Open Feature Project, 23 Jun 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OpenFeature - Project Meeting, June 23, 2022

Description

Meeting notes: https://www.google.com/url?q=https://docs.google.com/document/d/1u_1qnS0PiYvZxwtsBgZrsX6DrRKfpG56q-Kh0-qn_pU/edit%23heading%3Dh.wh8mkiotns4b&sa=D&source=calendar&ust=1655043850692349&usg=AOvVaw3jgs22qY0Eh_ZvErxDUON9

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

A

B

Hey guys how's it going.

A

We'll give it a few more minutes, because we usually have some weight joiners.

A

I'll share the document chat. um If you go to today's date, the 23rd uh feel free to add yourself as a participant.

C

A

Okay, it's three after so I guess we'll go ahead and get started um thanks everybody for attending. uh I think I see some new faces. So um what we'll do really quickly? First, um is if you'd like to introduce yourself, go ahead and raise your hand, and then we can uh take time to do that. um You can raise your hand virtually.

A

I mean with the reaction um that way we can get everybody who feels like introducing themselves. uh Okay, so yeah first uh go ahead: brian hi everyone, I'm brian jenn,.

D

I'm a software engineer at harness.

A

Very cool, um I think it's a victor next.

E

Hey yeah um vic from uh from dev cycle, uh product manager and sdk engineer.

A

Very cool uh kobe: how am I getting that right.

F

uh Yeah uh had a product at that cycle.

C

Very cool uh nick, hey um nick director of engineering at devcycle, with uh kobe and vic.

A

Cool a lot of deaf cycle people uh and thomas.

G

I think yeah uh thomas uh I'm creating a library called go future flag, so I'm interested in what happened here.

A

Great um so yeah good to meet all you guys, uh since we have so many people from dev cycle. Does anybody want to quickly give uh an explanation of what dev cycle is and why you guys are interested in.

H

A

um I'm not familiar with the company.

F

Yeah sure um yeah, uh I guess quick brandon, just pretty simple uh feature management tool focused on um kind of like the full life cycle of feature management, so um they're not just like not just trying to focus on like the devops side, but also how do we handle like beta management and all that stuff kind of like through one interface um but yeah. We came across kind of like the open feature um project and just wanted to be participants so interested in obviously like learning, more and understanding kind of everything.

F

I think we signed up a couple weeks ago to be uh participants, but yeah.

A

Cool very cool, um all right. So uh before we go further, I guess: um do we have any volunteers for a scribe? We usually have one person uh nominated for every meeting to kind of just add notes to the agenda um under the sub items as we go. Anybody want to raise their hands to do that.

A

Thanks, justin appreciate it, um uh okay, so let's go ahead and get started. Actually, uh maybe someone. Unfortunately, I don't see bayert here who was, I think, gonna talk about his draft proposal. First, so maybe we'll start with my items and then we'll see if we can circle back um yeah. Let's do that. Okay, so um there was uh this one should be really quick, but uh there's a linked issue uh 84..

A

I know that um justin and fire- and I went over this a little bit uh my back and forth and I think maybe we're landing in a place where um we can potentially uh potentially close this, um but I did ping quite a few people on this. So especially if you're developing an sdk, um I know dave's been working a lot on the golang one. You might want to take a quick look at the last few parts of this issue.

A

Basically, we're looking at maybe escape escape patches might be a good way to describe them, but like some places within the open feature, spec that you might be able to get an instance of the wrapped client sdk.

A

So you know if you're wrapping some vendor sdk, and it has a feature that maybe open feature does not have an abstraction, for there may be some desire to to get the instance of the client sdk or at least get the provider um for that, uh for that uh instance, and then, and then maybe in the provider, call a custom method to get that that better instance.

A

um Obviously adding things like that kind of has complexity, and it also, uh you know, moves away from a common abstraction and kind of suggests that you could use uh things that are not common abstractions, so something that we want to have some consideration around make some consideration around in the particular use case. That's brought up in that issue, which I think is pretty interesting. um Justin actually made the observation that the use case can be accomplished with hooks, for instance, so that would be a more uh open feature.

