 Let's see. So on the agenda, we have EnvoyCon announcements. It is here to talk about repo kitay. And I put a little note about the 1.8.0 release. Is there anything else that anyone would like to chat about? Well, yes. I would like to talk about introducing some PRs. So I work for Cisco. I work with Ed Warnacky. And I set a note out on Slack a few days ago about if there's still interest in the PR to refactor FDs and as a first stage in an Envoy VPP integration. So I would like to talk about that effort. Sure. We'll add that to the agenda. I think we're still happy to have that. All right. I'll just add that to Cisco FD. All right. I guess we can do the short ones first. Chris, did you want to say anything about EnvoyCon? No, just that the program committee met yesterday. We finalized schedule. Notifications should be going out over the next couple days. And my hope is to get the schedule online Friday or Monday. So just schedule will be up soon. So wait. I guess the other short one is we're due for our 1.8.0 release. So I made a 1.9.0 milestone and I started moving stuff out that I know is not going to land in time for 1.8.0. So I think as of right now, we should probably stop merging any particularly risky PRs. And we can just track on the maintainer side what's left in 1.8.0. I think there's like that SDS doc one and there's like a couple small ones. So it's all pretty simple. And we can target maybe doing that release either like end of this week, Friday, or maybe the first couple of days next week. I'm going to be traveling tomorrow through Friday. So I'll be pretty busy. Anyone have any comments about the 1.8.0 release? Who's who's driving it? So are you going to be driving this? Yeah, I can do it. So I will take care of that kind of stuff. What's that? You're on the deprecations, the script to you know, ping all the people I haven't done it yet. So I might have some questions, but I will attempt to follow the instructions that are there. And if I fail, I will ask you. All right. So yeah, so I will take care of that probably early next week, realistically, but but just just a general note and I'll drop a note in Slack that we should just be careful of what we're what we're merging at this point. Okay, I guess to give each day the most time, do you want to quickly talk about the Cisco FD thing? I think we're all I mean, we're all very happy to have it. It's a useful thing. I think it just needs to get it done. So is there anything in particular that you wanted to chat about? Not really. PR was pretty self explanatory. I will push it out. I will say I'm very new to Envoy and very new to all these tools. I've worked in a completely different space for my whole technical career. So like I'm going to push it out. And I everything compiles. I have a problem when it when it links to the tests run some of the tests. So at this point, it doesn't pass all the tests. I do not know why I think it's a dependency issue. So I would ask for people's eyeballs to point out something I'm doing stupidly. I starting to map out so so the PR was just about a generic IO handle and a type of the class of a new class that accommodates an actual handle and different types. And then following on that, I would like to start pushing some PRs and getting your comments on some new, a new base class that we'll derive from to implement Envoy VPP interface. And what we need is the ability to do reads, writes, connect, you know, bind and listen and accept all that in one type of class. So there's a few new base classes I will introduce. And the whole goal of these classes is to implement what's called the VCL interface. So VCL is a library that runs, will run in Envoy's base. And it implements APIs to connect to a black box on the other side, which is VPP and the host stack. So basically it's, I need methods that allow me to call in, call the VCL APIs. And I've got to figure out how to get VCL events that come in through the shared memory message cues to get them passed up into Envoy directly. So there's a few things to look at, but there's some base classes that will pave the way. Yeah, that's fine. I would suggest that you start with the FD refactor. Let's get that going. I'm going to be honest, it would be useful if you could find someone at Cisco who might be able to help you with some of the tooling issues. I'm guessing there's someone there who might be able to pair with you. People will look of course in the community, but it's going to be best effort. So if you're having trouble just with new tooling type stuff, it might be useful to try to find someone to help you locally. Sure. Once the FD stuff lands and we start going into the VPP side of things, I think our main viewpoint is going to be that as long as the brunt of the VPP code is an extension, we don't care as much from a community perspective how that actually works, but we'll obviously have to collaborate on any new extension points that end up being required. So I think once we land the FD work, that's probably time to circle back and do a new design doc on what you're proposing from an extension point. Sounds good. Okay. Great. I have actually just one quick item. I actually burnt most of this morning on this. So product gen validate is now actually really far behind in terms of version and we haven't been able to bump for a while due to issues in build side and this is actually a problem for three reasons. Now the first is what was the original reason that we can't support Windows. The