Cloud Native Buildpacks Working Group, 8 Sep 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Working Group: 2020-09-08

Description

* Stub Team Maintainer: https://github.com/buildpacks/rfcs/pull/106
* Distribution Team: https://github.com/buildpacks/rfcs/pull/113

A

Stephen actually on vacation, do you.

B

B

uh He did mention that yesterday, I think right. I think he did say that he was.

A

I just noticed on slack, it says uh it has this tree vacation tree.

B

Yeah because he mentioned yesterday that he wanted to that's why we discussed in depth that other thing yesterday, because he mentioned the uh he was going to be out for like a week or something or two weeks, maybe even.

A

And then everyone else is out for um yeah the.

B

A

B

Spring one.

A

Yeah joe's too cool to go those plebeian conferences now.

B

Yeah he's too busy writing his. You know: go code, he's moved on to better languages.

A

If he has no comment.

B

Yeah as I post for help in the girling chat like help me now, yeah.

A

Man, I'm so so concerned now with john leaving.

B

He's my he's my crutch man for the uh for like oh this go thing just did something really weird john's like oh it's just this thing over here. That's, like you, know eight environment, variables and it'll code. The way you wanted to I'm like oh cool, I guess and now he's not gonna do this.

B

I have to actually post to people and just.

A

C

A

B

All right, I guess we'll give it like what another minute and just call it.

A

Could I just talk more about stack max.

B

We could yeah. I was hoping to get get into that code, the last couple days, but now the bustle. I have a plan forward to unblock myself on the stuff that I'm working on, but then also some other people got blocked by us, not having a staging.

C

um We should have this conversation, not in the recorded working group.

B

Yeah, that's why I stopped saying whatever the words were but yeah, but the uh I'm unbelievable.

A

You've got to decide more work, there's more work for.

B

Something else that I have to do now, um so I haven't actually got to look at it, but there was some questions related to the the the platform group this morning, they're talking about stackpack and conoco, that's relevant they're talking about windows and uh one of them. One of the people said that they uh looked at it and thought that it looked it looked like it wouldn't.

B

Take too many changes for conoco itself to work for windows, so they were, they seem pleased with it and they would like to help test it when we get to the point where we're producing images, because I think no no, this is uh this is uh pivotal.

A

B

Windows, the windows yeah windows, team yeah, so they want to. uh They want to start investigating that once we produce images because I think we're at the point we're not producing images yet in that branch, rajo um we're running the packs, but we're not exporting yeah. Okay, that's where I thought we were, and so I think that's the uh that's the next big step isn't it is to start writing images.

C

Yeah, that's right. um I talked to steven about this yesterday and I think we'll be able to do the build the build stack, changes and merge that before we do the extend phase, which is where you actually produce an image during build. You basically just like make the changes in line cache them, but you don't ever actually create the image, and so the windows part would be the just the extend. I guess I don't well, maybe I don't maybe there's more to it than that. It seems like the build build image. Part would just work.

C

I'm sure there's some edge cases, though right, but any case yes, you're right. The extend is the next big piece and I'm hoping we can merge the builder piece before that. Just to you know, I'm mostly just worried about the go mod like I'm tired of having conflicts in it.

B

Yeah, so we have to extend the run image right because the build image you don't have to do anything because you run that just on top of and then you just continue running and then we're going to run, extend and we're going to basically run those same stack, build packs um or well. I say same it could be a subset of them. I guess right, I don't know, did we do the whole build run.

C

B

At this point,.

C

Or is it all yeah so yeah? I punted on that because I think we're gonna have to get here. I think that's fine, um yeah yeah, so right.

B

Now is the same: color pretty sure we'll need it. So does that mean we're going to have to? uh How are we going to get those those packages over to the run image because we're going to boot up this run image right? Your current base run image so heroku 18. and are we going to like volume mount in just the stack yeah.

C

B

Something is that kind of the idea.

C

Yeah the um it will be expected that the platform injects the extender binary and the stack packs.

B

Okay, yeah, so we're gonna have to something so life cycle, or I guess let's say we said platform right. So I guess this is not a life cycle.

C

Right right, yeah, so it's like life cycle can just assume that that stuff has been done so.

B

Anything that's.

C

Gonna build packs.

B

It will run basically is that kind of uh okay.

