Internet Engineering Task Force Interim Meetings, 15 Jul 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: CBOR WG Interim Meeting, 2020-07-15

Description

CBOR WG Interim Meeting, 2020-07-15

A

And it's recording so welcome to seabor interim the last before itf 108. The agenda is okay, uh jim already has posted it. um The code dmd link and the first item on the agenda is the erratas for 1749, so just making sure going through the list and making sure that we have um addressed all of them so see. I don't know if someone wants to take.

B

Okay, so I went ahead and listed all of the errata in the in in the minutes, um so the big ones I'm looking at are the canonical the three canonical rules, ones.

B

But we made the changes that most we made most of changes that they suggest. But technically we did that because we did a best document, not because it was narrated on the old document.

B

Should we tell barry to just say close as want fixed with a comment that actually was fixed in the.

C

Abyss yeah: I went through all these uh errata and checked whether we had to major make an issue and made an issue and so on, but I just didn't think about disposing of the reports at this point in time. So uh sorry about that, oh that's, fine yeah, but it essentially means you. You have to do the work again. So that's not so uh fine.

B

So does anybody object to saying there'll be tubbery on those three that we say.

B

Closing this won't fix but put a description on that says they actually are fixed in this document.

C

Yeah the question is: is this a hold for document update thing I think in in many cases these errata actually were?

C

Rather, you have made a different version.

B

Right, that's what the quantization rules are.

C

So I think we are still rejecting uh these. Even so, even though the notes could say uh that this is addressed in in the best.

C

Document, I don't, there is a won't fix. I think we have hold for document update and we have rejected.

B

A

B

So I think that does make sense for those three um five, four three four as near as I can tell you just removed that example. So I think the correct answer is hold for document update.

C

B

um The same thing would be true with five seven six, three.

C

Don't have them in front of me. I have to.

B

C

Get another browser window open here.

B

What'd, I say five: six, seven.

B

C

Does that exist.

A

It's five, seven, six.

C

Yeah confused in which order are these.

A

It's the one. A tag always applies to the item that is directly followed by it, and a tag always applies to the item that directly follows it.

C

Yeah, that's definitely valid and uh it's uh hold for document update.

B

Right and in point, in fact, the text now says a tag applies semantics to the to to what's what it applies to.

C

The check applies to what it applies to that's, probably true,.

B

I don't remember where the exact text was.

C

C

Applies semantics to the data item it encloses.

B

B

So- and I think, that's probably better than the text that was there originally anyway.

C

Yeah but the the report is verified and the disposition is hopefully on that.

B

B

Okay, so I think that's all of the erratas. I can send the piece of mail to barry.

A

Sounds good great, so uh next item was the keyboard tag for date, status mike.

D

I uh created the next draft, which is intended to incorporate resolutions to the comments from jim, michael richardson and yours francesca.

D

I didn't use text, but I took some inspiration from it and, as you've seen on the list, I asked barry to get it posted I'd, quite frankly, lost track. The fact that we were even going to have a blackout period given everything's virtual, but that should be up shortly. Among other things, francesca removes the comma from the number and it removes the rfc 2119 language, because you're right that used none of the keywords.

A

Yeah, that was just an id.

D

Sure I've never written an rfc draft that didn't use any, but realizing this is just making declarations. There's no protocol. I guess it made sense. I searched you know the document to see. Are there any uses of may or should that should have been uppercase and there weren't so took it.

C

Out yeah, I don't know if you have seen this, I made two more minor uh comments.

C

uh One is that you could actually supply the the strings and numbers of the john dennin example, and the other one is that the the augmented date that you are talking about actually is a date time and probably should be called that way with that rhythm.

E

um I agree with the date time thing: your date.

D

Dot perspective, I don't know what programming language that's in well,.

C

That's that's just I just copied the calculation that I did over from from my terminal window. So the the interesting things are, the string 1940 dash uh something and the numbers is 10 000, something don't rest of the text.

C

The programming language is ruby by the way. Okay,.

D

um Yeah, I could add those example values.

C

I think would be really useful for people to check that they are doing the right thing.

D

Yeah, I I agree. I'd I'd be glad to do that. I want to let get this one posted, but then I'll do that again.

