Internet Engineering Task Force 102, 18 Jul 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: IETF102-HOMENET-20180718-1520

Description

HOMENET meeting session at IETF102
2018/07/18 1520

https://datatracker.ietf.org/meeting/102/proceedings/

A

Good afternoon, I guess it's time. So this is the homemade working group. If you don't want to be here, then has a good time. Please I'm, Stephen Fry's, barber stuck.

B

Anybody need an explanation of this slide. Okay, so we've already done some bashing: we're actually going to move Ted to number four and meet in number three, so Ted, because Ted's put together some slides on. Where do we go from here, and so we figure that should come after my status report on security topic so and it's really quick things, but any other agenda bashing. Anybody wants to do.

B

All righty, um so we have Phillip here, is going to be taking notes and Dave is going to be jabbering and we have the blue sheets going around well. Indeed, okay.

B

So we have three drafts currently adopted.

B

These are the three they've all been updated within the past few months and we will be hearing about them today and the you may have noticed- or at least I noticed that Julius did a lot of activity on profile and it's now profile- oh seven, and so that's good we're getting hopefully going to get that all taken care of. Oh, that's it. Okay, so I guess that means next up is Daniel.

C

C

That's gonna be fine for Ted next, so outsourcing a home network, authoritative, naming a service so well.

C

You should be aware of this draft. Well, they have been put on old, since we believed at the time that we need to make some more progress on the home net. Specific naming architecture, so there have been a few revisions were just what the intention was just to say: hey we're still here and we still believe very useful.

C

The draft seems to mean a relatively good shape, so it's more. We might have a few updates to really integrate the home net naming architecture so, but it's more needs than the structure itself. So how do we position? I mean? Are it's not that we're doing two times the same thing? The position is that the simple naming architecture makes service hosts reaching discovering each other within the home net and with no configuration and reaching home net services or oaths from outside.

C

The network's is out of scope with a simple naming architecture, for example, and DNS sake is also out of mostly because it requires some configurations so the front and makes their hosts within the network reachable from the outside and so outsourcing made sense only because well, there is no need to outsource your alternative service outside your home net. If you are not expecting to reach those hosts inside a home net, it requires some configurations and DNS.

C

Ack is a in scope so to put into picture the thing we are interested is to generate the DNS at home net zone so internally and then to generate also the DNS icon net zone. That's going to be published publicly so outside in an outsourcing infrastructure. The idea is that from the outside, you can enrich the zone hosted on the outsourcing infrastructure and while when you are in the home net, you reach the one, the internal view inside the home net.

C

So that's, but the mechanism see we're describing it's a mostly the one represented by the red arrow, so we are clearly outside the home network.

C

C

So the home net requirement recommended that the home net own remain reachable in kind of in case of connectivity disruptions. So that's the reason we have one home net zone host inside the network, but from outside we prefer to outsource to zone so that you do not expose your home net to DDoS attacks, for example,.

C

The so you have to to generate the home net and then to outsource it. So the mechanism you use to synchronize the home Ned's own inside your network or I mean the one you generate with the outsourcing infrastructure. Is we use a hidden, hidden primary mechanism?

C

So basically the the outsourcing party is going to be secondary and you maintain the synchronization between those two, the NSX signing can be handled by your home nail home net, authoritative service, so I mean basically your home net can sign the zone or you can delegate signing to the outsourcing infrastructure. So in the current draft we detail what could be the motivations for one of the other, but we somehow recommend that it's being signed by the home net and the while.

C

You also need to handle the the reverse zone so which is usually handled by the ISP, so configuration well. We you need some configuration for sure. So the idea is to make this configuration as easy as possible and maybe two.

C

Yeah that well that's a well, that's an point. So what needs to be the? What needs to be defined? So we need to define the synchronization server. So that's where you're gonna outsource your zone and how you going to Atlantic eight with the outsourcing party, the authentication data, and you also need to know from the outsourcing service infrastructure, where your data is going to be reachable because well, the public's authority server is going to be handled now by the outsourcing infrastructure and also the register home net domain.

