AsyncAPI Related to AsyncAPI Spec 3.0 release, 26 Oct 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Spec 3.0 meeting (October 26, 2022)

Description

Agenda: https://github.com/asyncapi/community/issues/491

A

All right Welcome to the wait. What's.

B

The name of the.

A

Of this meeting.

B

We're gonna make a Lucas. Is that gonna be a thing we're getting whatever wherever we are.

C

That's how it ends up when you're honest with people and tell your stories they Mock You, publicly.

B

Luckily, I only have one meeting that I'm organizing. So it's pretty easy for me to remember what reading I have been.

A

uh I mean no. No a few minutes ago you were asking about the title of this meeting. Okay,.

B

Okay, that's fair. I can never remember what the title of this meeting is at least.

A

In public now, all right.

B

Throw me under the person or it's all good, always.

A

B

To do it, I appreciate it all right, welcome everyone for another spec 3.0, with a packed agenda actually for for once.

C

Do we start with betting like how many items we're going to be able to handle during this one.

B

Yeah we can do that. How many do you think.

C

How many do we have actually I didn't? Look five.

B

Four: six, four: four: if, if we have q a we got, six yeah.

C

Or but all four from mati yeah.

B

C

B

For me, but otherwise yeah.

C

Yeah I I bet on until foreign.

A

November one.

B

C

Think we are gone: Welcome To, The, Future.

B

Let me share my screen.

B

This is the agenda for today, so everyone knows.

B

I think it's viewable right, yeah.

B

um Actually magic do you want to start off with your update first, because then I can figure out what my update is in the meantime,.

D

No uh okay, yeah sure no.

B

C

D

Can I share my screen? Okay, so what.

C

It will be the better shouldn't. We do. Q a first, oh.

B

D

B

C

I can push my item in the Q a okay.

A

B

I, don't want to have any questions before we get into it or want to say something.

C

I designed this meeting so I know how to where are the vectors.

B

We're just gonna create fake profiles on YouTube and ask questions that nobody really has. But that's.

C

How you build up the community lesson? One! That's.

B

Fair, that's bad all right magic! Take it away! I! Think I gave you or did I give you a commission.

D

B

Okay, I will try, go.

D

And this one here this one actually okay, so you should see that okay, so the first one, the update on the my pull request to pull a request. The first one is to moving the tax and external documentation to info object and I, see that the the front accepted as I, remember, yeah the front accepted so I need a fast two, more uh accepts, probably from cash on the dial.

D

A

Yeah, didn't you get approval from Dave I I.

D

Yeah I yeah I get it, but it's uh you know in the previous changes before the rebase, so.

E

ah Okay, okay yeah, so they.

D

Yeah here and also I get from you.

A

C

Are four approvals.

A

They're still yeah they're still valid. ah Okay they're still valid unless GitHub. Oh.

C

Don't you sit on the right hand, side in the list.

A

Of reviewers yeah, on top on the top of the of the.

D

Pull request ah yeah, you have it there or there, but it's very strange because I make the push Force yeah, with the to make the rebates.

A

But probably the parts of the code that um we made suggestions or something uh didn't change.

C

A

That's probably why.

C

No, no, no, like the thing is that we just simply don't have a branch protection on this Branch.

D

Yeah right, so you probably yeah so.

A

You probably see.

D

Something like that make.

A

This course yes, I, don't think that's the case. I mean I I've experienced like, for instance, if I do a good push force and you change things significantly, then it will discard all the all the but.

C

That's the thing I think: well, maybe no.

A

C

Better I was targeting Master, probably that's again, Branch protection. So that's the.

A

C

Like if we would have Branch protection, you wouldn't be able to do Force push like on the master Branch. You can't do first push because there's a branch protection so.

A

Now it raises the question: like: should we yeah.

C

D

Should we just never, probably if you make the push force in the pull request that points to the master, you can make the push Force, but then you reject all approvals. As you skip I mean you forget about approvals.

C

Yes, yeah yeah about front is talking about his case from the.

D

C

Release when he did Force push on the release Branch, sir.

D

C

Be able to do it, yeah yeah.

D

Okay, so okay, probably after meeting I, will merge it. Yes, so yeah yeah.

C

And now the question is like.

D

C

What about the branch production then.

A

Yeah, that's what I was asking like. Should we enable Branch protection on next major spec.

B

Good question: yes,.

C

uh Did you answer already the question, but was it was it yes as an answer or but.

D

If you don't have that's a good question yeah, but if you don't have the blank protection, someone who earned the code owner can merge it by the RTM.