C

Yeah, actually, I think, it'll be slash, stack, slash, build packs, okay, like cmb slash, stack, slash, build like we're going to put them in a different directory. Basically,.

B

So it's interesting, so it's the right, so lifecycle is not going to the builder phase is not going to put anything into those directories. Is that right? So it's going to be done by the platform.

C

Yes, that's weird, uh I think.

B

It doesn't have to be the same ones in some cases like the same ones that just ran.

C

I'm pretty sure that that's right um but yeah, we need to talk to steven about that kind of stuff and then the run builder, because you know after after it's done. We need to do something with the layers that have well. No, we might even just chuck them, I think, is what he was saying yeah. I.

A

C

Was saying that, like I don't know, good chat.

C

What's up david.

D

I I thought that uh both terence and jesse left- I guess it's a really small thursday meeting.

A

Joe has that effect on people.

C

You're not uh partaking in the spring one festivities.

D

Not so much we actually just made some uh 9 p.m at night. For us um I figured I'd call in for the working group meeting, but otherwise, um I'm out of.

D

A

Now that you're, are you back back.

A

Yeah, uh can we bug you about your rc.

D

Sure that's uh I called in I I responded for today, but if you have comments you know.

A

uh But you're you're, okay, with sub team dealing with that.

D

I'm I'm okay with it. I definitely think it would be it's it's easier to pass it all off to the author, but I appreciate that uh we want to make sure that it's as appropriate as possible.

D

um I did try to put in like a bit of a bit of form like uh just because I sometimes have like difficulty when I look at something that rfc is tagged with. uh You know with um what's all again um that they're in uh during fcp, I'm sometimes not sure whether they're fcp, like I don't know whether they're, actually in fcp or whether it's just like a bit delayed et cetera. So I just said that we should add the label. If it you know, mfcp is going to be approved. It just needs issues created.

D

um Add it to the formula around that, but I'm happy taking that out. People think that's like overkill.

A

uh What was your question about? Sv like if it's actually going yeah? I think the fcp means that it's been approved pending, like any other new context, um so that should generally like, I think, for most rc's, and I think that this has been true for the most part. Right I mean joe could call me out. I guess I'm wrong, but like.

D

I I remember some with the emerged, closed disposition, but.

A

Sorry, what was the question? Oh just that, like once a decision is made for because of through once, if rc is hit, fcp like it generally doesn't change like it's kind of leaving the opportunity for new context to be brought up.

A

I think generally, that will also hopefully, as the project arose, uh be probably even like community-led like oh, like maybe community people will focus more, hopefully on things that are fcp or are more time sensitive, and it's like. uh I could see, for instance, like say joe's stack pack stuff goes and yes you're the mixon one, and it's like. Oh hey.

A

I actually use build packs and, like I have this youth case that maybe you didn't think of, and maybe it moves back to drawing board, or maybe it doesn't move the needle and we just go through it right.

C

Yeah, I think the idea is like 80 of people who are somewhat involved in the project. Don't care about the bickering and back and forth before fcp right, and they don't want to get involved. They don't have strong opinions. They don't want to get involved in the little discussions we've had on, like stack packs, so they might not pay attention to.

C

They might not pay attention to something until it hits fcp, in which case they know, it's like stable enough that they can take it seriously and don't need to like think about it as much as steven does or something- um and you know just like taran said, then it could go back to the drawing board based on something that they uncover there. But you know if they were comfortable enough to get to that point. It's most likely like a minor thing that they're concerned about.

C

I think, that's, I think. That's the intention. Yeah.

A

At least the door open, I guess for input, or I guess, like the other, the other way it's used sometimes is when uh like, for instance, on vacation right now and like maybe, we will move something because votes go in um and when he comes back, it's like it's like giving him enough window to basically and to like chime in for feedback.

A

um It's like hey, I know you're everyone's busy like uh this is this has gotten enough momentum and approval from everyone else like to move it forward right, like it's like vote now or forever hold your piece or like comment down forever. All your piece kind of thing, for, I guess the kind of core and sub team people uh like maintainers and contributors as well. It's like um so I.

C

A

Contributors and maintainers can also have that impact, even with non-binding boats. It's like hey, like this is actually a real problem uh for this particular team, because uh we haven't thought about this thing right.

D

That's fair, thank you for the clarification.