second is sometimes Mac is now failing because we're exceeding the CLI limit because of the massive command lines that the old version of product gen validate is generated and we kind of make changes to product gen validate to bump because of this because we're stuck behind. And the third is the one I hate this morning. I started to trace down some of these buzz bugs and it turns out that what at least what I was looking at was a false positive. We were actually somehow skipping product the proto validation because of it came down to the way to be the actual symbols for doing the validation where we used weak symbol overriding and the strongest symbol was being ignored due to ordering. I hate weak linking. I hate it so much. It's such a pain. And the bad ones even worse because you can't control them. But so the good news is the Microsoft folks have actually fixed this in the latest version of product gen validate. But and so this should just go away when we bump to that but we still stuck behind. So I know Chris has been trying to get this to happen. I know these ends been helping him but this is actually a I just thought I'd raise this because it's actually becoming acute for multiple folks in the onboard community. So I haven't been fully tracking this. So like so there's a problem some build problem I'm sure and like is is Chris working on it or or are we worried that like no one's right now but I'm not sure what the state is. I know that they had it and they had it a few days ago and I know he's going to call you. Do you want to speak about that? Yeah I think I think on last Friday I fixed the at least for the issue from Bezoside and seems the proto-plugging itself is returning some error with the test framework that happened in the PGV and I think the ball is in Chris's side right now that's my understanding. Okay let me I'm just hold on I'm gonna make a quick note I'm just gonna send that I'm gonna send an email and I will CC a couple people just so we can who should be on that. Harvey, Lisan? Yeah yeah well and probably best better to send to my private email. Right of course I don't know what that is so I will talk to you offline. Okay all right that's all for me. Okay cool all right anyone else have any quick stuff? No cool all right EJ why don't you introduce yourself and tell us all about repo Kitay. Hi I'm Itay and I like Kitties and that's obvious so what I've been working on on my spare time and related to my full-time job proper disclaimer is if some kind of infrastructure platform where you can deploy and implement quite easily and happily free bots for GitHub currently it's only GitHub. The idea is that the same way that you have like Travis CI, Circle CI and other CI systems integrated into your GitHub workflow by means of app and just adding another file the same should be done for bots. Up until now what you had to do is to use some kind of robot or something like that within JavaScript that can do everything and implement the app yourself and deploy it somewhere, somewhere to host it and do all these things. Instead what you are able to do with the repo Kitay system is just create a single file if you want in your repo called repokitay.sky. Put whatever you want in there in Skylark which is a subset of Python which is a limited subset of Python that suits these needs quite well and there you go you have the bots ready to go. It can come with batteries included so you have a lot of functions that you can use convenience to talk to GitHub like add the comments, the sign stuff, even merge if you want that's what I guess we're going to talk about this and also you can use third-party models if someone decides to route its own bot which I'll show soon. You just need to include it in your one-liner in your repo kitay file and there you go you have it. So if there are no questions I guess I can show it. So I guess before demo I think it would be good to talk through like whether this is open source you know what the long-term things are. I think that was Harvey's concern. I was under the impression from talking to you that it was going to be open source and I don't think it's necessarily problematic that it's closed source but I think there's things to discuss around. I kind of share Harvey's concerns where if we're going to stay with a closed source system and with that said I think this is awesome like I'm a huge fan of this I think this is going to save us a ton of time so I you know I would love to see this move forward but I was going to talk to Chris about this also and I just wonder if it's going to stay closed source if there's a way of you know doing some basic agreement such that we could either you know allow Google security people to like look at the code or like allow a third party audit from CNCF to actually look at the code. Like I think that would make a bunch of people more comfortable so you know we don't have to like talk about that entire subject right now but I just wanted to raise that and just see if you have any thoughts or see if anyone else had any thoughts on that just because I feel like that security aspect from the closed source perspective particularly for a new thing it would make people more comfortable if like we could get people access to it just for code inspection. I completely agree with what you said and I sympathize with everything you said so as I written the email that I sent out yesterday I would love people to do an audit on that because it will be good for everybody. Now