A

Idiomatic way to do the same thing so yeah, especially if you're developing an sdk, maybe take a quick look at that and and see if the proposal makes sense, um uh and actually you know, I might add, a specific link to where I reopened this issue, just because it was an existing issue that I reopened.

A

um Okay, so that's point one, um so the other thing I think I'd like to hit on is the uh contribs repo update. um So we have a um a repo, I believe it's I believe it's publicly accessible now it might not be because it's just getting started. uh Let me just see.

A

So yeah the the node.js contribs repo I'll, also link that in here, so um for those of you guys that are a little bit out of the loop. um We have uh this concept of a contribs repo, so, along with the um the kind of core sdk there's also um there's also probably for every language, a need for having this additional repository that would contain um uh providers uh open source providers for various vendors.

A

Perhaps um I mean there's no need to have providers live here um as well as hooks, so one of the key ones that we would like to have implemented is especially in our alpha releases, is an open telemetry integration hook, so basically a way to add open, telemetry metrics in spans to any flag evaluation in a standard way um without actually directly adding open, telemetry dependencies to our sdk.

A

um So so, basically, it's you know. The contribs repo for every language is a place where people can add, essentially extensions and additional contributions on top of open features via hooks or provider implementations that kind of thing. um So I kind of really wanted to throw that out there, because uh we also talked a few within the last few meetings.

A

It's one of the things I think justin brought up and made an issue for um about what our alpha goals look like, and what our alpha like kind of languages would be our alpha platforms basically- and I kind of wanted to to throw out there that this might be something that we have as a requirement for an alpha sdk release. um Maybe not!

A

um I just wanted to see what everybody thought about that, because especially we're talking about something like open, telemetry integration, that's kind of a key selling point for open feature, and if we don't want to bake that into the sdk, then we need a place to house it. It doesn't have to be super mature. It doesn't have to be a mono repo, necessarily maybe there's other paradigms in various languages that um that are that are better for that.

A

But I wanted to put that out there. I don't know if anybody has any ideas or comments on that. One.

H

So I didn't comment about java sdk. I'm I've not considered a no-tell hook as a requirement for alpha, because I think that there's plenty of information that we can get about the usability of the interfaces we've provided without necessarily having an hotel provider.

H

That said like if we had a no-till provider, I'd use it.

H

But I also have the bias that we're not as we don't have a ubiquitous usage of hotel within ebay.

A

Yeah, I guess the other thing is hooks- are pretty basic to write like even the hotel look itself in, like the current, like implementation, we have is just a few lines of code. I mean the interesting thing about the hotel hook is that it actually connects to every life cycle stage, um but it's basically just like create a span, a data to a span close spam.

A

It's not overly complicated, so it would be very easy for anyone who was interested in that to write their own, even if there was no centralized place to do that they could have their own implementation. So.

H

And maybe it shouldn't.

A

Be a blocker for for an alpha.

A

As far as as far as like patterns for creating some kind of contribution, repo like something like where you know arbitrary packages could be published, maybe from a single location.

A

I know that's a pretty popular paradigm in node- I I don't know if anybody has ideas around how we could structure repositories for other languages, to to support that, uh if a monorail makes the most sense or if we would just have a whole bunch of different packages, kind of living desperately that feels like we could become a little bit disorganized and challenging, but maybe something to think about, as we mature.

B

Yeah, I think I think one of the ones where the relying it might be a bit tougher right, because I guess when you publish the package, is really tight to the github repo and the tag on that repo. So if we had a bunch of different providers in there every time we rev that would be increasing the version for everyone, but that might be the only outlier. I guess you must need to look at our language by language basis and see us.

B

I guess anywhere where you actually build and publish a package. It's probably not an issue, it's more google kind of relates heavily on get. I think. That's the gotcha.

A

Yeah yeah it's a one-to-one with with go and repositories right in packages, so that makes sense. um It may be a bit of an outlier, okay, um definitely stuff, that we can come back to, especially as those languages get closer to an official, uh an official release, and I I think I agree with your conclusion justin. I don't think it needs to be a requirement um and like we, we are a little bit farther ahead with the javascript contribs now.