A

D

A

Is is a code owner yeah.

D

A

D

B

C

It's correct problem because it's still merged to the release Branch not to the master Branch like sure it can always be not.

B

A big problem: yeah yeah yeah, it's only a problem if a bot comes in and all of a sudden start to merge.

D

100 or something.

B

Reverted then it's a problem. Yeah.

D

B

Anyone ideas, so please don't do it all.

C

Heads off just put more trust in Bots dude.

A

Game gain access to async pay, but if and.

B

I have full confidence and but I don't have full confidence in people that build supports. Thank you.

D

D

So that's one: okay, I will merge it. So it's not the problem but yeah the next one.

C

I, don't think we still so we. What do we do with the branch protection we continue discussing somewhere offline, probably an issue right: you want yeah, okay, I'm gonna create an issue in this pack Ripple because it it also concerns the minor releases anyway. Yeah.

C

Okay, now you can continue sorry Machi for.

D

Interviewing, thank you very much, okay. So the next one is with adding the missed metadata tools in the server Channel operation and schema object. So here I've wrote, which one metadata India and front make make the review and hit. One comment very important, because I added for every the object I mean this main object: server, Channel operation message the metadata like the the filter, the like title name, summary description, stacks and external documentation.

D

Yes, of course, in places where we don't have it and front row that if the server ID you know in the map of the servers already is already a machine friendly name for the server. So it's a good question and I have only one problem with that that we shouldn't treat the ID as this machine friendly name in the case when someone want to make the reference the given server in several places and then, for example, use the server in the code generation.

D

So the name will be exactly the name for the given server and that name won't change yes between the the generated application. So yeah, that's my only concern about it. So what do you think about that? Of course we have the same problem in the message. Id in the message object. Sorry, because we have the name and the message ID and.

A

We can transition an operation ID as well yeah, yeah and.

D

We can also trade the message ID, as the name is, but for the server and the channel. We don't have something like that. So maybe the name for these two objects.

A

Next name is Human, readable right.

D

A

Machine friendly, okay, uh human friendlies are not human reason.

D

A

Name is machine friendly name for the server yeah, so is it isn't it already like an ID.

D

Yeah I I wrote you this one possibly case when we need the name, so you can of course treat the ID of the name, business friendlyness for the server, but in the next version of spec, but.

A

I I'm reading this, this response that you have there, which I don't know why I I missed it um and I, don't fully get it like if you want to change what the name of the object, but for code generation, for instance, but you want to the ID to still be the same. That's what you mean.

D

Yeah exactly, and you must remember to it, remember about it. Yes, if you make the reference.

A

So I aren't we then introducing some I, don't know code generation specific stuff into this bag.

D

Do you mean by some extension or something like that? No.

A

I mean by I'm saying that by doing this I, don't we introducing name which is just for code generation purposes. uh I cannot think of any other case. Okay,.

B

D

Instance, documentation yeah right.

A

So I'm wondering if uh I mean it makes sense if it will be human friendly in this case, yes for documentation, yeah.

D

But we have the title for it: yeah.

A

Yeah we have title for that exactly, but if it's machine friendly, why not having the the the ID right? But the problem is that you said like to generate code, so I'm thinking out loud right.

B

A

So so don't don't mind me so, but that is like very code generation specific right, so I'm wondering if this will be actually an extension or something like that um or I. Don't know like or maybe not maybe can be there, but things that what I feel is that it's a pain in that in most of the cases, what I feel is that we'll go we're gonna, have servers uh and then development and then.

D

Yeah yeah exactly.

A

Production name, production and, um and then it's like I, don't know it's weird, because you have to maintain the same name, usually the same name in two places: yeah.

D

And by this year you can have the name. Yes, I mean the name.

A

Field, well, you don't have to because it's not required, but yeah.

D

Yeah but you can use it yeah for it. Yeah.

A

D

A

For example, sorry, if you copy paste the definitions, for instance like I, have development and I want to copy the server definition and paste it and then start staging and then cop paste it again and and do production, then I will have to go through all of them and change the ID and the name right so um I don't know it seems to be more work.

A

For me, um I'm not sure like the thing here is that I see where when it could be used, but the thing is that will it be used right so or is it something that we're just like?

A

Maybe someone needs it because then, if if someone requests it, then it will be easy to edit right yeah I'm just trying to keep it simple like okay, we don't add it and and then, if someone says that hey I have a problem I'm generating code, blah blah blah whatever um we can encourage them to use an extension for instance, and then we see if many people are using this extension all over, then we incorporate it to the spec um or if someone opens an issue that and many people think that it should be included.