the reason I'm not currently putting it open source because I want to see if I can make some money of it I mean that's one of the goals of my goals here. If I can't make money or if I can find a model in open source that I can make money as a side thing I see what we can do. I would love all third party audit and I guess I signed to you obviously nobody passes this audit on the first time and any fixes that are needed will be made. I remember from my time we were working Google how they worked and it was wonderful. Yeah sure I'm up to it I'm up to any involvement involvement. I mean my the other thing that could be interesting is if I mean if you are pursuing you know an actual commercial use case and this conversation probably needs to happen with Chris and CFR offline is if there was some sort of you know essentially a support contract so we have essentially you know SLO on the service and then obviously you don't want to run an SLO if this is just a hobby but and they carry a pager all day long but you would to essentially get get to that point where everything all incentives are aligned correctly. Yeah I know I want to get to the state and I want to know what's how we make it a bit more I would say it's official I want to see what's involved and I'm not really sure I'm exploring stuff. What I appreciate about Anvoy is the thing that you know it's the first use case hopefully it's my full attention so yeah. Yeah I mean like I would I would like to see this move forward so we've we've got that email thread going I was actually going to reply to it I mean I think we can be pragmatic here it's like you know of course it's it's bots like if the SLO isn't there on day one we can just turn it off like it's not like a total crisis. With that said I think everyone has concerns like not that we don't have the same problem with using CircleCI of if we give right access which I think some of the more interesting use cases like fixing DCO and a bunch of other things would would require right access that gets a little more scary and that's where just like having the ability to have security people look at the code or like do an audit and just to put on my negotiation hat you know since we are obviously working with you and doing a service to help you like develop this stuff and get publicity and a bunch of other stuff it'd be nice to talk to CNCF about like whether we want to it doesn't have to be a complicated contract but like you know could we get like a like to like work with you on this and like have security people get source code access like those are the small things that I think would would probably make everyone comfortable. Does that sound reasonable Chris to just talk about over over email? Yeah I mean we're we're open to whatever a project kind of decides you know you have to we're fine with proprietary tools and sawn it's just you have to decide as a project and from like personal perspective the perspective enabling something that kind of has like right access is you know definitely potentially sketchy could lead to interesting situations but you have to kind of decide as a project we're happy to fund an experiment to kind of see how things go but it's really it's really up to you so. Yeah I mean this this seems promising to me and I feel like if it went well I find it hard to believe that all of the other CNCF projects wouldn't pick it up too I mean it's just like it's it's too useful so I feel like if we can make this work I think everyone wins like all of the CNCF projects wins it wins by you know helping with some development so you know I think if we can just make everyone comfortable here this could work out really well. Does that sound good to you Harvey like if we talk about that over email? Yeah that sounds great I mean you know just keep that everywhere we're coming from yeah I'd say Google we have a bunch of secure engineers who know as ISC and they don't even trust GitHub so. Yeah and we're not signing we're not signing every commit or anything like like the others things that we're definitely not doing that we could potentially be doing as a project. Yeah so we have our own issues solve just in moving code from GitHub internally and I don't think this is makes it that much worse but we should just keep yeah become a sense of the of that I think that story we need to tell ourselves internally. Yeah I mean maybe concretely we start with like a very small experiment maybe like a couple of bot or specific things that may not necessarily require right access in the beginning and you know CNCF's happily to kind of you know fund that experience and kind of see how it goes and then if things are going well we could expand scope maybe it's generally. Yeah yeah we can so there's actually I think there's four bots that I think will be used for and one's already been written that's the assigned bots that can at least assign an issue without having to be maintainer and okay I mean that's been done already I think similarly I think assigning a reviewer or requesting a review which is like a separate verb essentially in GitHub would be a cool thing to add similar to that. The next is kick CI or basically we run CI tests and okay I think it is looking at that there's two ways to do that you kind of do an empty commit like we do it today and that applies right access and then maybe possible I think just invoke the circle directly which means you have to hand out some API token for that and that doesn't I think require