Open Feature OpenFeature Project Meetings, 16 Jun 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: OpenFeature/Flagd K8S SIG, June 16, 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

A

At least I have some time to change the background.

B

C

A

A

And this is papa too.

B

All right, todd's booting up his computer, so we can take some notes, um we'll see what turnout looks like today.

A

And my son brought me some photos with me. I was sorting the documents in another room, so I'm a bit afraid what's going on there at the moment, but yeah otherwise should be fine.

B

Yeah sounds good. um You had.

C

B

That I need to reach out to the cncf to get like get the ball rolling on some kind of sandbox pieces. So.

A

The skill of announcements, if you, if you just want to let's say social media, then we can post it, and this is if we will repost it. I cannot organize it, but if you want official press releases through pr agencies, etc, then.

C

A

Need a cncf marketing machine and then we need to work with them.

B

But I mean like I: I would assume that they would email us some kind of confirmation that we at least like you know, were accepted right or is that because, like I haven't seen anything yet, I was kind of waiting on that to do too much.

A

No for sandbox projects, so the sandbox projects are not officially considered to be a part of the cncf. This is what we need to be careful of so cncf sandbox project. Well, basically, there is some best effort, I think, but that's it.

A

So unless you ask for something they are not going to provide or prioritize help, but if you ask sure it may happen, so that's the.

B

The plan status is a sandbox project. There's no official like even saying your sandbox now or anything like that.

A

We can officially say that we are cncf sandbox project, but we cannot save a cncf project. So this is something you'll be super, careful of because uh the cncf is careful about it, uh but yeah, otherwise, so doing announcements on our own asking the scenes you have to repost it in their social media. It's totally possible.

B

A

I just have to figure.

B

Out what the next steps are, because it's a little unclear, because I never got any like official communication from them, which is kind of weird and slightly disappointing. Honestly,.

A

So, for me, the question is what scale of announcements you want.

B

Well, I don't know I mean it would just be nice to have more confirmation than like uh meeting notes, type of deal but um yeah at any rate, we'll we'll deal with that. No.

A

uh So open uh feature is accepted.

A

They will graduate to services like death stars to list and con of uh member projects, but basically that's it, so they won't be doing any massive pr around that. But at the same time the cncf will gladly repost everything we create.

B

Yeah, I wonder about even like logos and all that kind of stuff because, like they have like the landscape piece with the sandbox section in there, so like getting all that stuff.

A

Yeah, it should happen and also we need to have open features somewhere on this landscape in general, either standard or feature flag management. So this can be achieved by a pull request. So there is I'm not sure. Maybe they will do that on their own, but you can submit a pull request as well.

B

Okay sounds good yeah and I'm not sure what the turnout's going to be. So we may not uh like for this particular meeting. We can chat about a few things um but yeah. If it's uh a little turnout, we can kind of pivot or you know, wrap up early. I suppose.

A

So I will send the link to the since your friends keep to the chat.

B

D

C

B

If it's worth mentioning, probably not worth mentioning anything.

D

B

Mention it real, quick, yeah, um so yeah, based on that tweet from a couple days ago um that go feature flag project. uh I think you had actually uh linked it. um I had reached out to them we're going to talk to them later today, actually just to.

C

B

Which portions we could potentially leverage and how we can maybe work together on that? It seems like a pretty interesting opportunity, even looking at it, it's very similar to where we see that flag d service being potentially in a few months uh they're. You know they're just a little further along. They have a few things that we may want to uh kind of omit.

B

If we, you know port pieces over, like the aws sdk, for example, but like fundamentally like it works pretty similar to where we see 5d, potentially being so yeah we'll see how the meeting goes. That's actually right after this one, so we'll we'll try to post some kind of update once we have information on time, but.

C

B

Of cool that was a nice kind of like organic interaction. I suppose, and we found out about them and yeah. It looks like a cool tool.

A

B

uh I suppose we can just provide a quick update on the uh the kubernetes and the flag, deep keys so far, uh and then there's not.

C

B

Don't think anything really on the agenda except there's any other topics we chat about that quickly and if not, we can go ahead and wrap it up um from a status standpoint. uh We have the basics, you know kind of working into ends just to prove out the concept.

B

um So you can, you can actually use kubernetes operator. uh You can push uh basically flag configurations. um It will, you know, attach the sidecars. We are running like that simple flag d service behind the scenes uh you know so that's attached to the same pod. We have a very, very simple http based flag d provider that communicates to the sidecar process uh and does extremely basic lag evaluation.

B

So basically the end-to-end low works, there's just a lot of uh pieces. We need to. You know, continue to figure out in order to make it ready for for prime time, but in terms of like you know, proving out that it actually works conceptually. I think we're at that stage now. Right now.

C

B

A dirt road from end to end, so everything works, it's very basic and we need to make that a highway that, but basically all the concepts have been proven out, um so we're just going to be adding different transport protocols, adding an evaluation engine, refining the flag definition or the flag schema.

B

It's not it's not exactly where we want to be right now, but basically uh yeah, all the big. All the big questions I think have been answered um and we understand from an architectural perspective, how we're going to build things out. So it's just a matter of doing that which is good, because that's also where we're gonna get help from um some uh consultants too right yeah.