A

I, don't know trying to approach it from a different angle, like start simple and then grow uh which will not produce any breaking change. No.

D

It won't be the breaking change.

A

Or try to anticipate something that will we will know, or we think it will happen, but I'm not quite sure that this will happen. To be honest,.

D

Yeah but, for example, the Json scheme also has the name and as I remember. The modeling also used that, yes, as the fallback for ID dollar, ID I, don't know I'm just.

B

I can't remember, but.

A

But again this this is something that modelina, for instance, is co-generation, so same case um modelina will use dollar ID for Json schema, we'll use other fields for.

D

A

And for async API it will use a different thing. So why not an extension right.

D

Yeah yeah exactly, but we have this problem with the name of the message: object. Yes at the moment, because it's the the only place except the Json schema. Of course, when you can Define the name yeah.

A

We need to get rid of uh I mean we need to unify like. If we make a decision here, we probably should apply the same thing to message and to other objects of.

D

Course yeah, so the probably the message ID can be treated as the name is.

A

Or just completely removed I will just remove the message ID, because the message I.

D

A

Will be the key? We have a list of messages right, not a list. We have a map of messages, so the key will be the ID yeah.

D

But you can change then, and this ID by another. You know key in the object. Yes, but.

B

Why would you have to change the ID of it is already in the map exactly.

A

Why would you want to.

D

Have but you can use this given message in several places by the reference and.

A

That's also a problem because then you might have conflicts like you might have two different messages with the same ID.

B

Right, why was? Why was it that we changed, um that we changed the messages to have an object or a map of messages, but ID and a message, instead of just being an array still.

A

It was my thought was that so from the operation we can specify which messages will be, will apply to this operation like.

D

A

E

A

Of messages from the map of messages of the Channel I want this operation to only apply in this specific messages, so for that I needed to have either IDs or or numbers for the from the position of the array right so.

D

A

Of the array is, unless readable,.

D

But to have the possibility to reference from the channel messages or what okay.

A

um Yeah I was also thinking that maybe we, that is something that is yet another uh proposal that I will make at some point, but um but maybe uh I know that we're applying dollar ref everywhere, but I'm not sure if it makes sense to to also use dollar ref inside the operation. Messages keyword, assuming that we add a messages keyword to to operations of.

D

A

uh Which I think it will be the case, because it makes a lot of sense, yeah.

D

A

In that, in that case, we probably can just say, like these messages will be a list of message IDs with a dollar ref or anything right, and it will have to be found following the channel.rev messages and then resolve in there. So uh not perfect, but I think it will be easier to maintain.

D

Yeah, it will be, but but I have only this problem with Rita's the message ID, because in the current version, 2 yeah.

A

D

The message ID should be unique across all operations and channels. Yes, so it means that okay, but if we read off and only use the key of the channel messages object, you can have the a case when someone use the given message by the reference in two channels and that's the problem, because you know the people can name it name the message by the identity in different way.

D

So the given message, the same message can have the different key in the tunnel messages list.

A

If that's what they want, I.

D

Don't think that's a problem so so then we probably also yeah sorry. So the message I did probably should be unique across the channels. Not the uh you know the list of the channels in inside application yeah. So we also changed that behavior.

B

A

But but my that that is like in inside this, not not all the it shouldn't, be unique on all the channels, but all the inside one channel right. So you can have the same ID pointing to a different message uh on another Channel.

D

A

Inside the channel, because it's a map, it has to be unique right. The.

D

A

Enforcing enforcing uniqueness now, uh if we don't, uh which I think it makes sense like I, might want to call it something different in a different channel for whatever reason I don't know, um but um in the case of um in the case of having a message ID, for instance, um it will mean that this message will have this ID independent independently on where you use it, and that is prone to have the problem that you might have this message.

A

Id repeated somewhere else in another message, right and um I mean unless the parser is catching this problem, I, don't think it's a huge problem, but it it makes it makes it harder to spot this. These errors right to to debug these errors. Like uh you have. It reminds me a lot to the ID property of HTML. It's exactly the same thing right like.

A

Why are why we long ago discourage people from using ID on HTML elements right, because of that, because you can have the same ID on different HTML elements and um yeah, you might have some I mean in the case of the browser. It doesn't fail, it's still valid HTML document, but if.

D

A

To use hashtag then the the ID then, uh which one is it right? So it reminds me a lot of this one. The only difference is that even version two I hope because I didn't check we're making it fail that you have two different messages with the same ID with the same message ID so.

D

Here we have the custom rule for that name.

A

Yeah so, but in version two.

