Internet Engineering Task Force 98, 28 Mar 2017

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF98-DOTS-20170328-1640

Description

DOTS meeting session at IETF98
2017/03/28 1640

A

Let's give it 30 seconds to give 12 to the dangerous and, if applicable, to the topic.

A

Okay, we could swap at any time.

B

Good afternoon, everyone, this is dots, I'm glad you're here to join us. My name is remington a do I not co-chairing with to buy his gun room, so let's get started so you've seen this I'm sure in previous previous sessions. The note well, of course, applies to this meeting. We have the blue sheets passed around. Please, please do make sure that you sign it. So we can provision the room in future meetings and wanted to thank our jabber scribes. In turn, takers, Frank and Bob very much appreciated agenda wise.

B

We have a full deck ahead of us. What you see there is not that different that what was on the mailing list, I think only one implementation report was added. Would anyone like to bash this agenda.

B

Ok, I take that as no let's proceed forward before we do that, just wanted to give you a quick status, update on where we stand in the working group, since we got together last in IDF 97, we had a virtual interim in February. We also had a design team meeting earlier this week on Monday to working group documents have been updated, and earlier this week we started a pole for adoption on the gated channel document. There's been some traffic on that.

B

Please do respond and we have at some time on the agenda to talk about that later. There are also a couple updates in the individual drafts we're going to get into a lot of these documents during our time here we have an updated use case, a show for new movement, since we last got together in terms of published documents on the IM DM.

B

With regards to the status of the solutions, it's actually hard to see there in the graphic, but we've we've had a couple of updates that you see reflected there in green, but in the case of the bottom three documents. The authors on that draft have said, for the time being, they're going to pause, progress on these drafts and put their energy into those top two drafts, and we can further this conversation when we talk about the solutions.

B

We'll talk a little bit more about milestones after we wrap up on the drafts, but I throw this up to point out that we are behind. We need to revise those milestones to reflect reality and we need to start really pressing forward with ours with our solutions kind of drafts, and we can talk about how to do that. Specifically again after we hear the presentations so to the agenda. First up for us is Roland who's going to be talking about our use cases.

C

So we have a clicking in one moment: hi I'm, Roland Evans over networks; okay, the cliff he is not working, I'm, not smart enough to unit. Here we go okay. Here we go so. um First of all, thanks very much to Daniel will go on in Frank John, who pushed this update out with the 04 draft of use cases. The structure has been simplified, so a lot of the stuff that was in the appendices um leading unnecessary and has been dropped.

C

We've added an orchestration section, not because dots is vital to the operation of orchestrators but to provide use cases a use case at least one escapes. That shows how a denouncement agaciro frustration system would interact with external entities via dots and what effect different dots messages you would have on the the actions of your frustration system takes in order to successfully mitigate.

D

C

D

C

Actually, a pretty lengthy section, that is an added there's, also a new intra-organizational use case. This is actually I'm having to do with a broadband access ISP who are protecting their customers against ddos attack scene and and how the use of dots would take place between their customers in their actual orchestration system and detox communication systems, even though the end customers themselves are. Actually you know what are there individuals, their small home offices? What have you there different entities?

C

This is actually considered an intra organizational use case because the broadband access ISP operate the entire network infrastructure. Here, it's not as if this is in like an enterprise who operate their own network infrastructure and have their own distinct transit edge or their purchasing transit from one or more multiple providers. So it's a little different. That's pretty intense extent, exhibit a pretty extensive section and the pro style is a bit more accessible as well. So that's really Oh war.

C

In a nutshell, 405 we have eight or nine proposed additional use cases that we want to put before the group to take a look at these are dealing with issues like multihoming. In a scenario where all the different upstream transit providers support dots, but they don't speak to one another or situation in which they do speak to another or a situation in which an endpoint network has multiple upstream providers, but only one of them actually supports. Dawson has yes mitigation service capability. We also have provider requesting assistance from a provider.

C

So an overload senere, an overflow scenario where, where a DDoS mitigation service provider is actively mitigating attacks and they're, reaching of their capacity in terms of transit and mitigation faster, so they signal to other operators for assistance using dots. And then we also have another use case in that group, which is where there's a federation of networks who have joined together for mutual DDoS mitigation assistance and in this particular case and a hosting colocation operator, are getting pummeled by a lot of traffic coming originating from a broadband access provider network.

C

And so they actually signal the broadband access provider network to suppress the outbound. You ask traffic in donating clear network, so those kinds of use cases will be in 05, there's a grammatical things and stylistic things that we want to improve from from 04 and they're the same thing in the orchestration section. It's not really changing the meaning of the content in a significant way. It should stylistic with anything else. We want to continue to harmonize our terminology, look at their graphs.

C

We have two additional terms and acronyms that we need to define as well, and one of our authors has changed affiliation, so we need to update the document vo5 to reflect his new village. So that's all I have questions comments.

B

So rolling win when, when.

C

Money you expect that though 5 I'm relatively soon, um seven of the nine additional use cases argue written, have to get the additional ones out there and then get them out. You know for comments before we look at incorporating so not long.

E

F

I quantify this with the date: okay,.

G

I expected date because not.

F

Long, its relative term, let.

C

Let's say roughly three weeks: we can. We can try to target that, but it's going to be dependent on how much feedback you know we get right back, thunderbolts.

F

And half the clue get we.

C

Can get you for that kind of time frame, but not not.

F

Very long, ok can I suggest with working group hit. First, you can actually, if you have already seven, you can close to seven and second, a feedback that comes I actually want to encourage that this feedback is given up on the mailing. Oh.

E

We love young, not.

F

Off channel and then it will be again helpful if you just make an update- and you know, don't don't be shy. Version numbers are cheap. We have plenty of them. So thank you. Okay,.

D

I Andrew Mortensen, arbor networks. You have the multihoming use case there with some dot servers that were not capable or was that, where.

C

We have an ongoing area where they're all dots, capable and I'm a looming area where there's multiples, but only one of them is dots capable. Can we talk about the operational impact that will have on the organization that is request? You know the things they need to take into account when they have to under it when you're under do not stab, and then they end up going from you know. Three upstream providers want right what kinds of things they have to take into account. That's actually a big part of dots.

C

That I think a lot of folks don't realize is that the technology itself is not magic, great. Oh, it's simply a lingua franca and in the use cases we hope to give the flavor of the other out-of-band things that have to be done in order to make dots work. Good.

D

Because one of the outstanding issues on the architecture draft is the multi-homing case, so hopefully the use cases. Yes.

C

D

Our last architecture.

C

Call in in one of the use cases its broadband access network, where there are multiple dots clients that are all signaling for mitigation systems, because it's a very broad-based hat and we talked about on whether or not it should be considered in scope for dots, the dots servers themselves to try to consolidate, alerting or whether that should be a meal up to the orchestration system. What have you so we then that the three cases included thanks.

B

So what did I ask if there are new additional questions? I, don't see how I come in Mike. Are there more use cases out there that you don't see documented in any of the existing individual drafts or in the current working group document?.

A

C

When we can get a discussion going, I'll go ahead this evening and send out the seven that are already written down written out now to help facilitate to get people comment on them. Hopefully, we'll disruption humans.

F

I'm wondering whether we should probe the room for who has read 04, because that is just recently released so did who has read 04 yet okay? Well, that's 76! A few people, a handful yeah potentially was too short notice before the meeting. So okay, then I think we need some more review and well, if you can shoot out some parts early, then maybe the reviewers will be more motivated.