C

So basically the your domain name, you want to host your home net. Well, that's something you have to configure! So that's what you need for the synchronization and you also need some parameters that you already need for the configuration for the synchronization to generate the home Ned's own. So basically, if you're gonna outsource eurozone on one given IP address, you need to configure the NS are set with the correct address and you also need to let the known know. What's going to be your domain name associate or your home net.

C

So there are no no match parameters to enter. They could be pretty easily done, so that's where we think the document is, we think it's rather finalized. It needs maybe a review some clarifications, but we think we're pretty much ready once the probably the naming architecture is also in good shape. So that's just. We think there is no much work.

C

That's what I mean should not be too much discussions at that point, so I'm just summarizing very briefly: I, don't know how much time I have, but there is another draft which is about how to you because well we said there are very little configuration parameters to have provide, but we would like provide some of some some way to eventually ease that's configuration, so easing the configurations means what can be automatically being provisioned should be, and the other thing is that if a user is entering those parameters once, for example, if you change this box, its CPE or whatever device, there might be a way that the ISP, so that's the party we adapted with the HTTP options, provide the necessary information so that any new box you just like them- and it's fine at one point- you will have to provision this this information, but it can be in the worst case, being only done once so yeah.

C

So it's not zero configuration in most of the cases, but it's it's a very little configuration.

C

So the basic scenario we had considered that the ISP is providing most of the information. So why the ISP well first I, was working for a nice big at that time.

C

But then, because of the reverse zone, the ISPs, or at least providing you at least the IP address- and most likely is the party that has to to handle the reverse zone. So, even though you don't want to interact with your ISP for your domain name and and so on, you at least interact at least for the river zone. So is one of the actor you have to. You could be considered as a potential candidate, and he has well specific relations with the the line. So you can authenticate that and well what can be used.

C

Dhcp is a good candidate as well. So that's for the basic scenario. In most of the case, the ISP identify your whole net with a domain name, so you can use this domain name. Even it's not a human friendly, you domain name is, can still be used. It gives you the IP address.

C

He has some means to a sense. You get your line in you in yourself, so special relations made to make the ISP still a good candidate for to help most of the users.

C

It's not that because we consider the ISP as a primary basic use case, that we are ignoring the other use cases and in fact we also described in a draft how a third party register I mean an and a registrar can be used.

C

That is, that is not your ISP. The big trick is that we use a cname and D name. You still be known to the world as a specific domain name, and then you redirect that to something that is being managed by your ISP, and we can also have the configuration to the third party being stalled. India ISP, provisioned it by the ISP so that the ISP prova sent and provisioned is the CPE with the appropriate configuration you can also be extended to an HTTP Jason like these parameters in the filter if needed.

C

So this is the architecture we had so DHCP. You send a message saying this is my public key and your request for a DNS zone template and some synchronization servers. So with that information, the DNS template server is basically the template for your home Nets own. So with you provision it with your home net register domain name with anything, you need, and then you just have to populate the names and hostname and IP addresses of your network, those that you want to be publicly available and etc.

C

River synchronization server is the same. So once you you do the reverse, and that's where you have to upload your zone and synchronization server is where you have to outsource your DNS zone.

C

So well, it's a it's just a provisioning, we're just using data to provision. So this is how to use the third party, for if you want to use I'm, not saying any names, but none ISP names. um You can still do that quite easily, either provisioning either redirecting under the name. You want to be called to your ISP domain name or you can also yeah provision. The ISP with the appropriate information, your home return, needs to use, etc, etc. So I think, overall, it's quite flexible and that's all.

D

Hey Daniel Ted lemon um we've been so I.

D

Think that this is important, work I think that it's useful I there are some things about it. That puzzle me but kind of the main question that I want to ask you is: have you tried implementing this so.

C

Me direct well I have already configured as its primary and secondary yeah I've already synchronized this too, but I know I haven't implemented it, but Jack is doing it and.

