Internet Engineering Task Force Interim Meetings, 8 Apr 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: CBOR WG Interim Meeting, 2020-04-08

Description

CBOR WG Interim Meeting, 2020-04-08

A

And it's recording so to see Bohr working group meeting. This is Francesca and Jim is my co-chair, some useful links mailing list and the jabber please join and etherpad from unit taking. Thank you Michael for taking a minute and WebEx information, because it's always good to have them somewhere.

A

This is an official ITF meeting, so the note well apply. If you haven't. Please take a moment to read and understand, and if you have any question feel free to ask me Jim or anybody or anybody around as a reminder, minutes are taken, this meeting is recorded and the presence is logged. Yes, please go into the ether pad and write your name at the end of the ether pad as a virtual blue sheet.

A

This is our agenda for today, so we're in the introduction part right now and then we will move forward to working documents where custom will talk about the Seaboard base status and we have a couple more topics in the other part.

A

So we want to talk about the date and time tag document, so I haven't seen Mike in the call he had asked for time, and so Carsten I know that there's been some discussion in the mailing list will continue that here and if time allows we'll talk about the CDL freezer document or how to continue working on CDL some time for wrap-up.

A

So this is a working group status update. So we've had six interim since ITF 106, so we are pretty used to having virtual interims and we think they're very useful, and we would like to see more participation so for whoever is here who never been to a working group interim before pleased to join. They only take one hour, often less than that, and we discuss topics that will pop in the main.

A

This time we think yeah, it's it's a check you can find minutes and whenever the worst slides slides for those meetings at this link and yes since ITF 106, we've had yeah super rate tags and super sequence that were published. So congratulations to the working group and the others with that, and also we've been working on the 70 49.

A

So the working group last call ended the 23rd of March. There was a couple of issues. They were open and we're still looking for reviews from implementers.

A

There was some action points from the last interim to read carefully and review the text that was implemented to close issue 164 and 165 and I'm sure Dustin will go more into detail for that.

A

The status update of sake and yes, just scheduling, we will continue on having these interims on the same timeslot, which is this one Wednesday between 5:00 and 6:00 p.m. CST, every two weeks from now until the 14 of July, and if we feel like we don't have topic to discuss where the meeting will be cancelled. But otherwise we will be having the meeting and I didn't ask. But is there any agenda bashing.

A

If not, then we can move to Kirsten.

A

Recording my sharing.

A

I hope Carson is back I.

A

Was too quick with the introduction, a.

B

Long time to get his coffee Oh actually.

C

The the problem is that the Chrome application chosen takes some polynomial time quadratic or cubic I, don't know by the number of applications you have Ultima by the number of replication windows. So it takes me about 20 seconds to actually select an application. I want to share, and the fun part is during those 20 seconds. I cannot talk, so that was the gap you just experienced.

C

Okay, so I'm not seeing Mike Jones at this point in time, but he might still be joining so I have a gap in my slides for Mike to chime in okay. So I want to talk about three things: the this you have a status, the the time tag and if we have time the CDD, air, freezer and I think we will have time mushing, the whole meeting will be pretty short, so we did a working class call and everybody seems to be happy.

C

Everybody seems to be sleeping or everybody seems to be busy with realigning the world with covered 19, which is my problem right now, because we we have. This is really University one here and yeah. So we are getting some comments from implementers, so it it has helped to a large sum of them, but this is coming in slowly. So I think we just have to plan for a little bit more time for for more comments to come in and and one one other comment you came in like 20 minutes ago, so I think it working.

C

So let me just quickly talk through the three issues. We have right now on the repository, so one observation is that the text currently says that not all implementations might do 64-bit integers, but it doesn't warn people that the same might be true about one plus, eight okay, so text with a 64-bit tag number and that's probably a very small piece of text that needs to be added by but I agree. This should be added and why we were discussing this.

C

There also was the observation that we are not going to get into 1.1 plus eight land very soon, because the the designated expert can control locations in 1, +, 2, +, 1, +, 4 LED. Now, unfortunately, the answer is no to that question, because a 1 + 2 + 1 + 4 specification required.

C

So the designated expert can not provide any back pressure there. Interestingly, a Jana has has asked me about allocations in in those spaces as well, just to make sure that they are doing everything right, but I'm I don't have the power of a designated expert there. So I don't have the power to push back, and that is the result of discussions. We have heard and I'm a little bit in the rough on those discussions so of course, I'm clinging to every every opportunity to bring this up again.

C

So do we want to put a little bit more control over the 1 + 2 allocations, maybe at least than we have at the moment.

D

Carsten is a specification required is and first-come first-serve its.