B

I suppose um we we do have the liquid reply team starting july 1st, um so that will be a pretty big boost in engineering power starting at the beginning of next month, so that should help out quite a bit they'll be working on this portion and I'm sure we'll leverage them for some of the sdks as well. So we can really start accelerating. You know r d effort there.

A

So one question: do we want to discuss post hoc today, or should we schedule it for the next meeting today we will have much more people attending.

B

Yeah, I would recommend, if we're so so, this meeting is really specifically oriented around flag d and kubernetes components. I think the post hog provider and that recent work is really interesting and I think it deserves a bigger audience and it's probably more general, it's less about the cloud native components and more about just general sdk. So next week is probably the best venue for it and there'll be a lot more people as well.

B

Generally, there is, I mean I can't guarantee that, but we've been having attendance around like the 12 or 15 people, right, yeah, pretty good turnout and so yeah. I think that yeah we certainly talk about it. I mean well, I think, we'll wrap this meeting up relatively soon, actually likely, because just you know uh we're missing some of the people in the audience that we probably really need to chat to about pressure decisions and things. But uh so we may have a few minutes to talk about that.

B

um I I have been working on the contribs repo for the node.js uh sdks right now, and so we can start moving portions to there. That makes sense. I have an incredibly basic hotel um created so far, although the one, um oh, I think is, is your name wired? Is that how I pronounce your name, can you can you help us pronounce your name there.

D

My name is bayek fired.

B

D

B

Oh no, I was just gonna say I looked at your your hotel hook and especially with the metrics and stuff. I like that approach pretty well. um So if you don't mind, I may start leveraging part of that in, like the more like, I guess, the the official hotel hook in the contribs, repo um or yeah.

D

Let's find that.

B

D

No boys no worries. I I copied from your your your code also, so I just announced.

B

D

C

Have you as a co-editor.

B

Or something once I do that, but I like the metric portion of it I'll have to test it out. Maybe that will be well they're all marked right now as peer dependencies, so like perhaps like they could be optional, so you could either do metrics traces or both it's up to kind of figure out how that could work from like uh kind of.

D

C

B

But um I do think, that's a pretty nice touch. um I haven't really. I don't have a.

C

B

Of experience with the hotel metrics, but um you know the ones that you were collecting made a ton of sense to me. uh So I think that that'd be kind of a nice enhancement to that uh that hook as well.

D

Yeah yeah cool yeah go ahead. No, you go first, no way.

B

I was, I was just going to change the subject, so um um if you wanted to respond to that, go ahead,.

D

Oh yeah, I wanted to change subject to. I had a question about something, but you go first.

B

Okay, yeah, I was gonna, say um I really. We really appreciate the work you've been doing, especially with the post hoc provider and just in general, using the node sdk.

D

C

We were saying earlier.

B

This meeting is really specifically about the cloud native components. I would, if you wanna, if you wanna, specifically talk about post hog at all, I would recommend you do it next week, because that's our general meeting and we have like I said, a lot more attendance, because it's about the project at a wider level and more about the sdk. So if you are interested in talking about it um next year, the first place yeah.

D

Next, I I had a question about the the cloud native stuff, so so I think I had asked this question before, but um I think it would also be nice if there would be like a demon or like something where it allows you to manage the 50 flex. So you define them in this in a format, and then this demon like in the kubernetes config map or something and then it um it would.

D

It would like sync them with the with the with the 50 flag fender, so you it will either add the fit flick or remove it or change the default values. I think that would be quite cool if that somehow um would be possible.

C

D

That's something we've copied.

B

C

That's something we talked.

B

About- and I think it would be nice if we could even use the same crd or maybe different crds, but we could probably use the same operator. I think it makes sense to use the same operator, maybe different cvs for defining them, not sure, but that's totally possible um all right. Most users have apis to create this. This stuff, um I mean right now we're just focused on evaluation out of process, but I totally agree. Like we've, we've talked with vendors that they're interested in what you're describing too. um So it's definitely something we want.

B

C

You want to like fl.

B

Describe some of this, this is an issue um I'm sure it will generate interesting discussion. It's been talked about.

C

B

It wasn't really recorded in any sense, but if you feel like opening an issue- and uh I think, probably probably flag d, either 5d or the operator no, you know.

C

What, if you want to discuss it? Well, if.

B

You want to discuss it. I would actually open an issue in research, because research is a good place to talk about issues that could affect multiple repos and then, if we create specific issues and to do's, we can do that.

B

They belong. There's a.

D

B

B

um I guess uh this is slightly changing topic too, but um I was uh using the latest version of the node sdk and the. If I use the errors um in esm, it actually uh was was throwing an exception. So I created an issue for that last night. I was just talking to todd. We may have to use like roll up or something similar to just bundle it. uh It's just a quick aside just to let you know that it's not totally working properly, at least for esm modules.

B

uh So it's probably just it's just a packaging issue. I'm sure I'm sure it is too, but um it's just yeah, just an fyi that that change may come out relatively soon, because it's it's not quite worth it. Oh okay,.