E

Where is Jack I.

F

Am Jacques Latour with zero, so I guess where she was been working for last two year on building a secure home gateway project based on open wrt, and the purpose of this is to build a well. You can look it up on get up if you search check your home gateways or labs you'll find like PowerPoint slide.

F

That explains what we're trying to do, but we're trying to set up a secure home gateway with all the latest and greatest stuff, including Dan's front, the naming configuration the goal is to have DNS SEC inside the house, outside the house for the home gateway to use a outsource, any CAD DNS secondary service for the for the external name resolution and internally signed DNS SEC inside the house. So we're gonna implement we're working on right now in implementing this in our prototype. So working on that plus many other things on the project.

F

D

How was the, how is the document? Did you find it easy to follow? Were you able to work with it.

F

It's a little bit long, haha so it should be yeah yeah. It should be yeah, there's a lot of stuff that can be I. Think there's too much details, but the essence of it is there, but there's a lot of details or there's all that many options that are documented defined so and we don't even know, for example, right now in the prototype stage.

F

We don't even know if we have, we do the signing on the gateway for the NSA, but if we do that they're signing that the NSX signing on the outsource DNS service, because we don't know if there's enough gems to do good, DNS SEC on the gateway with open, wrt and a limited CPU capacity. So that's what we're we need to believe with and understand. So it'd be great. If you could.

A

Send some of that feedback to the list? You know what you found easy and hard with the drives yeah.

D

I mean what I was getting at with this. With this, these questions and I'm delighted that you're working on this is that I think that one of the problems that I've had with the document is is that it is quite a bit to digest and I. Think it's probably not as the point is it could be, and so it might be really helpful if you guys could work together to get the document to be more implementable, more.

F

D

To the point so.

F

D

To encourage that yeah we're.

F

I'm also working with Michael Richardson, so he's more verbose than me on mailing lists. So this should help in Iraq outbound communication process.

C

Yeah I'm glad to review the draft I mean it's.

G

Mrs. Linse heard a couple questions for one is I. Remember there was discussion about home, doped, pourpour specials on for the whole network, in your draft, your external zone, how much is synchronized with the previous located Jenna Sloane and another question is, since home networks were intended to be used with zero distortion right then, if you have a million home networks and just leave them with no agent attention, it's quite possible that the records in the zone and devices and the home that alters at some point they become distinct relized in the document.

G

Do you poor? Isn't any measures to detect such disorganization and to make it all consistent again? Because if you take a network, then disconnected for your internet service provider for a couple months then connected to a different provider and variable two conflicting versions of the zone, because you change to the set of devices at the network, then in your document, is there a way to make it synchronous and consistent again? Thank you. So.

C

I'm, not sure I think there are two questions. Yes, first one is how we considered home net, dominate DARPA.

G

The top I think.

C

Darpa so um because we are mostly interested in also seeing the zone I, don't think we we would like to outsource home dot ARPA on on the global Internet.

C

So in that case it's it's a little bit, I think out of scope of the document, because we not really willing to have a dot offer home DARPA zone. I mean it's on the global Internet. Okay,.

G

So extremely the contents of the external zone or the contents that comes out of the whole network, is it a subset of the contents of the whole opera zone.

C

No, no, usually, what is being published outside is not does not should not contain home dot. Appa.

G

Right so the document doesn't describe a use case where I have already like five printers at home, and then they decided that this one specific printer should have a name and I, don't know universe or a condom. Internet.

C

Yeah, if, if you don't want it to be accessible from the internet, global Internet, you should not outsource this zone. That's a you have to two different from that zone, the one that is seen internally and you can sign it with the intersect and the one you do you're going to outsource, but the one we are interested in. The draft mostly, is the one year because we are describing the mechanism to outsource. So we are mostly interested in in how you can outsource Lee's own. That you have two different zone is another issue.

C

We discussed that, but we don't provide any. How does honor are populated, okay,.