F

It's always a little bit. People would be hesitated to review a draft if they know all the next one is coming in a week: yeah, okay, cool, then yeah I encourage that actually I think from a chair perspective in a couple of weeks or months. We need to go to last call. So I think this is the right time to start to do, reviews more. Thank you, Thanks, a room.

B

Thanks Roland next up is a discussion about requirements. Andrew.

D

It also occurred to me yesterday we discussed getting the use cases draft up into the dots wgf depository, so you can track the shoes there as well. So hopefully we want an update on that soon, all right. So this is an update on the dots requirements, largely this is going to cover the issues that we've been going through and our tracking on github.

D

So, let's start with the status, we do have a lot of issues a lot, but we have the via known issues on github in the github Tretton issue: tracker the open issues right now are the blockers to work in your glass call. So you can actually monitor our progress toward working group last call by tracking the open issues. Just because that's the case doesn't mean we're not accepting new issues. If you have concerns about the existing draft, please feel free to add issues.

D

You can do this on the github site or if you don't feel comfortable, that bring them up on the mailing list and we can make sure that they get tracked on github and pull requests are welcome to. So if you have changes suggested changes, you don't need to wait, you can go ahead and forth. The repository right now make changes to the markdown and then set up a pull request and we can just pull it right in.

D

So we've closed about half of the known issues since the interim meeting. Any feedback that we get on the pull request has been very helpful. There hasn't been a whole lot of interaction there, but what we've gotten so far has been very good and you should look for an 05 update, very short. They probably this week. I've got another poll request or two and we'll continue working toward the work of your last call I'm not going to go into too much detail here.

D

On the recently closed issues you can, you can find them on the issue. Tracker and I've also sent out I think a couple updates to the working for failing list on this, but I will touch on a couple of the more significant ones. Issue number nine pulled in some language from tears: signal channel draft about the path MTU, where the dots agents themselves are asked to try.

F

To discover the.

D

Path: MTU, if they can't, they fall back to some reasonable defaults. I think it's 1284 default case and if legacy ipv4 networks are a consideration.

E

To fall back to.

D

576 really necessary issue. 11 covers what sort of security state that dots client can assume on redirection. We currently have some language there. That says the dots client cannot assume that the security state is going to be the same on the redirection target, but it's free to try to do a session presumption using something like you. Do, a session presumption, I think probably of the other coast issues number 12 is the most significant.

D

This is trying to establish the responsibilities of both parties when it comes to mitigation requests and the status feedback you get from the server.

D

This goes into some details as well about the details that you're going to come back, so packets per second dropped on behalf of the client at twice per second total total ice drop. Total package dropped there's some additional things there, but they're less less important. I.

D

Think one of the major concerns that that big Dolson, whose feedback kind of drove a bunch of these issues had was about some of the indistinct language for things like mitigation termination headline. So we've tried to clarify that and the responsibilities of both parties when mitigation is actually considered to be totally terminated.

D

This has an aspect where to try to prevent route or DNS flapping dots client is, it can make a request, terminate mitigation, but the server has a small window to leave the mitigation running. So if the client decides to reenable.

E

D

Rapidly, the dots client, the dot server, is not announcing routes again rapidly. I, don't know if this is actually concerned for people, a sort of an informal survey that I've done with some operators. It seems to be. This is relatively important, but others and also said we can just do Ralph lap dampening so I would be interested in some feedback. If anybody hasn't.

C

Well, our networks, so um yes, I mean you don't in general, it want to have to be changing your network network state constantly on. However, you know from my perspective, and maybe this is how it's been discuss.

C

I, don't know this is not really something that should be built into dots itself necessarily, but it's an implementation, not you know some of some kind of timer and I think we actually talked about that in the 00 use cases draft- and you know when we're talking about the very basic case- um we're not ems, requests, negation and then turnings terminates in mitigation. That term is an optional. You know there could be an optional timer, where the mitigator keep some negation active.

C

So I, don't think that that's necessarily something that should be built into dots itself, but it is certainly something that I would believe an implementer would maybe want to put that knob on here. So.

D

One of the considerations here is we're trying not to encode financial relationships right, but one of the considerations here is to try.

E

D

Out at what point, the dots, clients, responsibility and ownership of the mitigation is actually over while retaining that flexibility that you're talking about so the dots server operators can leave the mitigation for some small period. But at some point the dots client knows that, after that timer expires. It's no longer responsible for the mitigation can just assume that, since it's asking the mitigation determinated, the dot server operators are free to. Let that continue running, maybe they've got.

D

The attack has continued to impact the Oscar server operators network, but not the clients and the dots client wants it over for the responsibility. Yeah.

C

Roll it over to our networks. um So in that scenario yeah we probably engineer this in the meeting, but you know the idea. Is you? Have the dots client makes mitigation request? Dot server accepts that mitigation started mitigation status messages are exchanged, then the dots clients is ok, I'm, going to issue a negation termination request great, and then you know, my assumption would be that the dot server would basically say no and then sighting in litigation is still running and continue to provide status updates. You know until it based on whatever timers are configured.

C

It decides that that will indeed close the negation down yeah.

D

H

D

This could be a configuration thing. I think I want to make sure that the dots, I operators retain the operational control, but I hear what you're saying yeah.

C

Is it control? Control is wrong word because this is a communications, medium rain and, as you said, we're not trying to enforce. You know the contractual relationship which are trying to facilitate the communication. So you know whatever on you know, timers or whatever you know special rules that allow, for you know, a force mitigation termination from a doc's client to a got server. That's almost really almost outside the scope of dots. Oh, my thanks.

D

So I guess the the last one. That's really worthy of discussion here on this. This.

E

D

Required an optional mitigation scope types, so previous revisions had listed some example scope tights, but they were examples. They were not required or optional. So the required scope types now are the obvious ones: ipv4 got a quad ipv4, prefixes, ipv6 addresses and prefixes and fully qualified domain names. The optional type right now there's only one of them so feel free to suggest others is a URI, and we previously also had listed in the examples e 1 64 telephone numbers, but you can actually encode those as a URI.

D

So that's I, think that covers it and you are I covers lots of different schemes as well.

D

So we have I, think eight or nine remaining open issues, one of them that we discussed a bit yesterday during design team meetings is the definition of what a session is.

D

If you look through previous revisions of the signal channel, if I of the requirements draft there's this kind of loose use of the term session more or less interchangeable, with this, the channel itself, so I've just gone through I, have a quarter quest now that replaces session the channel, and that made us revolve both of these things getting Dave's input on this would probably let us know if we're moving the right direction.

D

The issue number six is something I think is probably just a deletion. We already have in front of the security requirements, a requirement from usual off I think this is just left over I'm going to skip 16 for the moment, and then we have some not considerations right now, but they are implicitly client only so we need to have some serving that considerations and whether upnp is something we need to to add to the requirements or after a reference to anyway um so issue. 16 is how overlapping requests should be should be handled.

D

Let's say you have multiple dots clients. They detect the same tack separately. There's no acquittal coordination between them. They both.

I

D

Mitigation for the same attack, how does the dot server resolve that I've become less enamored of the idea that the dot start would be responsible for resolving? That I feel like this might have to be an implementation, specific thing. The dot server turns a mitigator or whatever is responsible for installing filters were doing the putting the appropriate countermeasures in place and, depending on what a mitigator says, is: okay, the dot server either rejects or adds the mitigation. It's about a quote.