E

D

Paying attention.

A

Yes, so I the shepherd review is ready. I just need to send you a confirmation of apr declaration etc, but the rest is done. I was just waiting on on the next version, because that's uh it's gonna do some updates, but I will post it now and then I will update it later when new version comes, but we're still waiting. So then you need to still do some modification to add this example that uh carson was talking about.

D

Yeah I can, I can do that this week.

A

Okay, because I I need I need to have that version before we can push the button so.

E

C

So then we of course, don't want to to you to push the button because he's already drowned in rough during.

A

70, actually he just sent an email saying that he will uh he asked or he will ask I don't know the secretary to post the draft uh yeah. He said he's doing that: okay,.

C

A

We have one more uh modification.

C

If I understand.

A

Yeah so yeah, I don't know uh I'm fine with. I don't know like I'm fine with waiting until uh next. In monday, in two weeks or 10 days,.

F

A

It reopens but uh yeah.

E

I will just wait for the next.

A

Version and then and then.

D

Okay, but you can post the shepherd review and send me the idea.

A

Yeah yeah, I will yeah. I will do that in the meantime.

D

All right, thank you.

A

No problems, okay, anything else.

A

I think that was it, um and the next item is the cyborg draft, uh so jim.

E

Lay the review I'm going to go because I have to get ready for an external appointment, so thank you for doing this. Thank you mike.

A

Bye bye, so uh the seabor oy draft uh jim did the review already and carsten. uh Has, I don't know if you have completely addressed it in this pull request.

B

C

And so there are things that we still can discuss. I think, but uh I also think I have mostly addressed it, um so this happened before I pushed 07 is. Do I remember that.

A

No, I think this the review was on o7 jim's review was on like super quickly on version 7.

C

Okay, so in that case I have to push a dash 08.

A

C

I didn't realize that because it's it's over github and I could have put that on monday- and I just just didn't- remove that sorry. Yeah.

A

No, no so, but so just I guess it would be good to know if you believe the request addresses all of jim's comments.

A

If you can remember.

C

Well substantially so.

A

C

Again, I think we can still improve, but we also can improve this. Why? This is a group document.

A

B

I think that the only things that were not addressed were do we want to have any control operators to compute relative addresses, and do we want to try to figure out how to apply things recursively, I'm perfectly willing not to do the recursive thing.

C

Yeah, I think, that's a discussion we need to have, but I think it's not not. uh We can adopt this if this happens and we cannot adopt this. If something else happens, this kind of decision.

B

Oh definitely not.

A

So I guess um now it's a good time to start uh or to for next version. I guess we can do a call for adoption.

C

You can do it for this.

A

Okay, um sure I I guess we can start a call for adoption. I don't know if, because this is um already a charter item, so I don't know if we can um make it faster or they just go the two weeks and and.

C

Well, I I believe that adoption calls should be shot one week.

A

C

Yeah- and I thought this always was a little bit too long, because if, if somebody doesn't uh react, they can still fix it in the working group. So it's not like. We need two weeks to do a detailed review and and all that.

C

So 24.18 is no help here at all.

A

Okay, so I think we can start this dream. What do you say.

B

Yeah, I don't see any problems, not starting with that.

A

uh If uh people readying the call or willing to support that would be good.

A

A

As we said, we should pull interest in in the calls as well, since we can do that.

C

So so hank can. Can you do a two sentence statement about where you need this.

F

Yes, uh so of course, again requirements swap over from from rats and stuff. uh This is about reference integrity, measurements. um Some of these actually use semantics, although there are other things outside the iitf to take a step back that is they're using certificates for reference integrity, measurements. They are already.

F

They are defined a lot of semantics and referencing them in a meaningful manner, which even means to decompose them down to the value and use. The value here in in, for example, cause with or other concise representations using zebra is super helpful.

F

It makes interoperability uh yeah like like how it should be, and uh therefore there is a very driven and current requirements you could say or need for a oil representation in suburbia tags that would be making the life for a lot of machines and can do a lot of errors and failures a lot of easier. I think.