A

A

Yeah, I think it makes sense to just do it for stuff that has fcp. We like create the issue, and uh uh I mean I I guess it could be.

A

The other alternative you could do is like just when stuff gets merged in, like you, we you just keep track of, what's actually been merged um and then make issues off of that afterwards.

A

But maybe that's hard to track. I don't know.

D

That's the that's the current practice. It's just. I think that it's easier to ensure that something is going to actually be created. If we make sure that that's like a gatekeeper.

A

Yeah um and I'm happy if it's just like the label, the issue is not created, can block like we'll block fcp merging on that. Just ensure that that happens. If that's something, we really want to do.

D

Are you you're happy if that is well to be fair? If, if it's, if it's an issue, then someone like the some of the people who would otherwise merge it in can. I can also create the issues um or connect, at least if they.

A

Really want to merge this in, like please create the issue.

D

And anyways it hopefully should be done pretty fast. I just wanted to formalize because I feel, like things sometimes either slip through the cracks or just weren't, clearly communicated into like. Oh, this actually did get implemented.

A

Yeah yeah, I think the I think the process matters because of just tragic the commons. It's like none of these things are necessary hard to do. But if everyone isn't on the same page, then it's like oh, I thought it was someone else's job exactly so that won't make sense.

A

Cool uh sounds good to me. uh I I also think it if an author is really incentivized to create an issue. It shouldn't stop them from doing it, um but I just want to make sure that there's a formal process for it to make sure it's actually happening.

D

Do you think I should add that into the directory, uh if you.

A

Want uh I it's not gonna block my vote on it, but.

D

Okay, maybe I'll just clarify that small point, but otherwise great.

C

I'm gonna drop y'all. I got some things. I need to take care of.

A

All right later, joe.

A

um The newest rc that, I guess is a smaller group but.

A

uh That maybe we're talking about is the distribution team rfc that you opened uh add a new sub team.

B

A

uh I pushed him to open harvesty about it today because he just opened a pr, I think, on the community repo that changed the sub team text to add another team, um but it, I think it's a big enough project impact that it probably deserves an rfc.

A

But it's a pr 113! I.

A

A

um But basically trying to find a place for um the circle, ci orb, as well as the bill pack registered work um because they don't seem to fit from previous discussions. The the uh kind of platform team, which I think seems to be the place where, if nothing has a home, it goes there.

B

Right that makes sense yeah it makes sense, distribution, team.

D

I mean the the pack orb: could the pack arb could live in the platform team.

A

D

Yeah, it does today much the same way as like the tecton could also probably live in distribution, maybe in the sense that it's a tool that we're kind of like allow or we're you know, organizing for users to, I don't know, distribute climate build back to another platform.

B

Yeah yeah, I think that makes sense. I think.

D

But then it's on examples.

A

B

Yeah, you know, because they're really examples right in the pack orb similarly, but also like. I would think that the distribution team would also be in charge of maybe like the brew pack stuff, but I don't know like I'm not sure.

A

B

A

Yeah I mean I I guess at what point is like what does. If that's true then like what does platform own pack is that it.

D

Yeah, that's that's like the question I know like when uh when I talked to javier I'm like about maintainer track, so he was like one of the big things is he's like yeah, like maintainers work on the distribution of pack and making sure that, like it covers every platform, so this kind of like necessitates adjusting all right. I don't know I have to adjust my mental model on what.

D

It could be that that tecton is more of a platform because we're like running the life cycle. I haven't even looked at the pack orb, so I don't know.

A

No, I mean, I think, it's a fair question. uh I always thought circle ci felt like it should belong in platform. I get why build pack, registry and stuff should be separate because that it's like this is one thing: that's not like the others. It's like not a platform at all.

A

uh It's not a platform where you're running the life cycle right, yeah yeah, I feel like that's kind of the separations like I feel like anything under platform, is like a place where someone is running a life cycle in some shape or form.

A

And like how to package and distribute like basic life cycle in a, I guess, more user-friendly way.

A

um So I guess, would you not want circle ci? Would you want certain ci pack or to remain inside a platform if you were to lean on.

D

I feel like in terms of content. Perhaps it should be. um You know I don't know, I haven't touched it in the I don't know seven months or so I've been to the project, so I don't know to what degree it actually gets actively maintained and it could be putting it in the in distribution would help it get that active uh like active touch. I don't know what more development it needs.