D

E

A

This parser yeah.

D

In the previous password yeah.

A

We have such a okay, so if it's covered by departure, then it it's different, because the parser here will be like uh Indiana. In the analogy, the parser will be the browser right. So if the browser is failing right, if the parser is failing, then it's and it tells you like hey. This- is repeated. This message, ID is duplicated.

A

Then it's easier, of course, yeah.

D

A

D

A

The if the, if the the whole, if the the structure itself, makes it impossible to to duplicate it, I think it's much easier and, and one last thing to maintain right. Yeah.

D

But but message ID can be also used to you know to some governance in the company yeah. So of course the people can make some storage of the messages use it in the your system, your company, the message ID, you exactly add some ID. So it means that if you try to write another message but you've write the same ID, so it means that probably you make something wrong. Yes in your system, because.

A

Information is still there, it's just on the map, it's just on a different place, but the message ID is still there. It's just outside.

D

Yeah yeah yeah, but that's that's the problem.

A

You can still enforce it, you can still enforce governance rules so.

D

You can still enforce that to what that, if you make the reference of the given message, you need to make the same key in the given to another word we'll be given.

A

I, don't know what do you mean by uh because.

D

If you want to enforce the key in the message object, yes, you need to check that the given message is the same message in different place. So these two you know two places. We need to have two two keys. Yes, the same key.

A

For that we have dollar ref right if you're referencing the same yeah.

D

Yeah, you can check by.

A

The reference in memory yeah, but but we have the dollar ref right. If it's pointing to the same reference, then it's the same message: yeah. If it's not, it will be a different message with the same exactly the same definition, which is totally possible. Why not right.

E

I may have to discuss that in an issue right recently like one month ago or something like this I remember. We we were discussing.

A

E

I think so, yeah I don't remember now: I'm gonna try to find it just to leave it there. For reference. Sure, thanks.

D

Okay, so maybe you should argue or something yes, so at the moment every object has the name. So we should read off.

D

Okay, I I will read of the name from this Pro request, because yeah it has some more problems that and that process and and probably I will open the the issue to discuss the message ID. If we should write reader and so.

A

Okay, yeah, that will make sense, so we don't forget about it.

D

A

So, independently of the decision that we take here well, I mean I'm, not so I'm I'm now pushing for let's get rid of this Fields, because I'm trying to force myself into a lean approach here right and it's simplistic- maybe I, don't know approach here or um so that we start again on version three, we start like. We don't need that. We assume that we're not gonna need that right and only when we need it uh or if we know now that we need it, we will need it. Then we add it.

A

But the thing is that so not adding. It is not a problem because then adding it afterwards will be fairly simple right. So, but if we end up not needing it and it's not a problem, then the benefit is that we get a simpler spec right.

B

So it's easier to add than remove exactly.

A

Removing will create a breaking change, so if we're not sure right now, I will be cautious and just not do it. Unless we have a.

D

A

That, oh yeah, no, we need it yeah, we didn't think about it. That's true! We need it. Okay,.

D

Yeah use case for Community yeah.

A

But I will keep it simple and to me right now it looks like a duplication of information.

D

Yeah yeah, but the same way because okay, um one thing because we sorry not here- uh oh my God, my internet, um so we have something like that somewhere and descriptions. Yes and to be honest, I, don't know why we have two these two fields. If descriptions can also include the summary, so maybe you should answer it up the summer.

A

D

A

It has a purpose it's for it's mainly for documentation purposes, actually, which I mean. If we get uh back to what I said, like um case specific like it's, if it's only for documentation, then probably it will be an extension I.

B

A

Know, I, don't think so, because I think this these are pretty common, but um summary is it? Doesn't support.

D

A

A simple it's a simple sentence: it's meant to be short right, it's meant to be a short explanation of something right and and description instead is like it's as long as you want description right and it supports markdown and and all of that right, so you.

D

Can write about some details? Yes, yeah.

A

Yeah, so you can, you can explain whatever you want there right. So, for instance, summary will be the channel summary. For instance, Channel summary could be used for navigation, so, instead of having uh send or publish um slash users you could have instead of instead of that, you can, you can have um I, don't know the the summary like create an user right. Yeah.

D

A

Something like that, which is more human friendly right, but you cannot use the description there because it will be. It could be a long thing that you cannot fit in a single line right and.

D

A

Might contain markdown and.

D

A

Summary is useful for these cases um when you.

D

A

To have a single short, uh simple short.

D

Okay, I understand yeah. Thank you very much so yeah and also this pull request, adds the text in the schema object. But it's not my case, but we have also the issue for it. So I am I am protected and the last one yeah, because one person of committee, which one because.