C

On azure networks, as far as I know, there's no concept of mitigation scope within Don's. Right again, as you indicate that's something that is really outside I mean outside the boundaries of the DA system. Now you know that the client say: okay, I I'm, requesting negation. You know, for this particular slide, your block right or for this particular you know you are I or what have you, but in the actual configuration of the mitigation scope within the dr orchestration system, that's operated by the des integration service provider is we need?

C

We don't want to get into that, because then we're going to start constricting guys to a particular paradigm of des mitigation that relies on a specific technology path. We don't want to do that. um Could you go back one slide, please or more one where's, the part about the fqdn. It's under the issue phenomenal.

D

Okay, sorry, it.

C

Was in this light, so do you really mean fqdn, or do you mean like the zone cut because fqdn.

D

C

A very specific meaning.

D

It is, it is a very specific meaning and the I initially just had domain names as the initial language.

E

D

Add some input from another one of the editors who suggested.

C

D

With fully qualified domain have so that's, what's there yeah.

C

I think we probably replace fqdn with something else, because I'd always that implies that we're only able to put a an fqdn na dots mitigation request in April client reality. We might want to say, like you know, star dot. You know, example that a tu been something like that.

D

C

Subdomain addressed okay, sorry Nikki.

J

Fleming and raising the overlapping request, handling I, understand why you know you prefer on a service I'm not having to deal with it. I'm, not convinced, that's right answer. My concern is that I don't want to get into a scenario where requests are essentially client, specifics right and a giving client that asks something would have to be the one that gets rid of it as well, and if there are any kind of conflicts right, somehow the clients have to coordinate amongst themselves right, I, I, don't think that's going to be workable, so.

I

I totally agree on.

A

I

J

Soap so I think there's an issue in terms of when you specify a neat mitigation for a certain scope right. How does that get reconciled between multiple clients that are not necessarily coordinated and I? Don't know how you would do that without having the servers and she'd be dealing with that overlapped and you're solving it on the.

D

Back end, so the one thing that that's been in the back of my mind, right is this is might be an appropriate place for the relay where we have that that the aggregation of the signaling and the relay is responsible for doing that. I, don't know that that's any different than just having a serger. Do it I'm figuring out where the overlap is and merging them. What what really did I say? Oh yes, what.

J

D

Gateway, so what I'm saying is no I'm being serious right, because when.

E

I understand yeah.

D

The in one of our in the architecture, we have this diagram right. Where there's this aggregated signal, I'm right at.

I

Multiple downstream.

D

Then the gateway is responsible for a sort of merging into a single channel up to a dot server and that that aggregation sort of implies the merging of these signal channels and the scopes.

D

But that's since a gateway is just a back-to-back server and client I, don't think that's only different.

J

Than having a server do it, let me rephrase my question: okay,.

D

J

You suggesting that we formally define what a gateway is and the responsibilities that it has, because currently we don't open spot with.

D

That one I'm not not right now, yeah I, would agree with them. Oh I.

J

Think we got a chip away at this one. You know I agree.

D

All right, I think, the remaining open issues- none of these are particularly contentious in a way that the overlapping one is. I think there is some ambiguity about what ownership and protected resources are, so we need to tighten the language there and get some consensus.

E

D

And adding a couple requirements for authentication and authorization, this can just be put into the security requirements section there's a lot of implicit understanding of what authentication and authorization is required. We just need to make that explicit and finally did brought this up before the last in your meeting about the need.

B

For replay protection, my.

D

Sense is that we need to be very careful here, because any ability to replay might have an impact on the server operators network. So we should be. We should have full of retha protections in place.

D

There aren't too many other upcoming changes. I think the the one notable one is that the operational requirements section has really evolved into the signal channel requirements. I believe there's one remaining operational requirement in there that we can just go to the general requirements, that's relating to the not finding stuff and then the remaining requirements under the current operational requirements. Section will just be retitled signal, channel requirements, any corporation of any minor, Corrections, typos and so on.

D

Please submit new issues as soon as you can. We are working toward working group. Last call, as I mentioned at the beginning of this is the queue of issues. Open issues are the obstacle. That is the obstacle to work. Your glass call, so this.

G

We know we can get.

D

Through that, the sooner we get to upper he professed call thanks.

G

I economy, car without to the heartbeat discussion, do you think there could be a zero heartbeat mode? That'd be difficult, there could be your fabulous of any caste. Equal game of Tata Tata ah could be easy, could exist that so.

D

In the architecture draft we have the anycast consideration section where I think. Maybe we should make this clear. We talked mostly about using it as a method for service discovery. Right. It's a way to this. You use a any cast at service address and then you get directed to the unicast server that you should be using I.

K

D

Think I think the yeah.

G

Okay papi: do you think that uh math regular, I said how to eat, though yeah.

D

Right in that case, with any cashier, you can send a number change in there.

G

D

E

D

Discuss that offline, I'm not sure that I have a good answer for you right off the top of my head. Okay,.

L

Amite eek, I'm just a economies, point they're only a zero heartbeat idea. I think that might have value. I mean that's. The odd implement eight implement potential implementer go spoken to who wants to spin things up on an ad hoc basis. You know because they don't want to be maintained, multiple sessions from multiple customers potentially tomorrow, and so that that could have some value their initiative with that could be handled, maybe via the data channel. I guess which could be just navigating across sensational, but if the signal channel had a zero, hardly function right.

D

And so then they figure just set it up, sells out of the channel user palace or event ID yeah.

L

Buying such an exit area down do.

D

The same thing reason: the session use get ego sessions. Yes, okay,.

B

Okay, other questions and comments.

B

Okay, in that case, we'll move on to the next draft, which is the architecture and andrew. You still have the floor.

D

Alright, so this is, this is going to be quite brief and we have a lot of other things to discuss a while keep us moving like the dots requirements. We are tracking open issues long. If you could have issue tracker under the dots WG group, we only have a couple remaining issues right now. The document itself is kind of paused, while we're waiting on changes to the use cases and requirements.

D

I think the the recent movement on both of those graphs is is getting us into position where we can make the tweaks that are necessary and we can get ourselves working group last call soon.

D

The issues that we have closed recently do cover things like service discovery, anycast and the the need for redirection in the dots of architecture. These have been closed for quite a while, but and all of those changes are in the existing drafts. So if you have issues or concerns about the the current discussion of any of these topics, please take another look and bring it up on the work of your mailing list of or simply open an issue.

D

We only have a couple remaining issues, as I mentioned in the question for the use cases. Multihomed dots clients are still something that we need to sort out. I would like to talk more with the use cases offers on this. So we can try to coordinate, come to something that's mutually agreeable and then finally, things like privacy and visibility for a Christmas signaling. This has been a concern about the. If you need to have let's say, a mitigator wants TP I into encrypted traffic.

D

Are there concerns about sharing private keys and, if you're, using this sort of backstock recursive signaling, but.

K

That really covers it. On.

D

The architecture draft right now largely it's going to be some coordination with the requirements and use cases drafts and then, if we can knock out the remaining two issues. I think we're ready for working group last call again submit any issues you have with the current draft as soon as you can so we can keep track of what actually is remaining before last call I.