right access is that correct? Yeah so I actually have news here I actually succeeded contacting the circle CI and kicking without mp commit and I didn't invest a lot of time in it but there are some quiz day but it is very possible as long as they get the circle CI token for envoy if you turn it on. We can do that yeah I mean the one nice thing actually about using the circle CI API versus the versus the empty commit is that we can kick only the test that actually failed so it saves it saves CI resources and saves time which should be really nice and it also doesn't pollute the PR yeah. The other one that would be really nice is to set labels that would also be great. There's there's there's tons of them and like I feel like once people learn how to write like learn how to write this stuff just being able to go through and like you know implement stale bop but like have it be part of the you know part of the repo and like have it be in skylar like that all just sounds awesome yeah yeah one that will require right access which I would find useful to be able to as a maintainer be able to tag a reviewer as ready to merge as soon as CI test pass yeah yeah nice to have but I don't think I don't think it's the most it's not the highest priority but it's it'll be nice yeah that's actually one of the main reasons I wanted I wanted the beginning right access because I was talking to people and that's the main feature that that people want just merge it when all the tests pass and I got my LGT or whatever you define is ready yeah I mean they're like I just just sitting here and brainstorming there are so many small ones I can think of like another one is for a maintainer to be able to not only request your view but one of the things that like when I'm looking through all the PRs and sometimes it's hard to tell if people have actually worked on it like since I've since I've asked them to to do something so like even being able to set a label and then have that label be auto cleared like if there's activity on it or something like that so like if there's a a commit or a comment or something and then I can just scan it and like I would have done all my reviews and then like there's a label like needs feedback and then like the needs feedback label gets cleared like when someone actually does something like that that would save me time like that would be fantastic and like these are all things that just seem like they don't seem hard with the type of system that that you're you're building yeah it's actually what you said just now it's really easy I can implement it right now in front of you if you want to see and if the demo gods will allow we're unfortunately I think we're almost out of time but but I mean this this this conversation is an awesome it's super useful I think we can either if you want to come back in two weeks and like do a demo then we can do that or feel free to actually just make a video and like we can just send it out to the lists and people can just watch it awesome yeah and let's say in two weeks I'll show you everything if you can send me a list of the most useful read only stuff that you want to see I will be very happy to stop working on that I also think what is name and see him Luciano yeah what is full name I don't remember Christopher I think Christopher M. Luciano wants to deal with this as well too like to put some stuff in and the main the main things that prevents me I'd not prevent but actually inhibits use of other people is I don't have a sufficient docks of everything yet and that's something I'm plan to invest a very very very heavy so I think maybe Chris will be the first test of this stuff and we'll see how it goes and yeah okay awesome um so why don't why don't we do this um since we've got that more private email going on between you and the maintainers and and Chris um I was going to reply to that today anyway why don't we just talk a little bit about like what kind of agreement or trial we would actually want to do here and see if we could just get some basic agreement and I think what would make everyone comfortable here is just for me and I think for Harvey and Google is just making it so that like like authorized security people can look at the code basically and I think as long as that is the case I don't know that anyone cares about it being open source or or not really I will very welcome it and very welcome any suggestions how to fix it obviously I will do it one thing that prevents it I think currently like one issue that I have is that whenever something gets modified in in the skylock code I actually create or got event I actually create another process where I run it there and I would love to run it in another container or another pod but I don't have enough money to do so yeah sure so that's something that will help if I have more credit and stuff like that yeah and that's and that's something where I think like if we got a little trial going with cnc app we can probably actually get you some money yeah yeah I mean if you need access to machines we could definitely hook you up so that's not just gcp credit or stuff like that or even if you want on the community cluster we could point you um I could send you an email it's a about how to sign up for that you just have some straight up bare metal so it's not the cloud but oh yeah I it's very easy for it saves a lot of time to me just to work with gke but we'll talk about the things no worries all right thank you all right bye thank you bye