A

So if we see that working or not working we'll be able to apply those lessons learned to other languages and platforms, um so uh now that bayern has joined, I'm wondering if we want to pivot back to your point uh for the draft uh proposal you have. um Are you ready to talk about that right?.

G

Hello, can you hear me.

G

Yeah yeah I'm ready um yeah, so I wanted to propose the the one that was already ticket about for the even tracking.

G

Personally, I think, based on the the custom defense functionality that I found in all offenders like launch, darkly and a few others.

G

um I think we can keep it really simple and just have like a track or track event or I don't know, I exactly want to call it um and then just like event name and then you can pass like something similar for the beauty flag, functionality where you can pass a context, and it will be then, like all the extra metadata data you want to send, along with the event.

G

Yeah, I was thinking, there's something like this and then uh this will be the function that can be implemented by the by provider and it could be like a no operating function when the profiler doesn't support it or throw an exception, or something like that. That was my idea.

G

um I only the only problem I have if people are happy with this idea and what thing I just proposed. I only not sure how I can like uh write this down, because I don't fully understand the exact structure. What is expected of like a specification proposal?

G

I did have this draft pr where I started writing something, but it doesn't really make much sense to me yet how I should fill in my my proposal was back so yeah. That's it in. In short,.

A

um Well, I can. I can certainly work with you if you'd like to, um if you'd like, to try and spec that out. I think it would be a smaller spec, um but perhaps some background is is necessary too, for people here, um because I know that not every vendor has such a concept. Some of the vendors I'm familiar with do, but it is a subset of even the ones I'm familiar with, so maybe um maybe to like. Oh.

G

A

G

So so at work we are using postdoc for our product analytics and future flagging functionality and they allow you to define future flex and then.

G

So you have like persons or like type people, you want to target for your future flag and- and you can send events and this events can then become a property on the person. So you don't mean you can send an event, but something happened or like user clicked on the order button, and that might you can then set like this. If you can then set something on the person like as ordered for first time and then you could target and you could use that property in your future flag.

G

So you say if you want to do an a b test where um use has already ordered something in the past and enable this or otherwise not.

G

So, that's why I would like to have this tracking functionality, because that that way can help control my future flacking environment, because the the events that get tracked will can be dependent on in the future flex that we specify in our environment and also some of the features that that I had erased earlier about being able to access the fender client instance in most or I think all of these identify user group identify and functionality.

G

Like that is all even based, so if I would have if we would have this track ability center custom event, then all those cases would be solve too.

G

um I'm not sure if, like other defenders, also have this possibility that you can, like you have to deflect the bent on specifically friends, have been doing something that's specific to um postdoc, but I think it would be nice to have this capability like you could, instead of you, could in the profile you could use, maybe google analytics they used to send these kind of analytics.

H

We have a similar concept to this at ebay for our a b testing framework be curious to know if there are open source experimentation providers that aren't feature flag specific and if there's a common tracking standard they use. I know ebay has gone through several iterations of tracking standards and I can pass this around internally to get api sniff test.

H

But it's certainly something that we will continue to use and if there was an open feature client for it that matched our api. I would see no reason why we shouldn't use the open source thing.

A

Yeah- and I know circle has.

I

A

Yeah, I know that both I mean I I I'm just reaching for memory, but I I believe both split and launch darkly have similar functionality, um so I am- and I I I know that some don't, but um I so I think, there's some philosophical.

A

You know uh differences in whether or not some feature flag providers think that this sort of thing is within their wheelhouse um yeah, but I think it could be something we want to support. I think the thing that could be tricky is that, because some vendors, don't even have this as a concept there's a little bit of an impedance mismatch there. um So I think that what was mentioned before uh I can't remember- who mentioned it?

A

Actually, maybe it was you justin, but like um there's, no reason that we couldn't have you know composite providers or um even multiple writers. So, for example, you may we may make the tracking provider distinct from the flagging providers so that you know you could have a tracking provider um which may or may not be distinct from your your flagging provider.

A

So, in some cases you know vendor that supports tracking could be both of those things, but also maybe you could plug in segment as a tracking provider or something like that, if you're using a feature flag meter that doesn't support that as a concept um and of course everything can go up right, so everything kind of by default would probably just go up. So if you just register feature flagging and then you call a track, nothing would happen.