E

Kaplan-Meier DAV um sorry, I, hadn't, read that part and just hearing you say: DPI mandame, el traffic. That's gonna hit a bleep storm yeah.

D

E

I'm not advocating.

D

For it, it's just a concern about do we need to mention the perfect concern about privacy and security if somebody is advocating for this in the architecture draft we're not necessarily advocating for the sharing of these might accuse.

E

Why, wouldn't you just have a requirement that are that's it I'm, sorry, I think a requirement that prohibits it.

D

We want to impose that on somebody that you can.

E

What are you wrapping this in? What I'm, using TLS.

D

That's that's likely yeah, so.

E

When you had a 1.3, that's already done that.

D

The there's the other static, if you hum and stuff that's going on as well right so I mean there's yes, I agree with you.

E

Yeah, not an ietf um you're, not in a working group right, not in a working group. um It might go through the independent stream. Editor I haven't agreed to 80, sponsor it and I'll meet Becker. Would okay, so we'll have to see what happens but.

D

E

Yeah, I don't know, I think it's.

K

I'm proud, no, I think getting.

D

For this right, this was an issue brought up by one of the other editors and we've been trying to figure out. Is this something we need to treat the architecture draft all.

E

Right I'll well when, at least by the time against to my review, I'll, look at it and see if it needs to be in the strap that all, um if you're doing a negative statement against it, that would probably be fine, maybe dropping it completely, might be your best bet. Yeah.

D

Okay, thank you.

C

Roland ovens are renette works, a couple of the issues that are still outstanding for architecture as well as for requirements. They require some fairly dense discussions. That might I mean we want to keep things on the list as much as possible, but that doesn't preclude having interactive discussions and then summarizing providing the list. So maybe we can think about having a actual meaning.

C

You know a con call for interested parties to discuss these issues sometime relatively quickly, because I think with some interactive discussions we can kind of go through and not some more things down more quickly than we could trying to trade messages on the list. I.

D

Agree I mean I would even settle for any discussion on the mailing list at this point on those issues with.

B

Their be interest for us to do virtual item around virtual number around this or.

D

F

Just one question a timeline for this: when do you need it? Next week, two weeks, two weeks from you, ok, so.

G

F

You come to us and tell us your gonna be doing ok. Thank you.

B

Other things related to the architecture, documents, comments, concerns feedback.

B

All right, thank you, and so next up on the agenda, we're gonna pivot to talk about different protocol and solution-oriented drafts. First up is draft ready, dots data channel that Nick is going to be talking about, and for those that did not see the announcement on man was. This is the draft we're talking about to pull for adoption.

L

Hey I'm nikki, I'm here to present on this one Terry could make it, and it's really this is. This is really more of a recap of the discussions we've had over the last couple of months and also the bits and pieces from the interim meeting.

L

The main reason why I want to recap: it is because, after the last meeting, we had a bit number discussions between myself and root area and a couple of others, and we decided that the as far as the data trial goes, the a lot of the efforts were converging anyway, so we decided to just converge onto this one draft.

L

So to that end, the data challenge using rest conf. There were young models in there, they're cooler, yang, they're funky, the updated draft meets all the requirements, as as it has been, have been pretty much in line this draft being pretty much in line with the requirements all the way, along still sitting on top of the stack, that's very, very familiar at the restaurant TLS tcp/ip with dot sitting on top of it, um the rest Kong for suggested, and if you seem to make some sense, it's obviously a subset of net.

L

Can't it's not a heavy use net com. So we feel that might provide less overhead as far as like implementation goes, but then kind of in it allows us to basically use some common app commonality across other efforts were.

L

It addresses the needs of what we need as far as like configuration so on. You've got the general kind of what you would expect post put delete, guess to retrieve config data and non config. There was a tea lights. Are people liveliness? It's generally speaking, the data channels reasonably connection-oriented anyway, so it will be maintaining that it goes along.

L

We've got a number of young models in there already and then we've got the people it exists and the convert as well as we go a couple of our options.

L

A couple of examples as identify a model filter model you identify one is it was discussed earlier regarding things like fqdn and this question on that it'd be a the idea is to basically assign an alias at the time with the actual mitigation that, instead of having to kind of define every single individual element, you just say Oh use, alias one and that preferences IP address this work this service of this. So it kind of ties in a few of those together and then.

E

It can be expanded as.

L

We go along, maybe we've got fqdn I, you are IE, wants it or in there and stuff, so they can just as we as we wish and as we as we see fit weekly pad that out and that's pretty.

M

L

Mean it's just it. I think we've we've all come along same path of them, the type of things you want in this particular draft and how it's been going in if we've all been kind of getting there. So that's me done any questions. Comments.

G

B

So just agenda wise. We built some time here to talk about. Are we ready for adoption of this as a working group kind of artifact we're not going to decide? Today we have a poll out on the mailing list because we travel and everything there's two more weeks to provide your your thinking on this after ietf finishes. But again we have time in the agenda. If you like to share your feelings endorsing this or you have some concerns and you'd like them voiced.

J

Fleming andreasen I'm supportive, as I said on the mailing list, I think makes a lot of sense. I think the direction that is going as well makes a lot of sense, which to me is more of a separation of theta channel NBC. Hong satellite is really is operational, stuff I mean configuration related data, and a signal channel can then leverage that data later on so I like it.

L

By the way, it's not to say that they're on any it's that it's still not being worked on them. Dave race and some comments before regarding the eye again comments around the idea session, but also the comments around hit, the ideas of cuddling and so on that we need to look at.

B

Absolutely nikka me just to clarify adopting just as a working group document is not accept and, and it all suggests, that what is in that draft is is concrete. It is completely valuable. It's just that we're accepting the existing text as a working for artifact I.

N

Mascis HD consulting I'm all for this being a working group document. It keeps one aspect simple, so we can keep our focus on the the really driving part, which is the signal channel.

F

Oop, okay, thank you, then, that moves us to the next part I'm.

L

Unhook this one as well, yes,.

F

The signal channel, which is more entry more challenging than any of the swings.

L

Okay, so um one of the reasons why I want to present this one in the absence of thiru is because I wanted to get some color around. Why? Why would sort of converging around this one, the shell even open? So there was a protocol document the week before, which sported protobuf zamora, beaconing type of singling. There was this one which was using co-opting, see bore, and there was some other variations on there again.

L

What we had a discussion after the last meeting and we we came to the conclusion that we would like to again try and converge on a single path, just to reduce the amount of duplication and so on. To that end, we've kind of decided that, in the absence of being able to kind of standardized on protocols at this particular time, then what we thought we could do is at a later date may be used as a content type.

L

We were using credit buffs in our original draft as an IDL and leveraging that in the new, a draft that down in new revision, satiric put together and stuff, it moved more towards that some boards, a similar apps ybor. Having a similar aspects as its protocol, things like the mapping and so on, the element mapping be in addition to that also having discussions with Fleming and so on.

L

What we saw was we didn't really want to go through the rigmarole of reinventing tcp or UDP and congestion control, and all these type of day to the coop we can handle that for a soak.

L

So with this one, the highlights is again: we have the young in there it's a mitigation request model configuration model and so on, including rules defined in the core young cyborg, and it compiles nicely see poor again is being used or prominent of signal map to see more types, we're using particular types and nothings to ensure that we can just again make it a bit more efficient, rather than kind of complete that constantly reiterating through the schemer.

