Kubernetes Network Gateway API Special Interest Group, 30 May 2023

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Gateway API GAMMA Meeting for 20230530

Description

Gateway API GAMMA Meeting for 20230530

A

Central government.

B

A

Sorry there was a there.

B

Was a little glitch Keith had to drop, he had something come up all.

C

B

And so I'm host and had to restart, recording anyway carry on.

A

Yeah so right which of these types of events, or you know, Edge triggers? Are we going to consider reasonable to attach experimental too right because those the compositional parts of the API right? We have rough timelines in our heads for when we think any of those things can happen now.

A

A different choice that we make in this group when service is refactored and when the refactoring of service goes live in the API, the last two being very far out yeah.

C

The last two definitely don't seem like triggers. We should be waiting for triggers that we should maybe think about a little bit yeah sure, but honestly, this level of thinking about it is sufficient for my concerns.

C

I think it really would be interesting to try to figure out if there are ways that we can do things to try to make it less likely for experimental things to end up in production before we, you know, find some of these issues, but I'm I'm not so far sure what sorts of things we could do about that.

B

Yeah hi: this is the age-old problem of every kubernetes API. Ever it.

C

Is not every API ever right, yeah, yeah, kubernetes or not.

B

You want feedback.

A

Yes, it's never into production until it's completely in production, at which point it's stable, even though it's experimental.

C

Right right, yeah I mean yeah. We could, you know, change all the version numbers and some magic field every month. Then then nobody would use it at all, but that would prevent it from going into production. Great.

D

Well, I think there are some philosophical questions that don't need to be stated here, but.

C

D

That there is very little appetite that I've heard for anything other than full steam ahead. So thank you for hearing me out. I, don't want to capitalize anymore this time. Well,.

A

I think the one thing I could ask right, because we did go through a design, a filtration process to get to this proposal in the first place. Right of the available options have new options come into the collective uh Zeitgeist that we should consider.

C

I should probably point out that, just to be 100 certain that we're talking about the same thing when I brought up attaching stuff to Services I was specifically thinking of binding an HTTP route via its parent ref to a service as opposed to policy attachment, being you know, attaching a policy using a policy attachment mechanism from Gap 713 to a service um I'm, quite certain that associating an HTTP route to a service is a thing that needs to be supported right now and I'm, also quite certain that it will be insufficient in the future.

C

um But that particular comment about. Oh you know if, if Tim or somebody comes up and hates the idea of you know of services, was referencing Shane's comment of yeah. We know that some of the zig Network people are not Terribly Happy with bounding, binding routes to services. I, don't think, there's any great new alternatives to that. Right now for the mesh case,.

B

Yeah yeah, I, I, agree, I I feel like the direction we're heading on at least for the scope of gamma as it exists today is both not ideal, but also there is no better alternative that will exist in the next two years. Probably that you know, and so we should do something with within the limitations we have. That does not mean that we can't improve upon that in.

C

The future and for the record that timeline makes me very sad but I believe it.

B

Yeah there you know, if you think that Gateway API is, is understaffed, uh which I will always say it is uh we usually get more attendees and interest and motivation at Gateway API meetings than we do at Sig, Network meetings and Sig network is much broader in scope than Gateway. So it's just there's.

C

Just not enough people way to make me feel guilty for not showing up for sick Network. Bro appreciate that working.

B

On it, no yeah, okay, so I I think we know what's next, it seems like at a high level what we should work on in the next few weeks is adding things to the 0.8 Milestone, making sure they have a mesh label associated with them uh and running with that for tracking 080 I.

B

Think we're in the ballpark of four to six weeks away from that, I would love to make it on the lower end of that, because once we as a project are ready, we still need to turn around and take that to Tim and Cal to and whoever else at Sig Network level to review it.

B

So it's going to take some time um yeah, that's uh that's what I think is our next steps and I just want to make sure that we're on track for that, um and we can continue all these discussions kind of in parallel and, like I, think Shane and flynna both said. We probably need some refinement at minimum on what we consider experimental versus implementable.

B

uh You know different stages of of this journey.

D

B

D

Again, maybe not right, as gem was trying to hit it, because that would right out of me right so.

B

B

I've I've had a a while where I keep on thinking. Oh man I want to improve. You know this this you know process or policy or whatever, but I don't want it to Target. You know these existing gaps and so I I. Don't really know how to do that, because there's never a right moment because there's always some Gap or some gaps that are in flight. But at some point we do need to make some improvements. There.

C

We could probably start talking about it and explicitly grandfather in gaps that were already open at the beginning of the discussion. Yeah I think that makes sense as a potential mechanism. Rob. Do you wanna? That is your screen right. You want to scroll up a little bit so that people can sorry other direction there we go. Do people agree with that summary of everything that we just said or shall I go ahead and edit that further.

B

Yep from me, yeah that looks good and thank you. I never know who's the one taking notes, but thank you so.

C

D

I'm not I could tell it's Flynn because he managed to key in the TM and nobody else. Would everybody.

C

Else would be able to do that. Dude! That's just option two on a Mac.

B

I had no idea yeah, okay,.

C

B

C

Strange in some fonts, but this is one area where Google Docs actually tends to do the same thing. The same thing, one.

A

B

Cool all right, well, I, think that wraps up our meeting for today, unless anyone had any last thoughts.

B

um Great well uh thanks everyone, good discussion and uh we'll talk.

C

To everyone next.

B

Week all right see.