C

A

Yes, thank you, okay. So then, I think we can move to the next agenda item, uh that is, the cdl module design continued and the cdl modules sockets and features inspired by castings emails in the mailing list.

A

Do we want to continue this discussion today casting or.

C

Well, it depends on whether people have an opinion on that. Yet I mean it's: it's pretty heavy material um and uh one one has to think a little bit about which direction you want to uh go here. But if people have an opinion that that would be good to hear.

B

Carsten I'd like to get a bit of clarification because I'm not positive, I understood things so, okay, I understand we're basically talking about two types of names: rule names.

B

And don't necessarily have semantic meaning, so what you're talking about is saying, we would say is, is basically saying something along the lines of these are the things we suggest that you use and another question is whether or not you want to be able to say to export things which don't have semantic meaning, but have some sort of numeric media.

B

C

The the evasion behind this this may be a little bit opaque, was that if we have something that defines the external interface of a module, that might actually also be the place where we describe how the semantically fit to that.

C

So that is, of course, an assumption that I didn't make explicit, so I apologize uh for that, but cdda right now doesn't give you a way to link rule names or keys and group entries uh to, for instance, rdf names or the fuis, which would be you needed to to link it to some anthologies uh out there, and what I'm proposing is is to think about the issue, how we are going to do that. We don't have a an answer due to the question here. I only have the question.

B

Okay, I think it gives me a bit more context.

C

So, for instance, if you look at the end of the name with this uh label blur, which is used in two different contexts, one talking about stakes and one talking about colors and we probably need a way. If we we map this to some external uh concept, uh then we need a way to actually identify this.

C

So rule names are unique, that's easy, but those label names may be used in different places with different semantics and we probably have to do something to to reach in there. So this is a little bit like enums are handled and in many programming languages where the the enums have names, but you also can can anchor the enum name in a type name.

C

So so you get a fully qualified name for the enum and something like this. We need here.

B

Because everyone in their structure, wants to have an e name called error.

C

For instance, yes, yeah.

F

It's a very good example: yes, and sometimes you but error- that's actually actually the perfect type, because you're writing a specification and you're aware that you have a early consumer and you are basically catering requirements coming from there, and you are aware that some of the keys or labels will be used there. Also. So this disambiguation, you can do with some foresight, of course, but not always- and that is the point.

C

Yeah the particular comes up. If you have a specification that defines colors for steaks or cooking points for steaks, and then you need to reference a specification that defines colors and suddenly you have this this. uh Well, it's not really a name conflict, but you you have two users for the same name, and uh so there needs to be a way to actually anchor that name into a specific context.

C

So I think that that's a bit of a weakness of our original cdl design, but since we didn't define external interfaces, we didn't address.

F

F

Yeah we did a special, specific implicit cut yeah with the design.

F

B

So at this point, basically, we need to have people read it and digest it and start coming back with questions comments, suggestions.

C

Yes and and of course, we could uh further facilitate that by actually coming up with solution proposals. So it's easier to see what what the the opaque words in this email actually mean. But you probably need a couple of months to come up with a solution proposal.

C

So if, if you can, can generate some guidance now, in particular by thinking about specifications you have and and how we would handle that uh there, that would be very useful.

A

Any way that we can um ping people that would have an opinion from other working groups that advertise this or you know, reach out.

F

I actually have we might have a early customer that we already talked to, and that is the suit working group. Maybe we can make them interested again, spend some time after now, there's cut off and now a little bit later. So um so I think there is.

F

Environments are coming from there also, and maybe we can make them.

A

Yeah, that would be great to have uh to have them participate in this discussion.

B

C

Yeah, so there are some 50, rfcs and and internet drafts that reference cddl. I'm not sure all of these have useful examples uh here, but uh going through those uh we might be able to to identify further groups.

A

That's actually something that we did when we prepare. If you remember the questionnaire work, yes,.

A

Yes, that was on a high level. Yes,.

C

And we could now go to them with a specific question, and- and uh again I I'd prefer to do this, based on uh some some prototype, some strawman designs and and then hear from people how this design would fit with what they are doing.

A

Yeah now, what I meant is that we already had identified several people who were using. I mean that list probably has changed a little bit, but it could be a good starting point.