L

Every time you send a message which was a big big issue: I had kind of between Jason and coding versus, say something like protocols, the mitigation response and the requests are now marked as non confirmable. The non confirm will requests ascent at regular intervals until response received. That was something else that we kind of were that was that was the way we thought needs shaping in this one and two good job getting it in there. Channel configuration star flag does confirm.

L

Also, we have sort of have the best of both worlds and we have printable messages when we need them non confirmable than we don't and we go from there. What we've decided to do I think as well as tranki, so that we have static types. We kind of tried to remove some the ambiguity of 0 in peacetime. We could have confirmable, you know, and so on. So we're bhutan is like these ones will always be at this particular characteristic. These of 14 bills of that characteristics.

L

There's no ambiguity in figuring out what you know, kind of various aspects of what peacetime is versus. You know what you know. What time is the heartbeat mat mechanism is using the car ping we're looking. Ttl is 1.3 their sport, never event, specific parameters which we feel is important and for extensibility and so on. Added new mutations, those parameters which are somewhat useful for the endpoint to be able to figure out. What's going on, it's like a very basic basic level of telemetry for the the originator to be able to figure out.

L

You know what, as far as like, if the attack on going as you tackle on going, what it's saying, it can then try a degree of status. Yes Larry, so.

D

With regard to sorry, Andrew Mortensen, the TLS versions, do you think that we are going to be blocked waiting for detail s 1 point 3 this morning, I mean it's. There are no competing DTLS drafts right, but it's still individual draft yeah I'm not asking you to choose, but none we we need to decide if we're waiting on it. Yeah.

L

Overall detail: s has been a bit of it. Well, the dtls has been a bit of a pain in this, um not so much because DC less itself, but the available libraries out there are painful, say the least, to try and get to do what we wanted to do. There's either you know they're, just either lacking or they're just so hard to work with this.

L

Isn't that so, but that's pretty much it so deal with this weekend, which we're trying to move towards this to consult wars, consolidation I'm, not sure whether we're completely ready to to say that this one's ready for adoption yet, but that doesn't mean we couldn't ask the question again.

O

L

Or questions welcome.

F

Maybe it's a chair, quick question: why do you say you think it's not ready yet.

L

My personal opinion is, we just need to refine it a bit more and that's, maybe just being a bit precious. It's it's more. It's probably it's I'll! Guess I! Guess it's more cosmetic than anything else. My opinion is to that. But my opinion is that I just feel we need to kind of like maybe just refine it and just have it flow.

L

A little better and I think that's something that's actually been working to in the most in the most recent variation, most recent version and I think it'll be out in the next version as well and I think we're very close to it. But I just think. There's a few tweaks and everything that just you know just before we go with it. Having said that, like I said, let's not say we could master coach.

J

Slamming andrea's, so I've ever support the notion of trying to converge on a single value. Yeah I, think I mean what this group needs to produce right. Our protocol specifications in the end um well, one of each one for today, I, don't yes, you're right, not multiple of them and the sooner we can get to that point like the better. In terms of you know, where do we focus our energy right? I mean we had the information model as well. The.

E

Reason that we have that.

J

My friend is because we have had competing protocol suggestions in there, like the data model is in here already like same thing for the data channel, so if we could find a way of converging on a single protocol of a signal, channel I think that would be very helpful to the group overall. So I would certainly support it from that point of you cook, and you know, I think it.

L

Was reasonable right as a starting point? Oh yeah, I mean I mean what's in there at the minute. I don't really have any issues with the content for this. We just need to it's just flowing likes. It cause my dear.

J

I mean a plenty of details, still right, but you don't get. You want to change. Yes,.

N

Not my skirts I do have two concerns from what you said: Nick, actually, one that the am Angie said: TLS dtls 1.3 blocking. What's.

C

The consequences.

N

Of that that no, what no, what is it that why do we want 1.3 and thus the blocking we may need to call that out. The other thing is the comment you mentioned about. Detail are actual detail, ice, implementations and issues regard to that is, so is their experience, which is showing us something that week that will be need to be concerned with or any gave, maybe some implementation recommendations which we may need. We need to stay say in the as as informational information, but something they had to put in the document.

N

With regard to actual experience, no.

L

I think those are both great comments and, as far as our invitation goes implementation, attempts of gun, I think what we found and where we have heartaches is because we've been trying to get this working for a little while and now very in our various efforts, and it's like just trying to get the one. That's you know the ticks, all the boxes that we feel we need.

L

Having said that, I think if we, if we do start to consolidate on single draught and we're starting to kind of say well now we can start to talk about the stuff. Then we can start to them dig deeper figure out. What is what we need? What's lacking, what.

E

L

Tolerate you know even yes and.

E

L

And then we can figure out well, is it is this a blocker, or can we go do something else you know is? It is the for example, I think Andrew used the tiniest cell, which had an implementation that might not be great, but would it be sufficient? You.

E

L

There's those questions to come up, but, yes, I, think your comments or well noted. Thank you.

D

I Andrew ones and I think I'm kind of responsible for the question about whether we should adopt now, but that was on revision. 09 revision, 10 has incorporated a bunch of stuff and I would be okay. Talking about adoption in spike.

G

F

That is indeed than a question to ask.

F

Maybe before I ask that we should ask who has read documents: okay, so okay, who has read draft ready, dot, signal channel version, eight, nine or ten, because that makes it cos, hid it okay. So that's a couple of roughly 10 12 keep your hands roughly okay, just okay! That I think that's a sizable amount of people so um master so actually.

A

Show me off now: okay,.

F

Let's be, let's be, let's be aggressive and that's awesome so who would be I'm gonna ask the question who would be comfortable, know who's in favor of adding adopting? This is a working draft. The answer would be yes, no or don't know so who's in favor, yes, crazy or shelby hand hum. I think this hummus, probably more adequate, sorry soham now or the ad has something to say cafe.

B

Get new book yeah, whatever your location, oh.

K

F

If you're in favor of adopting hum now, please.

F

Okay, who is against that I'm? Now, please, okay, I think that is reasonably good abstains. You can hum ok, I think that is quite good consensus to adopt. Thank you, cool.

B

Okay, so this brings us to to an exciting part of the agenda from a working groups perspective to actually talk about implementations of some of these protocols. So we have two up, so economy is first talking about the entity, implementation.

G

Hi I'm thomas newcomen, EDT communication. They I'd like to show you the demesne tation experience behind in the field short presentation.

G

Okay, now we are developing those diet, top of rare fish, a domestication from the these drugs, which.

G

With us, nah I just hate working well, although if he is not free romantic, it will be open sourced with psv, see clouds lightning. It is re implementation of a very able to invent through those protocol specification he accordance with the curing of da stressful fair, and we can show you a table to any time. So please contact us I'm available after this meeting us any.

B

G

The rest of ikea meeting she's.

B

Makin, a Meccano, stop you, which versions of those drafts did you implement? Okay,.

G

It is not written very, are pitiful this date, each other in the video lady trust. Our signature, honor is based on part of city hey. Thank you.

G

Okay, this figure shows the outline of our application. The core of application is the thoughts communications channel between those client and server the thoughts Gaia. Since those message is not service in court over de ellas as dots sugar, the dot server receives the signal, and then he validates it to forget a message format as the content itself of the posterior and then determine whether the students should be handed over to further process. Then adult sabah start mediation by handing over the signal too lovable broadcast a mitigator.