A

um That's my high level thought, but I think that type of thing would probably need to be handled by your proposal required um if I'm, if I'm tracking correctly- and I don't know if other people have ideas around that either, but I think there's enough of a con there's enough of a there's enough concepts around it in the industry that it's probably worth us exploring.

A

I know that in some cases it's used as kind of like a um like a finish line for feature flagging in terms of uh and in terms of really you know business value right. So this, if you have some arbitrary goal, that's represented by a track event. It allows you the ability to use it.

A

You know feature flagging to do some experimentation to see how that end result is actually achieved, so so it does make sense, especially in that context I think, but that's my high level take probably worth of specking out, um maybe not for an alpha.

H

I think the one thing I would say is please please not for an alpha.

I

Yeah, I think just just speaking generally, um I think this is where things are going to start getting.

I

You know just some some sort of like kind of hard decisions, they're going to have to get made, because that there's serious impedance mismatch across all of the providers from my experience and they people just have different opinions on how to you know how stuff should work. So the kind of assumption that you know that the trivial case um is is fairly straightforward because it is just a boolean right. I guess I mean, and even for some providers, it's not even as simple as that.

I

I guess from my point of view.

I

As long as you know like it's not going to be a simple like change, the provider string package, name or whatever and where you go like that's, that's that's totally unattainable.

I

As long as people understand when they're sort of going down a path- that's you know with a certain degree of lock-in to a particular provider um but yeah. I think this there's gonna stop as we as we start. Deeping down. There's gonna be more and more of these things where things just yeah noah will return for seal whatever, but I think that's just the nature of the beast right. It's like it's an opinionated product and you know I can't even think of as an open source pure like multivariate testing framework, not big one um yeah.

G

Yeah, I I also think we already have this a similar issue that not all pg flag vendors support, uh something different. Then there are plenty of ones that don't support the object like having a json structure or anything similar. Some only support like a name of a variant and like a boolean black and that's it so no diagonal numbers or like like any like dynamic, config, really style flanks.

G

So for doesn't support that. You have like, uh um like a json structure as a 50 flag value, so in my possible providers basically gets ignored, it doesn't do anything or it tries to like pass something into a json structure, but it doesn't do much so in that regard, I don't think there's like a lot of major difference between future vendors, that don't support like this kind of eventing.

G

And and like vendors, that don't support all the three main feed flat.

G

Types like boolean number and config.

A

It may be worth doing some surveying um before we spec anything out. um I mean we did that a lot with uh with the feature flag uh apis like there's, still, there's still research doc hanging out there from the early days when we were kind of you know trying to decide.

A

Oh, our api should look um when we did a lot of survey of the you know: existing landscape, existing vendors and existing open source solutions, and I mean I would say that there was generally a lot of a lot of cohesion, um but to ben's point there may not be as much cohesion with tracking so we should but yeah. It is probably worth understanding what we're getting ourselves into before we before we address it with this spec.

A

Probably so, maybe maybe the takeaway should be to do some research. First I mean we can open up a research issue and just just look at the vendors that do support tracking and see how similar those apis are. um Okay, yeah.

G

um Do we already have a list of like future flag vendors? I guess uh like the members of open future and anything else,.

A

Yeah, that's a good place to start is: is the um you can look at the interested in the in the same page as the interested party and community there's just the vendors. The playground also has providers with a whole bunch of vendors. You can look at that too.

H

The ones that aren't ones that aren't captured that I think are probably also relevant, are places like google analytics. I think, as tracking events and optimizely we'll have tracking events like marketing stuff. We'll have a notion of what tracking events are. We should look at those.

B

Yeah, but for what it's worth as well harnessed, if we haven't done this, but we get a lot of requests, but it's something we're looking at. So um if there's any kind of discussion not happy to help her feedback kind of what we know about it.

G

Okay, that's cool.

I

Just sorry, just to add one thing as well: there's also a big question about so like flagsmith. um We do those tracking integrations on the server side um kind of automatically.

I

I think, there's a big design split about whether they're happening um by actually, where the sdk run is, is runtime execution or somewhere else.

I

I don't know how the other providers do that yet.

