distribute it. Maybe let's pick it up, but just from what we did another project, it's like redistributing stuff becomes a full-time job for somebody.

H

In your case, Justin, if someone's also at the open source, Summit I, can smuggle them there from kubecon it's two weeks later, so it's not a long way.

B

Yeah it just messaged me: I can I'll, send you some for now, um but then I think for a more long-term solution. We probably can figure out uh other ways to distribute it. But if you send me your address or something I'll get some in the mail uh later. This.

I

Week, thank you, yeah I think we just make this aerobic game right. If anybody wants to stick, there's a yes here, you get 10, so you have to distribute.

B

I

C

I

Wanted to receive a sticker, so yes, now you're receiving 10 and that's how we actually have everybody have a sticker, yeah I think the production is not not the issue sending it to a couple of people is dealer.

I

um So if you go into an event and if you want to read this period, I think we can find a way to to make this happen. um I just share it and saying hey I would need some and then you more or less act. Just let me send it to somebody and they can act like with some further distribution.

F

I'm also willing to make the sacrifice to travel to the West Coast as long as somebody else pays for it and delivers stickers.

I

That's exactly what we were talking about: cost of distribution.

C

B

All right, cool, yeah, um yeah I'll, try to get that that taken care of, but yeah. It would be awesome if we can distribute more stickers. So maybe the pyramid scheme was the way to go too. So, just think about that.

B

um I did add one more item and if anyone else has any last items feel free to to quickly add it, um but I just wanted to remind people that uh there's some killer, Dakota um tutorials that Adam has been working on. um Some of them are based on Pete's recent blog post, which I think was really well done.

B

I mean they're, actually pretty fun to go through, like it's kind of a cool way to go from like kind of hard-coded feature flag values, all the way to more, like Dynamic uh feature flags and it kind of tells a cool story and and why you know why open feature is valuable. Things like that, so um definitely take a look later on with it provide feedback um and yeah I think those are pretty well done. So I had I had a good time going through it. So I.

I

Encourage you to try it like everybody. If you have some social channels, please also I mean, as we want to increase the reach, just also share them. If you can that's also helpful for other people, I mean it's just we can craft the message. Obviously everyone, if you can write their own message but to make them easily shareable with a Twitter LinkedIn, whatever channel that will tell more people what's out there about LinkedIn I.

D

Know that everybody is.

I

Usually willing to do it, but if we don't bring it up explicitly, people can just forget to do it.

B

Foreign I did add the link to the meeting notes too. If anyone wants to just directly uh check that out.

F

um Since we have a little bit more time, uh I'll try and find the issue. Oh actually, if it's one opened by Tom, um and particularly anybody here, who's representing a vendor, um uh yeah I'm, sick here, I'm trying.

B

F

Issue so there's um maybe actually it's a PR?

F

Yes, okay, yeah! So it's it's this one here. So um uh basically uh there was a. There was a question about making flag-d plugable for vendors um and I just put the o5 in there. This is something that has been kind of a hypothetical um for for flag D for a long time, but the whole idea is because the interface is inside flag. D, look very intentionally similar to The Go provider interfaces.

F

You could, with a little bit of massaging um plug, basically a go provider for a vendor, like you know, launch directly flexmath um whatever into flag T if you wanted to hypothetically. So this is something that we've heard come up a number of times.

F

um Tom has opened this uh this PR talking about it, um I'd love to know, if there's any vendors that are particularly interested in helping POC this or or if there's any interest here, uh and you can feel free to follow up on that on that PR or and I think there's an Associated issue, but yeah it's something that is hypothetically possible. I. Just I'd like to see if anybody's interested in doing it, for helping with.

K

Yeah, we might be interested in doing this. um Yeah.

K

uh Okay, yeah we're a super small team, but one of um one of us could go on he's. He can write, go um so yeah. We could definitely have a look into. It is.

C

K

Just the kind of standing up and shutting down that's different and the rest of it's the same is that broadly.

F

Well, I think, there's a lot of questions. um You know it's it. You know because they're, basically, you know, would we want to do some caching I think the answer to that is yes, um how how might the crds have to look in in uh in kubernetes to support this?

F

If it was going to be, you know, kubernetes supported, um which I think would kind of be taken for granted if it's supported in flag d, uh the other questions like how might we build the artifacts um because we wouldn't want to I, don't think like include flagsmith, for example, by default in in the flag, deep binary everywhere, so we'd want to somehow make it pluggable um or build multiple binaries for any vendor who wanted to support this mechanism.

F

um So there's a whole bunch of open questions and I just kind of like to get the creative juices flowing um so.

B

Yeah yeah, it probably is in depth as we should go today, but yeah definitely take a look at that. You know we can work.

F

B

The issue and if there's any particular question or we want to come up with a prototype or whatever it's just really kind of gauging interest to see if it's worth the effort or if there's just zero interest or Reason to to do it. um There are some advantages. There's some potential disadvantages um to this approach, so yeah, it's I, guess the jury's still out. If this is a good idea or not yeah.

K

I mean yeah I haven't seen this. My only thought was um like there's a potential danger of it becoming a kind of a it if, if the popularity of the flag, the project continues- and um you know, there's there's a potential that it starts to like confuse what you know like. Are you building for one interface? You know what what are you building for? What are you targeting kind of thing? That would be my only you know my only concern really.