C

First confessor, so officially the desiccant expert is not in the room.

C

D

Think I maybe may agree with you that I'm not happy madness is very happy on that.

A

E

A

One plus two.

A

C

Me just show the text that it's currently in the draft so zero to twenty three, our standard section 24 to 255 our specification required and everything else is first-come, first-serve and I think we should just put the 1 plus 2, which is 256 to 65535.

C

A

Kind of remember did this discussion, but, like I know, we talked about this. I cannot remember right now,.

A

Those would be first-come, first-served, so I think I need to go. Look at the minutes again. Yeah, chair, hat off, also I think it will make sense to have 1 plus 2 into X. We.

C

Also could do something like keep the 1 plus 2 open to first-come, first-serve, but only part of it. So we could say the first half 4711 is just a syntactic meta variable here, part of the 1 plus 1 plus 2, would be expert review and the rest of the 1 plus 2 together were the 1 plus 4 and 1 plus 8 would be FCFS.

C

There are many questions, but what is a good video for 4711.

A

Okay, if nobody in the meeting has opinion about this I suggest we take the action point to look in the minutes and we bring the discussion back to the mailing list and we take a decision really quickly because we don't want to delay this obviously. But we want to give everybody a chance to give their opinion.

A

My personal feeling is that yeah, it would make sense to have these expert review I'm, not sure about half of these expert review. Half of the half of these come.

B

A

Serve because then it I mean it's very one has to choose. Where do you split it right.

A

It's very arbitrary.

C

Yes, but I mean, if you don't know what to do, giving half to one is giving half to the other it's kind of the default option. So I could imagine that that's a sense of a thing to do that.

A

Could make everybody happy but yeah, let's bring it to the main list and Esther's comments.

F

Franchesca this is our I just wanted to chime in and say I also strongly prefer that at least part of the 1+2 range is expert review. Otherwise we're not using expert review which doesn't make any sense at all and I. Don't I, don't care a goodly amount say several thousand at least gotta be expert review. It seems to me.

A

Okay, thank you.

C

Okay and as I said, there will be a few words added just pointing out that people who do not do 64-bit integers, probably won't do one plus eight tags.

C

If you 176, that's, really an implementer consideration which I never have thought about, because I don't need it in my implementation.

C

Emile said that maybe we should reserve one tag number for knotek, so in an implementation where you just have integer, where you write the tag number that applies to something you can write that you can store this special value which, if we had thought about it in 2013, might have been zero, but we didn't so so it probably will be something else like 66 right now. So when you hand that up in a data structure, it means this item does not have a tag.

C

Now. Jim noted that actually this doesn't quite help an implementation, because there also might be multiple tags, so you already have to have a way to deal with multiple tags, but I still can't imagine implementations where it's just nice to have a tag number where we know that that's never going to be actually used.

G

So is to go on the wire, or is this something as a convenience for internal implementations, its implementation.

C

And the hard part of this is to probably say what happens when it's on the wire, because it's not supposed to be on the wire, take them, let's, to put it there. So.

G

There's no advantage to making it short, so that's a good point and I would say: if it shows up on the wire, then it should be violently ignored.

C

Yeah, but we cannot do this retro actively to existing implementations, so there will always be some implementations who will treat the presence of this tag differently from the absence of the tag that might be used by them.

G

What what would they do with a tag they just didn't recognize then world.

C

Probably rejects the thing well,.

G

That may be appropriate if it, if it, if it wanders onto the wire then, and that actually also would would identify very quickly that some some guy is leaking because they would it not wouldn't interoperate right.

C

Yeah so I'm not concerned with into a probability I'm concerned with possible avenues where, where you have multiple entities interpreting the same SIBO and they come to different conclusions as to what it is and and that's the security consideration for a tag like this I. Think.

C

So not none of this actually has to go into 70 49 this. This can be just a tag. Registration, but I think we still want to do this right if we handle this issue that has been raised.

C

A

What you wrote in the parentheses here, it can be done outside of flash in parallel to make sense to me.

C

Well, of course, it's more prominent for implementers if it's in the document, so I think they just argued that it would be better to in coverage no tag now.

C

But yeah we're not supposed to invent new stuff in infinite standard, and this is not exactly an internal unity problem we are referring to here. So Michael also things we should do it now.

F

Yeah I hope you I will wait. First, an NRI I, just I agree that it should be done now and we should put it in 70 49 and this and try not to have an interoperability problem, that we could clarify high-end clothes.

C

To answer Alex's jabber question whether we have to argue about this for six months: I hope not, but it would be the first thing we actually change in the document. That is not just based on InterVarsity considerations, but I hope we can get away with that. India's gee.