G

Sorry, so what was the question? You remain like the tracking of each offense, whether that's done on the server side or on the.

I

Yeah like if you've got, if you're tracking, against amplitude events against amplitude, for example like what's sending the event data to amplitude the client device or a server or something else.

A

um Okay, well, it feels like it feels like we probably have enough here to consider this. As you discussed, I think um what what I'll do is take away is, um is create I'll, create an issue in research to do to do some research um with this I won't necessarily assign it to anyone, but at least I'll create it and um hopefully catch and I'll try to capture everything we talked about here um and I'll yeah I'll ping you on it uh via.

A

So you can add anything um that you that you want to add: okay, yeah cool. Thank you. um I was going to do a demo, but I think we can leave that to the end. So, let's make sure we get to all our discussion. um Just now looks like you added something uh about sandbox status. Do you have a question yeah.

H

As a question, probably for oleg, do you know when sandbox status is going to be figured out for us.

H

I got him on a bio break.

A

Yeah um before, uh hopefully uh he's back soon, but what I would what I would what I would add is uh we have been accepted as a sandbox project um by the cncf, but all here I'll, let it like respond.

D

um Yes, so, first of all, we will accept it after we accepted there is a bunch of steps that need to be performed before we are officially considered, even as a sandbox project.

D

There is a checklist uh we talked with mike bremer, who would be working on that, but ultimately there is a lot of logistics to be done. Transferring accounts the main ownership, registering maintenance chapters on baby, etc, etc.

D

So everything will need to be done before we could say that we are officially on the cncf additional announcements, so I made a proposal to michael. He said he wants to work on graphics and I haven't received any feedback after that, so it's basically on his plate right now. So I from what I know mike is off this week because he's doing short show somewhere in austria.

D

So I won't I'm not sure when we get responses.

H

Be like a road map type thing set up, so we can get visibility into the process, to figure out how we can help if we can help what the status is.

D

uh So there is um a github issue which lists the main things to be done, uh so it's somewhere on the cncf. I can't find it. Thank you.

H

But I would be.

D

H

Would be happy with that? That sounds perfect.

D

Yeah, I'm just looking for that.

A

Yeah, when you find it just attach it to the meeting notes.

A

um Yeah and- and the reason that I think it's been a little bit under wraps like is, is because um I got I was a little bit intimidated, but about saying anything about it because, apparently how he communicated about it, it's very uh rigged rigorously defined like we have to say it's sandbox, we can't say anything other than sandbox, um so it's like until we had all of our ducks in a row with that. I didn't want to even put anything in in chat. So just keep. Keep that in mind like there is.

A

There is specific rules around how you communicate about being in the cncf, so um we're officially sandbox, but it's just sandbox. So I don't think it's. I don't think I like correctly wrong, but like we could, we could basically get a wrist left by saying we're in the cncf at this point because we're in this year, since we have only a sandbox.

C

Yes and there's cncf helicopters.

I

A

D

Exactly they repel down into your house yeah, what team needs to be aware of uh the fact that sandbox projects in the cncf are not a part of the ccf also doesn't make sandbox projects eligible for any cncf and the linux foundation services, including color, packs security or lfx insights, some services? Basically, they get provided anyway like if you want to do mentorship another effects mentorship. You can do that, but uh there are no services or paid infrastructure, provided I guess nobody wants to get a free, jira instance from the linux foundation.

D

But if you want, you need to become an incubating project for that.

H

Okay, so I wrote this in the notes we are in the cncf sandbox is proved. We are in the cncs, not a fruit.

A

Yeah yeah exactly so, I think a louise who hasn't been with us for a couple meetings now, but my understanding is that eloise was going to um immediately move on to the next stage as soon as possible. I think that was his plan, um so you know I guess more to come on that um I'll. Try I'll, try and get that. Maybe it'd be good to see if we can get him in the next community meeting I'll see. If I can.

D

Yeah for the record, it took captain now almost one year uh to get to the next stage since we applied, uh so even if you apply speed records, it's not difficult in the case of captain I mean open feature, definitely has more visibility and more industry pressure. I mean just taking people on this call from different organizations. I think that it can be justified well,.