K

You run the risk of like bifurcating the projects or the community in the effort between, like basically the on a non-kubernetes folk and the kubernetes folk, potentially right where why you know you just go well, you know where kubernetes spoke. We want to follow that pattern all the way through and we're going to build for that, and we don't really care about the rest of it.

I

I mean it would still work for both I think uh I, always like the idea, because it would be easier to switch over from one Flag Management System to another right now. I would have to more or less rebuild all the application Services where it would only have to change the um that that provider in the flag intermediary I'm deliberately not calling it flag.

K

Tea right now, but running.

I

The process and her tests a provider built in there, so it's actually very similar to what uh and exporter would do in uh open, telemetry honestly said open Telemetry support will do pretty much the same.

C

I

Which I think would make it very easy to switch things around at least theoretically?

I

Why am I saying? Theoretically, because you would still actually need to exchange this component and or the open Telemetry that there is the opposite, build like the core open television collector, and then you have to collect the Builder and you can decide which receivers you want to have in there and then you would eventually see vendors like even providing their own distributions uh for this intermediary component, um so I I think it could really help on an upgrade path.

I

If I want to move to a commercial solution um but I I agree, it would be very explicit that you don't need it for using like the the same way. I think you don't need it today for open to them, which we obviously to run an open, Telemetry collector. But still the end to your point has become like the default practice.

I

What I would still say if it becomes a default practice for everybody, there's usually a reason for it becoming the default practice as well. um I think I would would go for the experiment. I think would be great if you could do that and uh yeah I. Think that, then, will you talk about positioning if it really works, because if it's like really super smooth, we could even have like one contains all of like the uh then packaged with all the gold providers in there, and you could really switch from any provided to the editor.

I

I know thought we'll come back to me to another, how much memory overhead it's gonna go big to the binary it's going to be, um but I think it would showcase what we want to do even easier in a lot of situations, especially the first third party software that you might want to configure so, but I think we anyways need to be more clear enough, like these related to this one.

E

And we also get a list written up in this ofep of like the expected pros and cons. So as we're going through it kind of. We can compare against that and see how it ends up.

F

I think that would be a really uh valuable thing. I'm going to note that down yeah.

B

F

Like I I, first of all, I agree with everything. That's been said so far. It's highly experimental I think we have a lot of open questions here um so but but yeah I think before we even really do any pocs. It would just be good to talk around this kind of stuff, yeah and pros and cons are probably a great thing to start with.

F

um So that's a great Point Dan.

K

Were you thinking of it Todd as like a shim or the can then sort of sit between the um the flag, D process and the um the the open feature like provider, or were you thinking of it in some other way?.

F

um Well, I mean I, think, there's a lot of ways to do it, so I almost don't want to like get get into it too much, but I mean the the off the top of my head. The way we could do this would just be kind of like Eloise is saying, build a build, a flag, D or several flagged images where basically what's happening right now in flag, D is offloaded to a provider, so flag, D itself would would just run and it would be.

F

It would be a separate binary, a very similar binary to what we're shipping now, but with some additional components in it. That would then be contacting for command and control contacting a vendor back end of one one or another. Essentially um is what I had in mind, but there's more there's certainly other ways to do it for sure um you could use. We have grpc inflicty, so you could kind of use that, as the integration Point there's tons of ways to do this, but of.

K

um Not having to like you know you could you could build a Docker image and then you know have have that been run with different providers without having to change anything within that image. At all. You know, even configuration which I think is, is definitely of Interest yeah.

I

Yeah, it would just maybe change like a configuration file. I'll say: okay, now, using this same way, you configured in the hotel collector. It was just by just a different exporter. I.

C

I

That all of this I think actually works. Pretty will work, probably pretty straightforward. Maybe it's been some difficulties. The biggest challenge I think we should I see, is how would we then handle the situation? You start with, like the you have, a crd was a configuration and then you move into a Flag Management provider like how you want to handle this thing or you don't care.

E

I also have some questions around if we end up with like flight change listeners, um some impact there, so yeah definitely still have some questions around what it looks like.

B

Yeah, it's probably a good time to kind of uh just transition and say like take a look at that that ofap we'll we'll work through some of those, like uh maybe FAQ type things pros and con lists and kind of work. Asynchronously on that, and we can do a you know, a deep dive in a bit. You know the last thing I'll say on. That is like one interesting point for this like.

B

If we can have this like centralized point when we start looking at, you know, Integrations and like the open, Telemetry demo application, for example. This could be a really nice way potentially to just switch between. You know, back-end. You know vendors to manage these flags and stuff too so similar to actually how the open Telemetry demo app works for observability tools where you can swap between um you know the tools to give the a similar demo. It could be.

B

That could be an interesting integration point for us as well here so um yeah I think uh I think we're okay on that one. Is there any other items we want to discuss.

I

uh Not to discuss, but we didn't put it on the agenda, but just to bring it up here. We did not forget about the incubation proposal for open feature.

I

It just did not yet get picked up uh by this. You see like we have. We would be good enough to like get started on the due diligence document, so not that this is forgotten. It just did not move forward so far,.

I

B

B

All right, perfect, I guess we'll wrap it up here. Then, thanks. Everyone have a nice day. Everyone all right, bye,.