G

The chose go: bdp are the first, the plague of locusts, which enables our deviate law, stick or Die version in any digital network, so we will expand a capable in Tahiti to utilize other devices which really over ssh or other protocols. Finally,.

F

G

Third, server will be able to send out message to other dots enabled devices as adults gateway functionality.

G

The software is implemented in colon, Adam implementation experience. I can assure you that the cost education experience is venture to be implemented. However, finding good libraries such as God restaurant, especially DTLS, is rather yada, because we are trying federal, detest library. However, we haven't found any good. He can get get it library. We work with the specification of God cigna father.

G

D

It yeah entrance regarding the details. Libraries are you if there is no native golang, dtls library right now, right, there's some that implements the go lang dtls, which is rapping openssl, as I corrected that what you're using oh.

G

Yeah yeah no Mickey Cola like I, so I you see that I am calling it okay.

D

Yeah yeah yeah yeah did the one that I have been trying to use is a tiny detail, s which implements the sort of their requirements for co-op. I. Don't know if you take a look at that, but it's it is the best electron, hey.

G

I use the loose if it's here, how about which is only using a gated 1.2.

G

Okay, I'd like to show you two.

B

G

Family, if you need to experience, but one is about the coupling of data channel on the student other, the student center of state deviated. What year is this land package gates must be used for mutual authentication under the boat summer, couples that dot signal on the data channel system using dust, those trans identities, so two drafts require that same common name for playa cited in DTLA of signal channel on the theory of each other. I think this is strong restrictions for implementation and deployment.

G

So, if Todd is going to use other beauty, authenticity Brody's, it needs generalized conflict of C energy. We are we I mothering for the customer organization. Education, ID tip actually included.

M

G

Data makes it okay.

M

Oh hi Jeff Hodges. If you don't need to really be using the seein as the subject name, you should probably be looking at using the subject: alternative name extension and mapping to DNS in that.

G

M

Need to use a distinguish a next 500, distinguished name in in the subject you can put whatever you want and there, but you really should be leveraging. Subject: alternative name, okay,.

J

Fleming andreasen I agree completely with your observation. This is you know what we were alluding to I think before, as well with your authentication and authorization model and also the separation of the signal I'm data channel. The way it is right now is not right. I mean the signal channel and data channel I independent right. It's the data, that's being conveyed over them that you have to reconcile in the back end and we have to have an occasion or authorization model that works for that I. Think right now it says it's based on the dots.

J

You know, client, ID or something like that. I agree, I think, is too narrow. I think that's your fear, more flexible with me mad. So what extent we need to specify that versus that becomes. You know a back-end operator issue, I'm, not sure.

G

Yeah but separation updates turn around a single time. I, don't pick on the issues, so I gratitude change the point. Ok, infinity experience. We thought we realized that the Italian Air Force Hospital on see that the only post going to because it is used to sometimes so.

G

They are each so the they are separated. Ok, totally obscure channel are separating the thoughts cry out for this channel can be placed in different location. There will be less probability of suffering from attacks than the place optional channel only devices, so in order to cover those channels from combat grant or then this world and 450 ed hardy.

G

However, if that is really beautiful, I'd like to add some use case to our use extra typing at.

J

Least one by me and raise nyan absolutely. I think there is such a use case good.

J

L

Sorry, just a just 2, plus 1 Fleming and I graders in this case, I can see many occasions when that might be useful, for whatever reason I need to have I don't have separation in the domain or to have clients are again like we said before either single signal only or Dave channel or combined, but spread out yeah yeah.

L

B

You is that that's not captured right now in the current text. I'm confirming I.

L

Don't think it's captured in that I, don't think some text, yeah I, think it I think you could imply it, but so she touch you I.

G

Think you can find out by I I.

F

Apologize just to make it explicit, so it's not in the use case draft clearly enough, then that means it needs to be clarified there yeah. So.

L

Thank you, yeah.

F

Sorry for interrupting I.

D

Enjoy mortensen I was wondering if you had any if you've done anything toward client provisioning at all, my bootstrapping or, if that's still coming down the road.

G

Although yeah their client.

D

Provisioning yeah, so so getting the keying material. Ok,.

G

Now I'm death, using that I am software page only reading that typifies fighting about that equate of the mitigation who hey.

B

N

G

N

Bomb, a squid some this last slide, where you had the one data channel with multiple armed signal channels, pretty much flies in the face of your previous comment about how to connect the two authentications together to pair them up, because you basically have almost like a reverse multicast sort of a situation here or how they didn't ease are going to done. This is going to take some real serious of thought about how to do the the identity control such that you have what linkage you want.

N

It's an interesting use case, but it's going to take some real, serious arm, thought of how to do it and feel uncomfortable with it. Hey.

G

We can live our party.

P

Jefferson organisms, you mention I thinking the first or second slide- that you're open source the code right. So my question is you have a timeline for that or whenever you play into the never the code to release the code.

G

Okay, uh now I'm trying to talk to our raviga Department, oh I, don't know world how I don't really date so I cannot promise you. However, it will be between these a big meeting on the next time.

J

Clementa isn't just responding to Bob's comments before I; actually, don't see them at us with each other at all. I think they're very well aligned to use case in areas that get there. I think it's very common I, it's not the channels that are being bound together right. It's the data that they're operating on and controlling access to that data that needs to be sorted out. I, don't see anything at all. Consider.

G

Okay, I love rifle of the last word from me that I read at least that we made a heartbeat, but it is not even eight yet so I will get hot hot with our dogs. You know Charlie are falling through the operatives with each other drop, and then I will make a father feedback to this body. Would.

B

G

F

Thank you comment as a chair. This is awesome, so I've been long hoping to see the first implementations being presented, and this is really very, very good. Thank you very much very yeah.

B

Thank you Kaname. Thank you, administrative liebe, for you to the next presentation. Where is the blue sheet? Who needs the blue sheet?

B

Okay, if we could just pass the blue sheets up to whoever needs the blue sheet. Thank you, Fleming Europe. If you could tell us about your implementation,.

J

So uh mm-hmm I can tell you about series implementation. It's uh it's not mine. I will do my best. You know.

J

All right, so, if you recall the presentation that we did at the last I ETF, there was actually a couple of slides in there about the proof of concept that was done back then, and it's pretty similar actually in terms of the information that's in here. The major difference is that the proof of concept implementation has been updated to reflect the latest draft, so in terms of the overall framework here californium, especially on the server side. That's the same, as for the last time you go online, of course, and not download yourself.

J

If you want to play around with it, there are various api's in here. Isaac was mentioning right. There is these related to congestion control, our than I important message, transmission parameters, heartbeat timeouts, which we specifying as part of your ball toss protocol operation and then the new part here about messages being confirmable or non confirmable, also that's being leveraged. Well now, in the end of a concept that's here has done.

J

The client side is browser-based, so there is a plug-in, copper again. There's a URL here where you can go fits that that will enable you to essentially come up with your own messages, and then you can send those to the server and you can play around with those, and that's a sense of what's been done. Here is an example of those again. You see. We have an example here, non confirmable message, seaboards what we're using- and this is another there- is a way of actually decoding some of these messages back and forth.

J

If you wanted to see, there's a URL for that as well, so this is just basically illustrating you an example message here how it's been actually encoded and what's be sent across, so that is it for the proof of concept from terror. Any questions or comments on that.