D

But don't forget that open telemetry became a incubating project less than one year ago, so things might be a bit difficult, sometimes yeah.

H

Thank you for shepherding us through this process.

A

Thanks a lot um yeah so like I said I had a demo and- and we can hold that to the very end. So if anybody has anything else, they want to add, um go ahead, go first, raise your hand, um otherwise I'll just go ahead with the what I was planning on demoing.

A

Okay, uh so what I will be demoing is an end-to-end uh end-to-end use of um our operator and flag d uh working together to basically do cloud native like flag evaluation in kubernetes and just a cluster I'm running locally, so um uh keep in mind. Flag d right now reads from a file, so that's the only provider it has, but, as I was just talking to dave johnson earlier, um one of the cool things about flag d is that it basically is using interfaces very similar to uh that in the golang sdk.

A

So hypothetically, we could swap out uh the provider in flag d to talk to any uh vendor back end if we had golang uh goaling provider, uh implementations for that particular vendor, um so yeah right now we're using we're basically reading from a mounted uh a mounted config map that is mounted through a custom resource definition um for a flag. So uh basically you'll see the update flags in like cool native ways, so I'll go ahead and share my screen.

A

You guys can see what that looks like this stuff is pretty new, so I hope it all works without problems.

A

So I have a cluster running here. This is just basically a vanilla kind cluster. I have three nodes on it, um as well as a a master node.

A

I'm going to install some stuff, so first thing: I'm going to do is create a can. You can you guys see my screen. I hope you all can yeah. So I'm going to create a namespace or open feature. um I have to install cert manager so that we have certificates. So I'm going to do that.

A

That usually takes a while before it goes. Green, just wait for that to start up or install some self-signed inserts cool, so I should be able to install the cert now great.

I

A

Just running a build in the other tab here so now we have a kubernetes cluster with um all the basics in there now I'm going to install our operator, um so this is installing locally, but this is. uh This is pushed up too so, we'll see the open feature operator should start up here in a second.

A

Okay, cool, so what this guy is going to do is it's going to use a it's going to use a mutate again mission web hook to basically mutate incoming workloads to kind of load in flag d to them as a side car so again what's happening here? Is we have a flag evaluation engine essentially running out of process in flag d, so flag d? Is that standalone go app?

A

That really can talk to eventually any vendor back end and it's running it's going to be basically injected into whatever workloads we have that are annotated accordingly, so those workloads have like a sidecar running with them at a process that they can do flag evaluation with um and and that way you know, they're completely agnostic of how that um how that is, is running so there's absolutely no changes to the application itself. It's just communicating to that flag d process. So I'm going to go ahead and deploy our workload.

A

I'll, show you what the workload actually looks like it's very simple.

A

So I have this end to end uh file, so um this has a few things in it. There's there's exactly three things in it. I'll start from the end, um because this is just simple, kubernetes artifacts. This is a service that exposes um our application. Very basic, just exposes it over a node port. This is our application itself. It's just the open feature, demo, app that you might remember the fibonacci demo.

A

It's just a very basic node.js application. The relevant part is it: has it has these annotations?

A

So these say that open feature is basically on and it specifies a custom resource definition instance to attach to this workload and that's what's defined above so here, you'll see our uh you know the same end to end crd mentioned, and you see some flags are defined here. So there's a few flags. These are all the ones that we need for our demo application. So I'm just going to deploy this whole thing in one. Go.

A

And that should work, it looks like it created everything. So now it's going to create my workload, and so you can see here we have my demo um application, but also in the same pod. We have a different container for flag d, um so these are going to get pulled and hopefully soon start up now. The app does generally take a while to start out because it's a very slow nest, app that comes online soon. It looks like it's running cool okay, so we have a workload with black d in it.

A

I should be able to access it here.

A

Yes, I can amazing um okay, so let's just do a really basic change. So this background is red. I'm going to go into that same file. I showed you before.

A

And I'm going to change the value of the hex color to blue, save this I'm going to reapply the same configuration.

A

So you can see some things were unchanged, but one thing was changed, so the feature flag definition end to end, so the actual values of the feature flag were changed.