A

D

We define this I would expect that this would be used by some of the people who were there was it people somebody wanted a. This is a tag to say that this field is not present and I'd be willing to bet that the note attic field would be used for that purpose.

C

Yeah so yeah take this to the list, but I think the sense I get is it's worth spending the effort doing this now and I mean we're waiting for new input for more input anyway, so we might want to to do this now.

C

Okay, if no no other.

C

The the etherpad right, I, understood, Jim, saying something different from what I just heard.

C

Jim, do you want to amplify your point.

D

Jim is the etherpad writer, so he wrote down what he thought. He said.

C

Yeah, okay, we have known for that an undefined and so on.

C

I, don't think that that Jim.

G

What would the no tag, no tag be followed by a see, bore no I.

D

Think that's what he was doing, but I'm, not positive. That's I I was never completely clear what he thought he was going to do with that no tag.

D

It was some way that he wanted to deal with. This thing is not present so that tag followed by no would be certainly one way to do it he's didn't. He didn't want to use null or undefined, because that was not quite clear.

D

How would you say that the null is not pres not been? How do you say that Noah's not present.

C

It's the meaning of no, that it's not present.

D

I was kind of going into semantics. Okay,.

C

Maybe you can write a slightly longer exposition of your thoughts of the mania, so I get a better chance at understanding. It.

D

Yeah I'm totally going to do that.

C

Yeah I mean one problem. Of course we have is that these are all github issues at the moment and we have the usual problem of connecting the github issue commenter to any discussion that is going on on the mailing list, but yeah I think we know how to do this. We just should not forget doing it. Yeah.

E

This is saying we have this sorry for being late. We have this and other working groups. Basically, all the notifications go to the lists. It's can be spammy if there are the meetings and you work at stuff, but again that might be useful to.

A

C

Yeah I also prefer being explicit, bring.

A

To the list, what I mean for editorial issues or things like that? It doesn't list.

E

Requires discipline I.

A

Think this issue needs to discuss in a list for.

C

A

To summarize, what I've heard is many people supporting the win? This right, like adding this to the draft right now.

A

We need to discuss more on how this is. You stand.

C

Okay, thank you so the the issue that came in half an hour ago- or maybe he said he said he five minutes ago- was about the security considerations about hashes. It's wearing on that an attacker can turn oh one hash into an O n hash if they know how the hash is implemented and the security considerations currently only mentioned one mitigation strategy and of course it should mention that there are other mitigation strategies.

C

So I think that that's a no-brainer to add this to the security considerations, its its implementation advice and, of course we can expand implementation advice indefinitely, but maybe pointing to the strategy Java uses here is a good idea.

C

So that's actually all I have on C Bobby's at the moment.

A

Very close to be done, I like to say custom. You already mentioned that we've gotten very few comments.

A

Are you reaching out to implementers or have you reached out to implementers.

C

um I have and I still can expand that so I'm, the more I do this, the more spenny it gets because I'm engaging more and more venues. That really aren't quite meant for this, but but it's probably not too much of an etiquette violation to go ahead and ask for help, but I also think that some people will will need a week or two to actually react.

C

So I did some of them last weekend. So maybe we should plan a little bit more time.

A

Anyway, you needed to submit a new version correct yes, so what time plan should we have do you want to include this? Also in the new version or.

C

When is our next interim two weeks, two weeks, so maybe the plan should be to it will submit the next version in ten days, so we have still time to.

A

That, whatever comes out of discussion about the tag as well, yeah.

A

New version in ten days and keep.

A

A

Then we wait and we ask for more reviews from implementers or.

C

Where we can stay in the current state indefinitely, of course, right now is not really such a great time to ask for people's attention, because there are all these other things going on with readjusting your life and all that so I think it makes sense to to have a little bit more time edit, but I also think that, maybe by the end of the month, we should decide whether we can ship it. Oh yeah.

A

I started doing the Shepard review by the way, so I'm I'm gonna wait for the next version to finish that and then we can't ship it. You.

E

A

End of this month, that's that sound.

F

A

Other opinions is that too late.

E

This is saying: I, don't think that is too late. I think the argument of we are in a prolonged session frame right now and having the other turmoil in the world is a good justification to have this month.

E

A

I wanted also to say that considering the comments we got, yeah I mean waiting a week or two more doesn't.

A

Many more comments, anyway, yeah and also one thing. I know that in the data tracker is that the responsible ad right now is none so.

A

D

Wrong, if you go to be about Paige Barry is our responsible ad about.

A

Yes, by the way, welcome Barry and thank you for being our responsibility and thank you an ex a for all your work. Nice.