B

Okay, thank you for making. So we just had two implementations. Are there? Is there a third one lurking that anyone wants to bring up or aspiration for implementation.

Q

Barry green for akamai I think um in hallway conversations leading up to this and and I did last week, a read-through of all our docs from the point of view of implement as Russ white and I were talking. We don't want this to go down the path of cider, alright, so I think we need to instigate implementations all right, so I think we're in a phase where it's not just asking who's implementing I. Think it's dumb. There are there's um efforts where you can have several companies get the governesses.

Q

Ok, we're going to instigate we're going to pay some University go. Do it or you can you got governments who throw money at projects like this? You know it's I, think um now's the time to start. You know in several people you know involve of the drafts or say, like. Oh we're not ready. Yet this is okay, no now's the time to start to instigate the implementation, our roadmap of deployment right because we will we want to make sure that we come to fruition. For this work we got to get our roadmap of deployment.

Q

I mean to get a the co. We have a prototype or the ballot make sure it's done. You start that now get the funda now right, like I, don't know you know on the you know like, for instance of you guys see see if you know right right. If, if we don't we don't, if we don't have vendors starting ACC process, it concept commit process.

Q

These are the sort of things we need to start getting going with it, because, as you start it now, you know we get to more I ETS. You know we're making some good progress on the specifications. Things will get closed out and as we get more people using the docks, you know what we'll find and discover things in tune the work.

Q

So if there's I think if there's anybody interested in this I think we can get a side, little instigation, sort of of discussions going on and start rolling this out and then in the next idea report out for the working group is to have a sections of saying. Okay, what's the result of our efforts of getting things moving on more implementation, operational deployment by two cents.

B

Thank you completely agree. Any other vendors researchers want to speak up about their plant implementations.

B

Yeah it so two things that your bison I've been talked about it our first now that we're talking about individual specifications. Do we want to start doing this as part of the hackathon that's held saturday before ITF, perhaps in Prague we're Singapore, and given that we have two implementations at one point, we start talking about interop between them, so things on our minds if you're interested, please do let us know okay. So the last item on our agenda is not directly related to some of the specific architectures we've been kind of talking about.

B

So next up is talking about IP fix usage, so Marvin.

R

All right, this is Marvin, don't play from Huawei and I'm gonna. Give you a brief update on the interrupt. Ok, this is countin. I would go through the background. The changes from the last iteration ideas issues in this step, the main main purpose of these presentations- to do some clarifications. So the background as you'll see. Why would I be fixed draft appear in adults?

R

Looking group, the first thing is: um actually our draft targus same problem space that we want to solve that dots attacks, but the approaches we approaches we are taken by taking are given differently, as you have learned that dots is more about signaling building the signaling mechanism, but our job is more about building a data analysis platform, so so.

E

That's the difference and.

R

So this is yes, as the working group chairs just this is a little bit of the the scope of dots. So this the reason why we submit this revised. The draft is because it's a little bit about the history. The original draft was here, so we want to present the new a PT job here. Maybe we can hear some opinions from the floor about security issues, alright, so the nest like ah what we propose in this trap is actually a collection of IP fix information elements, as you can see on the left best.

R

The original setup I've been fixed elements in the last iteration, so it is in the new trap that we have armed with in the IP fixed elements, and we have actually removed a lot of IP fix Edmonds, because the removed I be fixed elements, they can be implemented using the existing animals through some combination or complex structures. So this is a maybe a lesson learned for you that if you want to add some new IP fix information elements, you should really look at the unique 16 ones.

R

Alright and so what's left in the new trap, is it's all about TCP connection tracking? That's a the concept would like to promote so, as I said before, we like to butte a data analysis platform. So uh so we want to collect a certain information from TCP connections, and it is is by a by using big data technologies. I am using this password because, because the information we collected from the typical anxious, don't have to be very accurate. Actually we just want to monitor the changes that behavioral changes of those characteristics in those TSP connections.

R

E

R

Ip fix because we need some mechanism to export those information from kiss p, TCB traffics and, as you know, that I Vivek see small standard and is widely bought it. So we just need to add some new information elements. So this is the main idea. Okay, if you look at the diagram, what slack what's what's lacking the idx information elements is the sum of the some of the information related to TCP that set up all the timing: information, for example, the timing information that is related to this participe three-way handshake, which is used to.

E

R

A TCP connection, some attacks may our result in a very abnormal that changes in those times, so we believe collecting of clicking these kind of information is and processes it in arm in a statistical way can help us detect anomaly in TCP traffic's. What so? What do you mean by TCP connection tracking, as you know, that TCP has certain? Secondly, in flex, so we want to have some flax set for all those that is basically in flex.

R

If we observe a sink, then we observed a sink ACK, that's normal, but we said that flex only if those blacks occur in the connection in a normal sequence. If it's not in the normal sequence, then we will not set those information. We will not not set those flags. So by clicking this information we were. It would be easier for us to tell apart what the TCP traffic is doing all right.

R

We also want to incorporate some of the informations about the the packet and the time intervals between different TCP segments, because that might be armed exploited by the attackers. So this these are a basic set of information, would like to collect around his feet, connections and nest, which will realize there are certain issues regarding our solution. The first one is a sim message.

R

Asymmetric traffic, as we know, their internet is more and more is becoming more and more asymmetric, so this will limit the applicability of our solution because we need to collect the information from a conversion point that can observe both the uplink and downlink TCP traffic.

R

They also, they also some concerns about the performance. Yes,.

P

R

If the traffic is really a stream, the bottom of the traffic is really a large that might require a super power device deployed on the conversion point to collect such kind of information. So these are major two issues that we we are now are facing and basically so, as I said, this is just some clarification from the front for this working group so I. We appreciate that way to some discussions in the email list and we appreciate the old comments you have given to us.

R

Thank you very much, and so the next step we are we're going to take, might be when I submit our IP fix information elements directory to iono the ANA, but the reason we are doing another reason we're doing this presentation is because we want to solicit comments for you as much as possible. So thank you adding questions or comments.

R

R

B

Q

So what I would suggest, because this? It's ok, that the Vettes to having users as of that mechanism as a working group, but this would be an individual submission document correct it.

R

Currently is non in.

Q

The working group right so so, then you would do an individual submission documents submitted to I Anna diana is going to have a question who's going to be the technical reviewers right.

E

Q

The question to the working group is we'd, be appropriate technical reviews sorted to.

B

Clarify the IP fix registry is expert review expert review. It's expert review right specification right.

F

And- and there is a list of names for expert visual voice for IP fixes sensors, so I.

Q

Think you just submit and see what they start to getting the feedback from them directly. Alright, thanks.

H

Yeah Julie egli you're just going to submit it in there gonna say yay or nay and it'll be super simple though it should be. No big deal right.

E

A

B

We use the mic if we want to follow up on that.

E

Kathleen Moriarty Adia I'll, follow up with Joel I, had some thoughts while you're presenting as well and thinking you know see if the what the chairs had suggestin seeing at the the I can fix doctors, they were okay with it another thing and just looking at the what you're proposing it looks like a lot of the types of things we see that come from the opt area in terms of the types of measurements, and you know when we look at scree considerations for them, we just couch it and in frame it down to its minimally necessary and and try to do the best we can to protect the data.

E

So there's ways to look at it and there's probably a way to get the Sun either here or an OP.