B

um Let's go to Dennis's next question: that is, if you want to explore this further, maybe we can discuss it further on list about home that ARPA and I. Think thatís is that a question was around synchronization or was that related? Yes,.

G

Of sync and the contents no longer wishes what was in one copy and then valid. They become different than some point official. They do connect again and there is a conflict of their records because they're different did you Teresa such a huge case and how to deal with it. Okay,.

C

G

You just say you know: okay, no.

C

Iii think so the problem is saying: you have an internal view: I mean a own that is being published on the home net in a zone that is being published on the global Internet and.

E

C

The question is: how do we handle the situation where these two zones are desynchronized.

G

If you have another couple names, I can explain the picture. I can just ask you a list. Yes, okay, okay,.

H

B

I

That better yeah.

B

I

This time you can hear me yes,.

H

What did you do.

I

Is that right? Yes,.

B

Don't push anything.

B

You have to keep it pushed or something.

B

No, we can't hear a word. Your sister, probably.

C

Is gonna say nasty thing about service?

C

Can you share a.

I

B

Nothing after that, it's the weirdest thing.

I

B

J

Forward, okay, so I.

J

Suspect Barbara.

J

Okay, I'm sorry eyes. Mr.

I

Presentation, but two years ago, when you presented this work, I had a question which I, don't feel, has been adequately addressed: protocol which is called din DNS, which consists in simply, you know, making a DHT TPS request to a DNS server and exporting your address. There widely deployed and widely used and I still don't understand. What problem are you solving here that does not solve by the trivial dynamic DNS protocol that is I.

B

Think he's finished and I think you got most of that.

C

C

I'm trying to figure out where you I think we provide two mechanisms to to to synchronize between the HomeNet and they're outsourcing infrastructure, which is the DNS update and the regular way. Dns synchronized primary and secondary and I understand that you would like third mechanisms, which is based on HTTP. So.

I

I'm thinking about a simple end-to-end protocol include the device that wants to export its name over the Internet contacts. The DNS server gets a lease on the name and recontact set say every hour in order to renew its lease.

I

Ok, this widely deployed nowadays, it's a very, very simple protocol. Okay, it's em to end it's not bounded to a particular link. It's not bound to a particular routing domain, and so what I fail to understand and I checked your draft and I haven't found. It is what problem does this protocol solve by the trivial end-to-end protocol.

C

Well, I think we're reusing mechanisms provided by DNS to synchronize to zones.

C

Now, if you're su I, don't know the mechanism exactly you're referring to but I I suppose it's I have a zone and I provide push my zone to the DNS. You just one name. You just push one a a.

I

B

It sounds like Daniel you're not really familiar with the dynamic DNS mechanisms enough to answer Julius's questions no.

K

Yeah Mike so busy. We.

B

K

I

I'll take it for the list.

L

Mark Andrews, what Julius is talking about is way. Microsoft does dynamic, update their servers, reap records having dynamically updated, I prepared a tile. So if you're using Windows ad, the the host, the the node or host periodically, does a dynamic update and adds their own address records to the zone and then that just lets the day without and then the Windows server does something above and beyond the DNS protocol, which is in to reap them after a certain period of time, and you get there as a that provides a garbage collection mechanism.

L

There is ways we can do similar things, but the the DNS update protocol is permanent. You either. You add your record, you remove your record, but whatever changed it. The protocol level lets that is specified an update. It's permanent, okay, I understand what you're you're you're talking about ad. How ad works, Julius so ad works that the that you keep refreshing your data and the server does a bit of reaping as a clean up as to the garbage collections. In case, you've lost your power. You walked away whatever that's.

I

Right so the question is: why is the relative complexity of this protocol needed and what does it solve in HomeNet that does not solve like the simpler mechanism.

I

So I realized that this protocol scales much better now I.

B

Okay, apparently me techo thinks you've had your chance.

I

J

I

The point is why what's the Luke, what actual problem is solved by this more complex protocol that is not solved by the simpler one and I? Don't understand that and.