D

Wants to add the first one, the at the tracks to the component to to you know to make the reference between the text and also adds adds to attack to the schema scene. Yeah.

D

So, to be honest, I don't know if I should include the tax in schema, object or create another pull request. Instead,.

A

Of discussion I mean that that will mean yet another field that is not Json schema yeah.

D

Because we we have the external documentation, discriminator and deprecated yeah.

A

Yeah, it will mean, like I, think we should Instead try to go through like a point where we don't have custom Fields ourselves, uh and so we get rid of whenever we can. If it's possible, we get rid of this customer. We.

D

Can even you know.

A

Or maybe we can make this proposal on decent schema that they have tags they have tax or something like that, but yeah I'm, not sure if this is going too far, Maybe.

A

D

Five years later, in.

A

Any case you can still, if you want to have tags or anyone wants to have tags, they can still have extensions right.

A

So yeah, but that said, Json scheme of thing- uh that's not an amazing, kpi, right or well I think it's sold it. We already agreed I already committed the mistake of having three custom fields: foreign.

A

I had to create a schema object because I needed a place to say we.

B

A

Three, we have Json schema, plus three yeah.

D

So so, okay, so probably I, will remove it and start the discussion here in this issue. One.

E

More question uh could blacks become, as you say, from an extension but could could be included in the extensions catalog. Will it make sense anything from.

D

The extension catalog doesn't work, that's the problem. Well,.

E

It's gonna, it doesn't work yet yeah.

A

Yeah, a mentee from Lucas is working on that right.

A

C

I, listen, I, I, confirm.

A

Okay, eventually, it will be something.

C

But like eventually, whatever is an extension should end up in the spec. If it's used, yeah I, don't think we ever talks that something will be extension forever.

A

Yeah exactly so, if, if well I mean some things, I I, don't agree, some things, I, don't fully agree. Like I said there are some things that might be good to keep it as an extension forever. So it.

B

Depends on how specific it is for one.

A

Specifically, if it's just if it's like I, don't know vendor specific or something like that, then probably it will be. There then.

C

It's not extension from extension, catalog right.

A

Well, I'm I don't agree again like, for instance, if I have an AWS.

A

um um What is it called um event? Bridge uh specific extension that might be useful for many community members, so it's it's still valuable that it is on the event on the extension catalog, but uh it the purpose is not to make it into this into the course pack ever, but it's still valuable that it's on the on the extensions catalog right, so many people can um can benefit from it right.

C

A

C

Then it's binding, so we don't even have extensions but.

A

Bindings is not for AWS, it's I mean it's not for this. It's.

C

D

A

Event: Bridge like a specific product right, you want to have some configuration for eventbridge or something that that doesn't belong to the bindings. So that's why I was saying who I.

A

Bindings is for protocols or yeah.

D

A

D

So, okay, okay, so I think and yeah. That's it so Venus. Do you want to continue with your update or do you have any questions about my polar request? This one.

B

I think it makes sense that that YouTube basically came to the conclusion of like to keep it as simple as possible. Yeah.

D

B

In terms of it's.

D

Much easily remove it later, yeah the name and the taxi will be out yeah sure.

A

But don't close the issues uh stuff or or create separate issues so.

D

A

Exactly I will create this there's a place where people can leave a comment or yeah right. So yes,.

D

I will add in the 3.1.

A

In any case, still like, let's leave this feedback on the on the on the pull request on this pull request and share it on slack and so on. So just to see if we get more feedback other than the feedback from the.

D

People I, don't think so, I, don't think so.

D

Okay, sorry so.

A

We don't because otherwise it will, it will look like we made the decision on the on the call, and we don't want that right. So we still want to give opportunity to people to to chime in and and protest.

D

B

Least, I can't say that we didn't try to reach out exactly.

A

So it shouldn't be on our side. Okay,.

D

So yeah, so you know, do you want to.

B

Yeah sure, okay, everybody loves a good reference discussion. So let's.

D

Go through it yeah.

B

D

B

It's getting late.

A

It's getting too late now.

B

No, so this is a pretty big discussion, so I'm planning to do a little section at a time so for this meeting, I'm focusing on this use case and I want feedback based on that alone, okay, um I'm gonna, send you a I hope. Youtube is picking that up. So it's probably enough. Let's see if I can write it I'm, sending you a link to the specific discussion about this specific use case. So at the moment we I decided to say that this is has a known Behavior. So this is something that we currently say.

B

This is supported or should be supported by tooling.