H

Yeah I mean I, think those are Jolie egli again, I think those are surmountable obstacles right, I mean when you're dealing with proprietary data in a management system. Obviously you have to protect it in certain ways, and you know in security, sensitive environments that requires that you put bumpers around the description in your document, but that is actually the normal operational paradigm for this kind of information, because you wouldn't wouldn't want to leak it even if it didn't have this stuff on there.

B

All right, thank you for that perfect. Thank you. So.

H

B

So on the agenda, we actually have time for open mic. So if anyone has any thoughts, comments or concerns to share with the working group, please please in the queue so.

D

With regard to DTLS and the version whether we're going to be blocked on 13, if you look at the individual submission of the dtls 13 draft right now, it says that DTLS, one two and three are interoperable and are operable. You can clients as long as support both it's fine I think we just proceed with details 12 for now and when detail s13 is available, we can, we can deal with it.

F

To be as Cano muah be a non working group chahat, yes, big plus one go for detail s, 1 point 2.

O

I'm a I'm robi hamilton I'm with chemical abstracts can't see, but it actually is remember so I have no idea for sure what mitigation means here and if there's of the implementations that I've seen so far, is there anything specific in those.

O

As far as mitigation goes reason, I'm asking is were a DDoS attack target on a regular basis now and every IP address that tries to connect to us is different, and so, as far as I can tell, there could be 500 billion people trying to connect us so I, don't know necessarily even how to tell other than the fact that suddenly we're out of sockets on a server. So anyway, that's question.

B

Angel roll under you answering this question that question.

D

Yeah Andrew Mortensen over networks, so mitigation is we are attempting to define what mitigation means in the requirements draft there's some terminology there. So please take a look at that specific definition and see if it covers what you're talking about I, don't think the specific form of mitigation, whether using something in a firewall or a DDoS mitigation device or just trying to put out flows back rolls to block, is within the scope of dots itself.

C

Roland Evans or networks. So if you look at previous iterations in the use cases draft, there are some examples that do talk about some mitigation technologies. But the purpose of one of the purposes of dots is the big technology independent in terms of the negation technology, so we're not locked into a given paradigm rate on some of the you heard earlier in a meeting we're talking about some, the initial proposed use, cases will be coming coming out and those feature different technologies and so forth to provide some I guess: color commentary great the help.

L

Hey just a or color to the discussion as far as the way I see it, I think the simplest way it power-crazed andrew in the pastas mean the world idea of dots is to basically say help it hurts here and then, as far as the mitigation goes, that could be anything I mean it could be unplug it. You know it could be anything at the end. Could too is in scope as far as well I could be, but that we don't want to prescribe that, because that people's ideas are different.

L

You know, as a people's definitions are different and it's that's just a whole world of pain. Okay,.

Q

So I think we're skirting around the questions very green outcome: I um because you're talking about you're in serious pain right now, so here's my suggestion, because the work that we're going on here in ittf as a you decide. This is two years out. This is a lot of we're trying to make ourselves better. But what you can do is I'll point you to a little guy.

Q

I get people say: here's how you can bring your vendors in to figure out what is right for you, but you add one new question all right based off of this with implementations, every vendor, you pull in to say: how can you help me Yusei? What's your deployment planning your roadmap for dots all right? These are asking: what's your dutch rome at, then you pull all like the vendors in here: pull annoyances.

Q

What's your dots rome at because you're in pain now, and so the offline I'll help you out in that part, but at the question, what do you got? Is it dots and they'll? Look at you go like what's dots. This is don't get a clue get on board with the future.

N

Thank you for the question, because this is an excellent example about how people read things into our documents or look for things, and we need to ensure that the architecture doc, which is over people, go to. First all.

N

There's a dots architecture clearly calls out how know where mitigation fits into this and that week, I believe we do we're mainly to strengthen just a little bit more to clearly point out that this is how we signal for call for help mitigation is over there, so that that so people coming to hear saying, I'm in pain, will know that they have to go over there for that pain.

N

Addressing for right now that that's all I've been through many of these before, where the documents did not clearly call out what's in scopus out of scope and where else to look.

B

Yeah I would plus one that not comment that nabob just made fresh eyes on the document are wonderful and if you don't feel like you see yourself reflected in the use case, please let us know, because we want to update the text to do that.

B

Okay, other topics to bring up to the mic.

F

B

So catching actions from the discussion, so we agree that we're going to accept the decidua as a working group document. So if the authors can resubmit that as a working group document will will will get that approved, I would also remind everyone we're still pulling for adoption on the data channel documents, since we started that before the working group. So again, if you have feelings either way, please do drop that on the mailing list. There's time and then with the time we have remaining, not that I'll. Take that long.

B

Just wanted to review the proposed milestones and I would reiterate what you're about to see here. Only Tobias and I have seen and we'd like to discuss these with you, so we of course, are behind on everything we have the use case, requirements and architecture documents, as a straw, man to say working group last call after prague in September. What is the reaction of the working group to that and all the authors on this draft I.

D

Enter Mortensen speaking for myself, and not necessarily for my co-authors I think we. This is even conservative for the requirements in architecture, so.

F

You mean you could do earlier cool. What should we write there?

F

Sorry for being so specific Saturday?

F

Ok, ok, so meaning to its two weeks or something.

D

Sorry, anyone saying not to be clip I think probably buy, buy a product. Ok,.

I

J

Fleming andreas, I think you do need to stagger the requirements and the architecture document a little based on the discussion at use cases, while I use case and requirements yeah. Hopefully we can do that a little bit sooner, but you need a little space for the architecture to reflect whatever changes go in there and also, hopefully reflect that you working with last call comets that come in.

B

So if I internalize what you just said, use case requirements july ish so.

J

August/September.

B

Architecture so.

J

I'm only altered on the architecture document, so you know I'm quite happy with September I'm, not as ambitious as my co-author, but that I will let the authors of the use case and requirements documents for themselves.

I

O

I think the use case should be addressed by two month.

O

B

So what I'm hearing from the the authors is July use case use case in requirements? I'm, sorry, we're not editing. This live, and then architecture document staggered later for September further from the rest of the working group, because Kathleen's not officially listening to any of us yet because will submit this. What's the reaction I mean any comments. Is everyone comfortable with that?

B

Okay and then then it comes to the actual protocols, and so here we got to do a couple of things. So, first our official milestones actually say we're: making a data model document and a transport document. uh My counterproposal to that is. It looks like we're converging towards a data, Channel document and a single channel document. So first I would argue that we want to change the nature of the milestone itself and then tossing out a gauge of I'm sorry of December on both votes.

B

So does anyone have any feelings on first rebranding, what the milestone itself is and then the timing of it um hi.

L

Nicht ich um no I'm, for it. I think that um that's let me mentioned before. As far as data model information little goes, they were probably more necessary when there was a number of dates. I think now there's a degree of conversions. Then it can be just those particular task me just converged into that that function, so yeah.

B

Other reactions from from the working group on those.

B

Okay, well Kathleen we're gonna submit mr. you officially for your review, so you can stare at them just as we described okay, so this actually brings us to the end of the agenda. We have time so if anyone wants to come up to the mic going once going twice: okay! Well, thank you. Everyone for participating, thank.

F

You very much and we will send out the invite for the upcoming virtual interim discussion in a few days and yeah. If I close and.

B

If I could have the blue sheets, whoever has them if you could run them up to the front, who has not signed the blue sheets yet and where are the blue sheets.

F

K