A

So now, if we go back to k9, uh we should see this is still running. We may this. This is the logs from flag d. So eventually, what's going to happen here is you should see this guy say that he got an update for a particular crd, so what's happening right now is kubernetes is basically trying going through a reconciliation loop and there you go. You can see what happened is.

A

Oh, I see that you changed the definition of this custom resource, so I'm going to go in and change a config map and it changes the underlying config map that flag d is listening to without restarting any processes. So no pods were restarted here. Nothing like that was changed at all. That's kind of one of the feature plugins, you don't want to do any redeployments or whatever we basically just dynamically change the flag, values and look the app is blue. So that's how that works.

A

um You really control anything this way, so uh in in a real in a vendor situation. So so this is again like our inbuilt crd, with some with a feature flight defined in a vendor. We're you know, thinking of having something like this, like um provider.

A

Vendor x, right and then you wouldn't need to define a feature flag. You might if that vendor supports kind of some kind of api integration to define flags in their system via this definition, um but at a minimum you could say: hey, my providers are under x and your crd would probably also just take a reference to a secret to authenticate with that vendor and then flag d would be getting its values from that vendor.

A

So we're close to having that working. I think we really need the golang sdk to be finished before we go ahead and um implement that but yeah. I hope this demo kind of kind of um clarified what some of the goals are with some of the cloud native stuff. So if you have any questions, go ahead.

G

uh Yeah, it looks really cool. I have a question it. I don't. I don't know that much about kubernetes, but so, if you would have this demon and you, for example, you have like angenics running, would you also use it to like redirect traffic to two different pots depending on? If a future flag is on.

A

Or not absolutely yeah, so flag d itself. Just is a process, that's injected um really into any pod. So hypothetically, you could do what you what you mentioned by by deploying an nginx pod with some custom. It's been a while since I've done this, but I know nginx supports um custom lua extensions, so you can basically write some lua on top of nginx.

A

To do anything you want, so you could hypothetically write some lula. That would say: hey um look at flag d on this local port and.

A

Any other questions.

A

Cool um that's all I had um so all of this is is committed between the two repos, the open feature um operator, repo and the flag d repo.

A

Currently, what we don't have is any kind of rule evaluation in flag d, but I actually have a pr open for that. If you're interested go, take a look yeah, that's all I had. Does anybody.

I

Else have anything.

A

They wanted to add before we uh finish off the meeting.

I

um Just really briefly, uh we merged in uh the sort of first star, but the python sdk. It's there. If you're interested in python, um we haven't started writing any sort of adapters yet but um yeah. If anyone has any contribution to that, be good.

H

Very cool I'm going to cut an alpha release for java today.

A

Cool where's, the where's, the part of your action there. It is.

A

Yeah thanks for thanks for that, both of you guys, um I'm I'm okay at java and uh definitely know nothing about the python. So it's really encouraging to see people who are comfortable with both those languages contributing, um and it's been a good learning opportunity for me to just watch the python uh sdk development as well as the java java. One.

A

Okay, thanks guys.

C

Okay, um yeah so um myself, kobe vic, who had to drop um we're all pretty excited to be part of this. uh To be honest, I I came pretty unprepared. um How can we better contribute to this kind of meeting like for the next one? Is there any place that we could get started like.

H

They would you're with a vendor. Is that right, yeah? It would be awesome if you could look through the spec or one of the sdks, like the java javascript one and validate that that hits the fantastic thing like you can make your thing work in that context.

H

um That would be really useful and if you wanted to try a language of your flavor, whichever sdk you have and try to integrate it into, like basically write a provider in java to try to see if you can make it work. Okay,.

C

Yeah awesome, we'd love to this seems really exciting.

A

Yeah, just as totally right, I think, writing a provider is the most helpful thing you could do for us. That's where we've found a lot of uh we found where we need to like make changes. You can just validate our strategies.

A

Okay, great thanks! So much yeah, no problem! The other thing I would say in in general, um is you can add yourself to the interested parties uh in the community um doc I'll. Add that to the to the google doc um but yeah, there's a there's. Sorry, there's a community repo and you can add yourself as interested parties there. Just so can we have your stuff on record if you're not there already.

A

A

Okay, thanks again.

D