B

To be back, you responsible ad for the document or for the working group, I meant.

A

For the document I'll.

B

A

Assume it was you, but you know it just doesn't.

A

B

The responsible ad gets filled in as soon as an ad touches it. So if you request publication, then it fills it in until then, if I make, if I flag it as ad, is watching it'll get filled in, but I don't generally watch the documents until they come to me. So.

A

According to this, discussion expect to see a support base coming to you by the end of April.

C

Into 8150 to miss how much leg do we have.

A

According to mine stones, yeah.

C

I think this essentially should be in a cluster.

C

Maybe Jim does.

D

This is going to area directors as soon as the Sheppard report is written.

C

Okay, we are not that much behind we have a half a month or so great.

C

Okay, it's Mike here.

C

C

Hello, so I make a prompt to a slide which was a bit empty at the moment.

C

So there was a request to have a date tag. We have several time tags, but we don't have a way to identify a whole day. You cannot identify a whole day with a time check because, depending on timezone, it may be a different day in different parts of the world. So you really need a different tack, so we now have Mike's tag which is RFC 339 based. So it essentially looks like this and there's also the order idea to do an epoch based one and actually was it.

C

Take number of 100 I forget, but it's it's a lowercase D.

C

Which would have days since.

C

Eunuchs birthday.

C

So paid days since 1970 or 101, which is essentially add an offset version of M ready, so one could argue whether this should be an MD ID, but mmm babies are not exactly well understood by implementers, so maybe just doing this offset by 40,000 something so it starts at the eunuchs day. It doesn't make a big difference, so these are out there and the question really is because these are so fundamental is what what do we want to do about?

C

Those and I have argued against using a 1 plus 1 tag for the day time, but this is not really a strong opinion because of course, a day number is really short. It often will be a too bad number, so the whole thing is pretty compact, so this this might be a good reason to do that.

C

So when, when we discussed this on the mailing list, there briefly was was an idea to merge the true draft and then, of course the question is: do we maybe have a single tag for both?

C

We haven't done that with tag 0, & 1, so we have take 0 for text-based times and takes one for a book based x so that that would be a change, but it would be a change we could make so I think we should think about this for a short wire and so far, I think the the tendency and the main list was to do the same thing. We did with 0 and 1 and and just have two different texts.

C

So the outcome would be to have a one plus one for epoch and one plus two for text based.

C

And that would be those two numbers, one hundred and one thousand four. So.

E

This is Hank absolutely in favor of creating texts for things that are used on a daily basis, so to speak. So I am fine with distinguishing these two via text, because it's just simpler but I am very confused by the term epoch based, but because, if thousand 2005 or eight does not have an epoch, what is 0 1 1? So that's, probably the epoch of that Bay day based counter the start of the Common Era is probably also an epoch, and this is apparent in epoch based they counter.

E

Having said that, I am okay with calling it I, don't know text based an epoch based, although it's probably not right, we.

C

Don't have to give it a name. You.

E

Refer to them like this, so so, okay! So that's what everybody does exactly for for for common usage and for daily use it again, it's okay to reference them somehow, and that's fine and and as long as this doesn't temper with precision time tags I'm. Certainly the court.

E

C

Okay and the other things that are related to this are time of day and the time zone, so jerks originally proposed a apart from having the date would have a time of day tag.

C

He proposed tag, number 116, I'm, not sure that is so free and time zone tank.

C

You could essentially take apart and an hour C 3, 2, 3 9 style date, time into three different components.

C

With the time of day, it's pretty clear what you would do there second, since midnight, and the time zone would probably be an INR time zone designator or we just got a message from Mike.

E

So we only store in court.

C

Once I said, did you get that man.

C

I, just he will have an updated date. Egg draft published this week. Okay,.

A

We also agreed on merging yours and Mike's document last interim.

G

A

C

A

I think I thought that was your document and Mike's document. So.

C

My document is quite different: okay,.

A

So that was the New York's York's suggestion yeah.

F

A

So that would be the that would be the the update.

C

We don't have to discuss every tag allocation here in the row, but I think this is important enough for day-to-day work. I think said that we should get this right so essentially, at this point, I would say: let's just wait for what what Mike is coming up with.

C

C

Ok, so I think we can just wait for Mike's new version on this one, and then there is this time tag which has been around for a while and actually it's being used their implementations these days, the tag 1001 is being used in implementations, and this was just meant to be a more sophisticated version of tag 1 and and provides some extensibility which tag one doesn't have, and one question was whether maybe at some point we want to adopt this, so we can handle the whole thing in one go.