B

It sounds like there needs to be people have time to think about it and access the internet and research more on dynamic, DNS and things like that, and so we'll take that under the list. Okay,.

C

B

C

I think I'm fine.

A

C

I

Many people have read these.

C

A

In the last year, or so so, I don't see many hands like this yeah, an author, the person implementing and one another. So again, if we got to get some more people reading it, if we think it's going to move along.

B

Okay, so I'm going to present really quickly on what babble is doing about babble security and what then we might want to think about in HomeNet. So what babble is going to deliver, it appears is going to be a DTL s based on an H Mac based solution.

B

The reason they'll be working towards both is there are different use. Cases wanting to make use of babble in those different use cases have different requirements. The summary of trade offs, of course, H Mac, is much simpler. It is secure, but it's not encrypted, and it does have just few features as opposed to DTLS, which has a lot of features.

B

I mean it's more complex, and so currently there are two H Mac drafts and 1k TLS draft and babel is expected to pick one of the H Mac drafts for adoption, and these are the points that I have been asked to convey. Let's see so given both a DTLS and a an H Mac option, there is an expectation that babble working group will probably recommend one of these independent of that, but informed by whatever they do home.

B

That needs to consider if it makes sense for one of those two options to be mandatory to implement within the context of a home net use case and in that event, home net would also need to decide if any of the optional to implement features that may be defined in the babble drafts would need to be mandatory to implement, and then the biggie biggie is home. That would need to define how keys and credentials get distributed in a home net context.

B

Any question your comments.

D

Ted lemon, so I.

D

Don't think that the opinion I'm about to express will come as a huge surprise, but I think that we should use the the DTLS document version and that we should distribute keys public keys using h, n CP, which is seems like a fairly straightforward thing. To do.

E

David's Kedah Z uh I disagree with Ted, but I felt I had to cover its head. uh H-Had CP needs to be secured as well at the mechanism we have for that is HS CP over DT LS, which is going to need to be secured by something else already probably tell a search so might as well reuse those adieu Babel over to chance with the same certs, I.

D

Think we're actually agreeing and you just don't realize it I.

E

Just wanted to come up here and say: I disagreed, but yeah, not all right. So the same thing y'all decides.

B

Okay- and we will continue this discussion on the list- hopefully I.

A

Was able session? Was there any indication of timing as to when they might get still.

B

I think they were updating the dates in their Charter and they were aiming for. Was it October, ish or I'm, trying to remember?

B

Okay, yes, Phillip, why don't you.

M

A

Want to see of us, okay,.

B

Me note-taker Philip asked for thee for David and Ted to please post what they said to the list, because Philip did not capture it: Thank, You, Philip, okay,.

D

Okay, my name's Ted lemon I'm kind of a repeat offender here so I just this is a we. We wanted to do some kind of detailed status, update, slash question and answer session on the simple home that naming architecture and we have a fair amount of time, although not an extraordinary amount of time to do it in so I'm, going to try and go over the what I think the open issues are in some detail: wow, that's small!

D

D

I can barely read it.

D

Okay, so what's been going on, it probably looks from the outside like zero progress has been made, because the most recent update to the document really doesn't add much what's actually been going on is that when I was working on the document say a year ago, it became really clear to me that there was a lot of stuff missing that needed to be fleshed out before we actually went forward with the document, and so I actually devoted a bunch of time to that and and Stewart joined me in working on that and and very kindly provided some some funding so that we could do the work.

D

So so that's where that's been going I've been doing a whole bunch of foundational work instead of working on this document, but we've made a lot of progress on the foundational work and it seems to have provided us with a great deal of clarity about how to move forward. So I think that's been. It's been a good investment in time.

D

So one thing to talk about is that there's been some some controversy over whether there should be two naming architectures or one. The reason for the two would be that the simple hum that naming architecture is a bit rudimentary or has historically been a bit rudimentary, and so there was talk of I was mostly the one talking about it of having a an advanced home that naming architecture that would address some of the emissions I.