F

Again, that's true.

A

Okay, um then, the next item on the agenda is 108 scheduling, so we need to post the draft agenda today.

A

So right now we have a couple of items, so I I think I wrote down everything that people asked to either talk about or present um so right now we have the cdl updates uh captain. I think you asked for presentation time for draft board and seaboard pact.

C

A

C

That would be nice if everybody had read this.

A

Yes, I mean: if it's on the agenda, then everybody will have, of course.

A

Then, for for the date tag document, maybe we don't need to discuss it since um yeah. Maybe we can just give us a short status update. The chairs can do that and I'm not sure that we need to actually talk about it.

B

If we push the button, I don't think we need to talk about it.

A

C

Except if we already have feedback.

A

Yes, which, I doubt.

A

Maybe we can have it in the agenda just as a placeholder and then we see if we just give a two-minute status update or is it like that and then I put in the uh the oil document um again do we need to talk about? I mean if you have an update casting. I.

C

Yeah, I definitely will have an update on on sunday on sunday july 26th, uh so, uh depending on how the adoption call goes, we might be discussing ietf sibo attack, zero, zero.

A

C

A

We get feedback, we should definitely think about it. um Yeah I don't know otherwise, uh if you you have answered jim's um review, but I don't know if you want to talk about something else. Otherwise,.

A

Yeah and that the chairs can do that without necessarily taking too much time. But if you already have like issues that you'd like to discuss in the working group that that would take more time because we only have 50 minutes. So we have to be smart with the time.

A

Management but yeah, we can keep some time anyway, like uh let's say five ten minutes depending on what's the status.

C

B

Carson, have you heard anything from sean.

C

No okay- I I still have to address him personally. uh I didn't do that.

A

Yet, okay, anything more that we want to add to the agenda.

C

um Well, at some point, we probably should decide what the cause of action for the notable tags document is.

C

So I'm not sure we have to decide it in this meeting, but.

C

I think we should come up with a plan.

A

And carson, how much time do you want for the packed document to go back.

C

Well, that depends a lot on on what questions people have. So this this document has been discussed between various people since 2013, and I think I had a presentation that in 2017 about it already in in the working group- and um the question is: are people happy with with that, as as where the the various lines of thinking converge or do we do want to do that at all? And so on? Those are questions that I think we need to discuss.

C

So this is definitely a document that has uh the need for a good discussion before work group adoption.

A

A

Yeah, I I'm just thinking about the times now, because uh yeah of only of the 50 minutes is very short. I think the shortest we've ever had.

C

A

But maybe I will uh let's say: cdl updates. Do you have an idea, or I know it's hard, it's hard to give an estimate, but.

C

Yeah, the meeting probably is a week too early to discuss the implementation of the abn f work, so that should be done in early august, but not not in late july.

B

Is the only thing on the on the on on that document? Just getting some implementation experience, because I don't.

E

C

B

I don't remember any issues anymore on it.

C

So the design looks good, but uh people need to play with it before we can say. This is the way we want to do.

A

A

Okay, but we will then draft something we'll keep the time quite vague for now, and hopefully we can uh get it more defined before the meeting.

A

Anything else that we want to talk about.

A

A

A

If there is nothing else, I think we can close the meeting and.

C

Let me just do the quick advertisement for the asdf buff.

A

C

So please go to the asdf buff on july 28th and also I might report that I have submitted the first thrawn for jsonpath document that I sent that to dispatch. There has been some good discussion in the on the sibo mailing list. You have seen on the json mailing list and on dispatch itself and the the interesting question now is uh whether we can do this in one of these groups or get sucked up by the http api uh working group, which will have the charter to do everything that you need to do.

C

Http apis, which is a bit much for my point of view. But we need to have that discussion so that that might block uh the discussion where the the jsonpathworks was.

C

So my evil plan, of course, is to to make jsonpath useful for cbo in general and not just for json, but you you probably already guessed that.

A

Yeah, wait. We guessed it, but thank you for the advertisement very useful. um Okay, then I think we can close the meeting and maybe see you in cfrg and otherwise see you in two weeks for the virtual meeting.

A

C

A