C

What is our strategy here, because I get I can continue working with Hank on this in my spare time and and eventually complete it, because the tags are Reggie registered, but maybe we want to have a common thread here.

E

So this is Angus edition in put on this three weeks. In average, I received the question: how do I write a timestamp and I'm sick and tired of explaining the whole thing, and that is this? Is this is good work I mean seriously? Then people might like be confused a little bit about it for very first beginning, because we already have tags yeah, yeah, no, a timestamp is global and and interoperable timestamp is not easy. Clocks are bad. They they function, not very good drift. They they, they had a timing.

E

Timing of some some some global actions like like I, don't know, revocation whatever that's very poor example. I know it's important to to to chime Bay, otherwise their windows, and then things get messy and and all of this, and we have a lot of timing constraints if we go with long, very long, long, long on tubes that reach to other planets and then suddenly timing becomes very, very important and- and so I think this this all this work we aggregated here is very, very useful, but not for the everyday use so to speak.

E

Some people have a very requirement for this, and a lot of people are just like well, this is might be overkill, but it is not. Let me assure you so that's my my additional comment to this, so we we aggregated as a lot of attributes for time, a representation here of reference points of linear time and F, for example, that and and so I'm giving back towards to couple.

C

So these are the things that could be added at some point. There are other things like precision and, and so on, that could be at it. So I think this should simply be handled based on whether people come in and and ask for how do you represent eggs in in sea war? So people have already been asking about time scales.

C

They haven't been asking about precision as far as I know so yeah this. This could evolve for a while, but we could also decide at some point to ship it as an RFC. Oh we. What we also could do by the way is generate a document that combines a number of useful tags and describes them. So so we don't get fifty RFC's on on tags, but just a single one that it's a little bit like an index of our seas of tanks that go beyond our c7 g-49.

A

B

Yeah, let me review this and think about.

A

Somebody would think that that having multiple RFC is easier rather than one document that is summary of attacks.

A

Interesting tags document also for I.

A

Don't know what I'm not I.

A

Don't have a strong opinion on this either I mean.

B

My general preference is to consolidate things in one place, but there are. There are good reasons to split it. So I don't know. Let me let me review the discussions.

A

Like we can, we can have an intercept, interesting tags right now, but.

A

More tags will be added to it.

C

Just come up with a rhythm, we publish it every five years.

C

That's a way serious and in in the time between it, would live as as a this document and and we could register tags against there this document. So we don't have to publish this this every six months or so, but having it in an editorial stable, useful way without thousand places where people have to prove some some descriptions that all differ in the editorial style and so on. That would be a useful service to the community. I think.

C

A

C

C

Well, who wants to work on this? Of course it would have a ton of contributors.

C

Are the people who registered something that that qualifies as interesting right.

B

C

B

Talking about a working group document that it's up to the work you to assign an editor and that's not necessarily the person who came up with the text, it's just the person who's responsible for keeping the document up.

E

But, but also we have a certain feasibility threshold here. Why I am full support of the ideas that I currently in the document, apparently I would always wage the additional content. That would be the best periods we are just discussing, as is that is that, does that reach a threshold? Does this validate the feasibility of of having having a new check here and yes, I think I would agree with very first of all not to split and scatter this around. That makes things very hard to read and actually to find, maybe you overlooked it.

E

That would be like confusing, very very on that that side of point of view, I think that an aggregate is a good thing and having that again as a living thing, if that is a real, stable and reliable possibility as this iterative business that we plan actually to do I'm, absolutely fine with that. If that works, I, don't know if that works, but it sounds like a feasible approach.

E

A

Taken and I agree with you and everything you said, including if that works. So what you have to maybe think a bit will that work I agree with you. It's.

B

Always able to plan to do these updates and then it's also the working groups get busy with other things and don't have the interest in keeping it up or the working group gets disbanded at some point and there's nobody keeping it out. So it it's never going to be a complete reference, but having one place to go, to look for these things and have it updated periodically is a good plan. It's worth it's worth. Thinking about.

A

We have three minutes left I, don't think we have time to talk about city, DLL, freezer document today, Caston, let's move that for the next interim. That's okay! Yes,.

A

Anything more you wanted to quickly add for for this, or we continue the discussion in the main list.

C

Maybe people could read the the section four of the freezer document before the next meeting and sure we could have an informed discussion. What we want to do with this.

A

Section for the freezer.

A

Yes, then I think we can wrap up here. So we have a bunch of action item in the minutes. Thank you, Jim for taking the minutes and we will be meeting in two weeks. At the same time, using.

E

The same WebEx.

A

You want to add Jim.

A

Nothing else, I think we can close the meeting. Thank you all for coming.