B

And what I mean by this is that yes, this one so in Json schema they have something called bundling with references. This means that when you load up or when you use references, it would automatically match references to this URL to the ID of the schema without you having to explicitly say that this is a local reference to schemas a user signed up. So a reference library should automatically understand that you want that. You want to reference to this specific schema.

A

So so, in this case it it doesn't, it shouldn't need to go to this URL to graph. Yes right, it will just go and grab it from schema component schemas right. Okay, yes,.

B

The prop the problem is that this is not something that's allowed or is specified in the Json reference specification, which this is part of. ah That's actually a lie. No, it's true! So when we say reference here, it's not a Json schema reference. This is an async API reference, yeah yeah and in that standard, there's nothing that says that it should automatically match this with this schema, like nothing, says it yeah.

B

So this is very but and- and this is my issue with it- that if we allow this in official tooling, it all of a sudden become a feature that you might rely on to always work, but if the standard doesn't Define that behavior other tooling, that say they support async API could have a different approach and say well. This should never automatically match this, and this is my kind of my problem with the standard behind it and that's one of the reasons or one of the drivers for, and new standards for references.

B

But that's a later discussion am.

A

I am I the only one that finds this confusing.

A

B

Like confusing.

A

I mean confusing in the sense that uh uh am I the only one that that feels that this payload dollar ref right on the message will actually go to this URL to grab them. The message.

A

I, don't know anything about dollar ref or anything which is actually the case uh and um in my whole, life I will ever imagine that this is gonna match the one below. So this is like super intricate.

E

There is one detail here: Jonas. uh This is actually how Jason's schema reference. Json schema until I'm telling Jason Kima how the references work in not in draft 7, but in I, don't remember, yeah.

B

Number 1909.

E

Yeah, this is exactly right how it work. Yes,.

B

E

Only difference the.

B

Only difference is it's placed Within here within a definition. Subject: yeah, like that's the only difference, yeah, there's nothing that says that explicitly States in async API that this is an expected Behavior. The.

E

Way they they well I, don't know if it's in the documentation but I. Remember perfectly. We had a meeting with Ben at some point uh like months ago and the way he's playing it. It's like I'm telling this just to think about it like the way it references um are treated. It's like they are just identifiers, doesn't care about protocols or whatever, yes, pure plane strings, let's call it. Yes, they are just plain strings and depending on the tooling, it will behave in one way or another, but they always talk about kind of I.

E

I, don't remember the name. It's like Registries, or something like this, something that you load all your schemas I think agv, it's how it works right that will AGB. So you load all your schemas associated with an ID. So.

D

E

Like a a set or a map whatever, so all the work of getting this from the https, my schema registry, whatever um it's done on tool inside, doesn't have to do with the with the with the reference or with a spec. You.

B

Know what I, what I said so.

E

Completely different, so they only say this is a string identifier. Whatever you do it's out of the spec, uh so.

D

A

Know what is there now that you said that they're here, like it I got one question then. So, if I remove this dollar ID from user signed up schema.

A

Yes, tooling, actually go to this URL to resolve to yes.

E

Wait: yes, yes, and no! Yes, and no so.

A

It's an idea sometimes, but but in some other cases it's a.

E

Urine it's up to the to their tool. It's up.

A

E

Not in the spec, it's not in the spec okay. So, for example, there are the the hyper jump. I think it's it's going directly to the URL, but it's a decision on the tooling right. Correct me: if I'm wrong, Jonas, yeah and.

D

But that is your characteristic yes, so it means that, yes, you.

A

D

Have the right or invalid spec so.

A

That is yes, so then, adding this to the fact that we were saying like uh let's unify all our references in V3 right and we decided to use dollar ref everywhere right because using plain text, IDs seemed like a bad idea right, but uh by this. If, if we follow this approach, then we will be actually doing the plain text: ID approach, just with the dollar ref in in the beginning, but but um yeah the URL will actually be.

A

This could be something else, I don't know just the word user signed up uh without anything else, just the word, because it's a playing ID, it's a plain text, ID right, so so we we will actually be merging the two right. The plain text uh IDs with the dollar ref IDs, which yeah I don't know, I.

E

Mean it's really: it's.

A

Highly confusing for me.

E

Today: okay, no I just.

A

E

Say I find it super um useful yeah.

A

E

You have everything you can do whatever you want right, but I guess the the the concern here from Jonas is okay and how so we want all of these or whatever and how we ensure that. How do we say how.

B

To ensure consistency, yes,.