C

That's fine. I.

D

I I I don't use ethan yet so.

B

D

It's kind of a yeah, it's.

B

Such a weird time in the ecosystem, right now underneath.

D

I I first need to get over it. Everything needs to now have the dot js in all your files and in typescript. So once I'm over that hurdle, I probably will switch, but it takes time.

B

uh Into a single js file, then we avoid that problem too, because everything just is packaged into one single, like you know, transformed javascript string um and we can keep the typing. So that might be one way to do it. I mean there's disadvantages to that too, because the I, the files become a little bit harder to edit. If you want to monkey patch something in you know in your local environment.

B

So I I don't know, but we could consider doing that and then we wouldn't have to include the js extensions everywhere to solve that import problem. I'll look.

C

D

It's not a big deal, I mean, if that's how it's supposed to be now. That is just just um personally. I need to get a get ready for it, but yeah I mean, if that's the way microsoft wants to do it and it's fine, you don't need to change the whole coding style of the project because of me it seems.

B

To be an interview, I'd scripted, node, that's the problem.

D

B

um Yeah and so the other couple pieces uh shift topic away from the kubernetes side for for now, um you had mentioned a couple pieces like uh accessing the provider itself, um so I would definitely recommend opening a proposal on that uh conceptually it makes sense. The question came around like it's confusing to export, that from a client uh so like I probably would recommend not putting that in the proposal to pull it from the client again, because then it's it brings up.

B

The question like: are you supposed to use like underlying provider, or you know, use the abstractions and and it's it's intended to use the abstractions?

B

You could potentially pull it from like the the global uh open future object, though you know, since we're doing the we're setting the provider there, we could certainly probably expose a getter as well: um oh yeah yeah, so that.

C

That might be the.

B

D

I was thinking to just make a proposal for the event tracking and identifying users so.

B

It's okay, so that'd be like a basically like a general like admit uh or event, emitter.

C

B

Or an event like what.

C

B

For like user events, or would this be for like um sdk events like startup and shutdown.

D

Yeah, like yeah, like you, see, events like some vendors, like post hoc in uh lounge, yeah support, sending custom events so yeah exactly.

B

So the one thing to.

D

B

For that, I think that's that's awesome. If you want to work on that proposal, that would be that'd be amazing. um The one thing that we had thought about, which I think would be pretty cool too, is for the vendors that do not support it um or maybe even the ones that do support it. Maybe you could register multiple, like I don't know, we'd have to come up with terms for it, but basically like event providers or something um which could actually end up being like segments, and some of these other. You know.

C

B

Like, for example, to do more concrete, like I think, launch darkly, maybe it's maybe it's launch directory. Maybe it's somebody else split, there's a track um method which just sends arbitrary user data but, for example, cloudbees doesn't have that it's all feature flagging and no user tracking. So and if you're using cloud b's, you would register a cloud business provider but then also register some separate tracking provider that would just handle the tracking, maybe segment or whatever, and it would be optional.

B

So I think that's the kind of ed face we have to probably yeah basically nothing's uh registered, but and then the other idea is.

C

B

Potentially registered multiple, uh you know, event systems, theoretically, so just just things to maybe consider if you're gonna work on a proposal. That would be greatly appreciated, though, if you want to throw something together,.

D

Yeah yeah, I will be super super.

C

Interesting and.

B

It sounds like you have some experience from the postdoc world on this, but I would love for those events to correlate to the uh the appropriate um like feature flag, where possible, all right so like I'm, not exactly sure how that would work conceptually, but perhaps theoretically, if you were to do like a feature flag details like in that response, there could be, like you know, an event tracker in there. That has like the context kind of captured. If that makes sense, uh does that make sense to you.

D

um I think so I think I think I understand what you mean. I just don't know how to.

B

D

So like it could have like basically an object.

B

In there that has the event tracking method on there and that would have like the feature flag. You know um flag key predefined and stuff like that, so you wouldn't have to keep passing that in the track event, um just something to think about as well. It's not that one part's, not mission critical. We can kind of layer those parts on later, but I think you know.

C

You want to be segmented.

B

Or something you know what I mean like this: this needs to be like tracking in the context of feature flagging, so yeah.

D

Yeah, I think I will start with the the most easiest or simple simplest proposal. First and then we can integrate on it. I guess yep.

B

That sounds great cool. um What else do we have? I? I don't think I mean. I think we should try and try our best to keep this specifically about cloud native stuff again just for people who have been watching. So I don't think.

D

Yeah yeah yeah, I didn't mean to let you switch topics, is that.

C

B

D

B

Didn't get this question generalizing something officially on the agenda and we're just kind of making. You know progress on the stuff that we talked about at the last meeting, so um I don't know if there's really much more of an update, suppose we can probably just go ahead and wrap it up. Then uh yeah. I think so, there's nothing else. So, okay,.

C

Okay, thanks for coming yeah.

B

Thanks for attending and uh yeah, we'll talk to you over slack, probably at the next meeting.

D

Yeah, okay, yeah. Thank you same and have a nice day.

B

All right talk to you later bye.

D

Everyone see you bye.

D