A

I don't know if it needs a whole lot like. I think we pull. We allow people to specify the version of packs you you don't even have to update it based off of pack releases right.

A

um But yeah I mean.

D

Yes, the default is 11. It's 11.. We should probably update that yeah the default um default yeah.

A

I mean I don't know, maybe we need to update it because we need a chance of default.

A

uh Maybe that's something we should do as part of the pack release process or, like part of that.

D

A

Do you guys ever talk about the circle crowd in your.

A

A

Do you talk about techcon examples? I feel like javier breaks up every now and then.

D

It's come up. It was like on our. It was on, like our internal team um backlog a bunch of times, so we I think that there's kind of more active usage or people at least post more about it, and apparently sometimes we point it to give an example on how to like implement the life cycle.

D

So that yeah we we like adjusted tech time but then, like yeah, I adjusted it to use the creator, like maybe.

A

Do you have to manually update, uh like the lifecycle version stuff in the tecton example,.

A

C

A

Like the next version of life cycle, does tech time pull that in automatically? I think, if you just pull the example or do you need to update it.

D

I think if it's it's set to a specific tag of the of a docker image of the light, exactly I think.

A

So we update that image. I guess.

A

A

D

I'm definitely down for having to build like registry in this and just not care about circle ci, but.

A

Yeah um I mean I just think if travis the one who's maintaining circle ci. He should just do that as part of his work on not platform or on.

D

Top is he that's his? uh He maintains that.

A

Yeah, I think, he's still not maintained, certainly or for the most part. Okay, because we we use it internally at.

A

I think we moved off of our own implementation of a pack our own, not or but distribution yeah solution. What what should have been an orb.

D

Okay, that makes sense.

A

But mostly just like anything else, that's right, like it's just a convenient way to be able to not have to download pack manually by hand with pearl inside of a circle file right like it's pretty simple. It's like you use this function or I don't know what are they called encircle, whatever use this orb to be able to get access to, some pack commands right where it will download and cache and kind of do those things for you.

A

A

uh I guess the only other thing that was that's been interesting, that um I don't know if we want to bring up again is probably the I feel like the built pack default process discussion yesterday was I'm still thinking about what we should do about that.

D

I I'm kind of in favor of having it live in the life cycle. I feel like at some point we're it's not so much of an like of an opinion that we're forcing on people we're just the default that they're allowed to change if they want.

A

To so you're happy with it, as is.

D

Yeah, I think it well, it recently got changed. I think it's part of moving your passport before it got changed. Yeah yeah! Well, we only realized it was going to cause issues when they changed it, and then we tried updating. You know the lifecycle version of pack or, like oh shoot things just don't seem to work the same um so now I think pack is providing what is providing web for the default, but.

A

So wait, life cycle did a release, but if, if you use- and I guess an older version of the api- it more or less works is that right, I'm not gonna.

D

Do that I think so, so that's why window? I think windows has to be on the older api because of.

D

A

So you liked it the way it was before.

D

It was just something that seemed to work, and I didn't have to think about it um now.

A

D

Now and now we do.

B

It's all the way back to joe's like thing, I'm just never changing this to begin with, like kind of the whole multi-color binary stuff, I think because he's like, oh docker, runs, gonna be broken and people are gonna care and people are like hey. Do they really care? Question? Is yes like now.

C

B

Care about docker images being produced with no uh in the run command, because.

D

B

I thought we decided, then it just didn't really matter if it didn't have a default. You know.

A

Yeah, it makes me think about a whole rfc, where I felt like we were saying these things at least joe, and I were saying these things and people are like, and now I feel like we're having that discussion again,.

D

I was excited about that just because I was annoyed when I tried to make a like a docker image, use it for a pack using pack and I wouldn't be able to like do it in a way that was sensible because we weren't able to set a like an entry point or like a different command.

D

I think so. I'm happy that we're opening the door to do that, but I think the default process ended up being adjusted by that also yep.

A

I mean I don't know, maybe there's an already argument made that the default process should just be executing shell of some.

A

A

I don't know um if people don't have other topics they want to talk about. I'm happy to and this early.

B

A

I think I'm going to hop off here.

B

I'll see y'all next time.

D