D

But in this example you, you can only use the reference to the schema, but, for example, if I want to make the reference in the next major version, of course, not in this spec to the channels and then I make want to make something like the plain text, not the exactly the local reference tool to the given General, but by this ID. So the channel object need also this dollar idea to recognize that you wanna make the reference to this channel.

B

So I'm not talking about whether we should introduce dollar ID into any of our components. That's not within the discussion here, because that's a whole lot of thing. Yeah.

D

But we need some, you know so.

B

It's it's only relevant in in this specific use case. If we have to keep on track, it's it's very specifically for when you define or with your reference payload that and especially if it's a Json schema, because uh because those schemas can define an ID, for example Avro.

B

You probably can't have any of these Behavior right, because you can't there's, there's no mechan, there's no, there's nothing within the standard that says this is something that you should be able to do so this is very Json specific, but it's to create a consistency across our tools or any tool in in that matter.

B

So it's I, don't know.

D

But if you have the given schema with this ID in separate file, so in the from tooling perspective you you need to know that you have the separate file. Yes, yes,.

B

And you would have to manually load it into the tool.

D

Yeah yeah exactly.

B

Before loading this, in order to understand that you shouldn't be able to go out into the the web to fetch this because I locally have it in my cache, basically yeah.

A

We should probably in this case we will probably go through all the schema definitions and the reference all of them and put.

D

It and then make the difference and.

A

Then and then resolve all the rest of the dollar revs.

D

A

Because otherwise, yes, they will go to URL, always right. Yes,.

E

Kind of like a cache yeah.

A

Yeah, but it would work kind of a cash in a specific order. That's what I mean you first go to schemas and then the rest of the document I.

D

I, don't remember where, but I found one example, maybe in our case or maybe in another Repository, but if you um name the given scheme minus with some domain in this case my schema registry Arc, then probably you need this domain for every IDs, because some tooling, as I remember, don't support the separate and the domain name.

D

So it means, for example, if you have two schemas with different domain names and if tool yeah check, you know Traverse or schemas, and check that the IDS and and check the first ideas it treats the first domain as important. So probably the second one or the second reference to the another schema is another domain probably doesn't work as I. Remember, yeah, that's I mean it's very very hard to explain that yeah yeah I need to to probably I have to write some examples that you can understand, but yeah.

B

D

B

Down to a should our trolling support it on that.

D

A

A

A

I mean, if you ask me for my opinion, like don't introduce this level of complexity. Please let.

E

Me make decisions here, but no no.

A

That's true exactly we don't make decisions but you're asking me for my opinion. My opinion is hell. No then you can ask on the I mean you're. You already asked on the on the discussion right. So yes,.

B

So I sent you a link and I want each and every one of you to write a comment specifying this.

A

B

A

You very much so I'm gonna I would say that I'm gonna come and tell no no, without that other context.

B

uh So, just to spice it up, this is allowed in open API, 3.1.

D

B

So, let's open this yeah yeah.

D

But in open API you can also Define. Yes, in which dialogue do you want to Define this reference as I? Remember so you have two options: the newest one with IDs or the the you know, the previous Behavior.

B

I, don't know that specific part, though I.

D

Haven't read that yeah, but okay I will check in open API specification so.

B

Please do okay.

B

So yeah, that's, let's finish on that note. What an interesting discussion huh only 10 more use cases to go through with references so.

C

The outcome is that we add the dollar ID at the root of using API file. Right no.

A

B

Let's not make it more complicated.

C

B

C

Was a request to a dollar schema in.

E

B

C

Which was actually made sense.

A

It makes sense yeah.

B

Definitely all right.

A

um I mean all of this actually makes sense. I understand the value of all these things. I just think that is not solving any problem for us and it's it's only making it more difficult. That's it say, I understand, what's the value of this approach and so on, but.

B

That's it, oh, they have that feeling.

C

Like but what really is the value like, so you don't have to repeat the URL or what.

A

I would say the value it's in in many cases. It's like you want uh you. If you use URLs, you will have uniquely identified messages, for instance, so you will have Uris right as as identifiers um like, like Sergio, said before that you can have in schemas. You can have some kind of cash because they they are URLs they don't have to. But if you make them URLs uh it's easier to make it unique across multiple documents as well.

A

um So you don't have this problem that I found users signed up defined into different documents, with different messages with different schemas, or something like that, because yeah it's it's harder because well, actually, I don't know if it's harder, but I mean it. It seems easier to me that um they, the ideas, become normalized I, would say right.

B

A

And also, if, if they are defined on the schemas, uh they don't have to go to internet right to resolve them so important.

D

To remember where the given schema is in which file, but we only know that, given schema exists in.

A

Your system, but the thing is here, is that uh so right now what we're proposing people is that they do uh hashtags, less components, key mass and the name of the schema, which is referencing. A schema by its location like where is it defined, is inside components schema whatever and with the ID it can be defined anywhere right anywhere else. That is always going to be the same ID right, the same URL independently.

A

If you move it here or there or you or like magic, was saying like uh I changed the the the with the previous discussion right, um if I, what, if I change the the key on the components right component schemas I, have the key user signed up with the schema definition? What if I change user sign up to I? Don't know all lowercase uh or something like that?

A

um In this case, it will not break right because yeah, the key is one thing and uh and um and the message definition and the message ID is a different one. Right, I, just don't think it's providing any value. Press I understand the value for some people, but yeah.

B

Yeah, okay, so I just wanted to highlight this before ending, because this is what we wrote as the expected behavior that by default it should be turned on that this should be automatically matched for the payload, but it can be turned off as well.

B

Yeah, that's kind of the conclusion part, especially because we have no merit for what is expected.

A

So this is actually what you're saying is that, even if we don't support this like we turn it off right, uh it's still compliant.

B

If I mean it will still be, it will still be compliant because there's nothing within draft 7. At least that defines this behavior and.

E

It's not until.

B

That's where it gets tricky, um that's for next time. Okay,.

A

I think it's good yeah. Yes, we keep it short.

B

So so Lucas you are actually right, I hate to say it as usual.

C

But I'm like to conclude I'm and with the idea I'm even confused in pure Json schema. We have it in our schemas. Yes for some time because of Fiona's.

B

C

No sorry I I, I I, don't get it like, um maybe because I I just did not read the feature but like it for me, it's super confusing I, see definitions and I look into definitions, but there it's in there and the tooling fails.

D

Yeah, but in this case, if, if you exactly call the schema with the same domain, but this doesn't exactly exists and you exactly want to make the reference to this domain, you know to your local. So then you can have a problem.

D

Yes, I mean it's very, very educated, but you know I I really like it so, but.

E

You don't need that. Let's think that you, you won't need that.

D

Ever yeah yeah exactly but yeah.

A

We won't need that ever as well, sir.

A

I think so, of course, by the way Dennis.

B

Thank you. Now we only missing three other comments, work all right, I guess we didn't make it to your feature, request magic, but I'm gonna, write it up for next time as well. Yeah.

D

B

Have we have a request reply for next time, uh an update there, so, let's, let's see how many we make yeah we're, probably gonna, have to bet again how many issues we are, but.

D

We should have the meeting, not biblically, but you know in every route all right, Thinking, Out, Loud. Yes,.

A

I will only be able to join if we keep the same schedule. uh I will only be able to join one uh like every two weeks like now.

B

Okay, I I, don't think we have enough to discuss or having it once a week. I know that we have a few things now, but then the last two or three weeks. It hasn't been that much yeah.

B

B

We can always bring it up as an issue.

C

Every two weeks, it's too much pressure, yeah.

B

Every week every week.

C

And I mean every two weeks: it's already left, that's right. Okay,.

B

E

Are you okay, Lucas.

C

No, no because I remember like because, if you understand he reminded me of this dollar ID and I, now.

B

A

C

Because we have to fix our schemas now, just because of this ID like we want to be perfect, so we have this, but now uh GS code tooling doesn't support it, so we have to work around it.

C

So it's always great. You.

D

Should still use the draft seven? Yes, so.

B

It's a maybe I, think I, don't think we have tested it if we changed all the schemas to the newer drafts. I, don't think. That's part of the detail, though, so it's yeah.

B

Because, as far as I could tell we got regarding the vs code, it was, um it was only part of it where references didn't work, uh something.

C

Like internal failing, it's failing with references, let's see, what's what happens when once we fix these.

B

We can always make them absolute in the pencil yeah.

C

Yeah, so we have a workaround yeah, definitely and I'm gonna work on it, but yeah.

C

That's not how it should work. I.

B

Think I did have a custom bundler before switching to a library, but it's somewhere within git history and Sample requests that I don't know where is and so you're welcome to try and find it. Where I did the manual bundling I.

C

Think that I mean my my idea is that the way I plug in into current bundling process is that I leave your bundling and then at the end, I run a function that goes through what you bundled and fix it. That's.

C

That's my engineering. Yes,.

A

Perfect, it's a additive in engineering.

A

It only adds transactional engineering.

B

All right, let's finish it off, we are out of time and last remarks before saying goodbye.

C

B

A pleasure right as always see y'all.