D

Don't it's necessary more and the reason I don't think it's necessary is because actually the document is it's written if I were to fix all of the bugs, and it would probably wind up doing everything that that really needs to be in the the the sort of must-haves that I see in the advanced home. That naming architecture and that's not to say that so I, the the original home that name a architecture that I did like three years ago.

D

Maybe they had a lot of stuff in it that I'm not talking about here, but just the sort of basic naming things that we care about, or that at least that I care about I think are covered by what we have so far, and let me explain a little bit about that. So the documents divided up into a bunch of sections and two of those sections are the section on name and the section on Authority. The section on name is about what name to give your home net.

D

We concluded that that would be home DARPA and the section on authority is about how to publish records under that name, and there are a couple things to say about this. First of all is that in the name section we actually say that having a global name for the, that is to say a name, that's in the global DNS hierarchy. In other words, not home, DARPA is out of scope. But later in the document, there's a place where we say that we have to support somebody, configuring, the home net, with that global name and I.

D

Think that's true so in resolving that conflict. I think that we would I I. Don't think that we need to get very detailed about that and by the way, when I'm saying this. My goal here is not to dictate to you what I think is true. It's actually to to see if it upsets anybody and if they, if it is upsetting to you, then you should come to the microphone and yell at me and not just sort of grumble so anyway.

D

So the point is that that I think that being able to configure a global name in your in your on your home net, somehow makes sense and probably isn't all that hard I'm not proposing that there be some standard mechanism for doing this, but it should be possible as I say, I'm not proposing that we do a protocol. That would do it automatically I'm just saying that it should be possible to do it.

D

I noticed some puzzlement, so so, basically, what I'm saying- and it's ties a little bit into what Daniel was talking about- is that it should be possible for the stateful, authoritative function that in which you would store your names to be either local or off net. If it's often it it could be a global name instead of home, that home DARPA, it could also be a local, a global name locally. But it's I.

D

Don't think that the home, that naming architecture needs to specify that in detail, I think that if we just have the mechanism for having an authoritative nameserver, a stateful, authoritative nameserver on the home net, which is optional not required in a way that when there, when such a service is available, either on the home net or off the home net, that the home that can know about that, so that the right thing happens, then that gets us the thing that I really care about, which is the ability to have a stateful, authoritative nameserver, and we don't need to require that all hum net routers do this.

D

We just want it to be possible. So I think this is doable I. Think it's actually not too hard to specify this and I would like to specify this in the architecture document.

D

Part of the reason that that I'm saying this now is because one of the projects that Stuart and I have been working on is the Service registration protocol for DNS SD, and what the Service registration protocol for DNS SD does is.

D

It creates a an easy way to do: dns updates to an authoritative server if you have one so that you services that normally nowadays are discovered using multicast dns can register themselves in the authoritative service, and so, if that's and that protocol is designed to be pretty simple, pretty easy to use and if we can have that, as a value add feature on a home net in the home, that naming architecture not required, but but we specify how to do it.

D

If you want to, and we specify how routers that don't do, it can defer to routers that do if one is present or if multiple present then I think that we've solved that problem.

D

So, let's see yeah, so if you don't have a stateful, authoritative nameserver, then it just falls back to discovery proxy, so discovery proxy, meaning that you have mdns on every link and you have discovery proxies on every link and the discovery proxies are when you, when you look up a name. It just goes through the discovery proxies to resolve it and there's no stateful name service of any kind on your home net. So that's very lightweight.

D

It's very easy to do, and at this point we have pretty clear specifications for how to do it, which we can reference, so so we're kind of I, think all of the stuff that I'm describing here is stuff that we know how to document, and it's not hard to document and that's the point that I want to get across here.

D

The one remaining thing that has not been documented is how this stuff gets socialized through the home net, so you've got a bunch of home net routers. Maybe some of them are capable of doing a stateful name service and some of them aren't. How do you figure out what you're doing, and how do you make sure that it works on all the links? We need to specify that that's that's one of our to-do items so.

N

This, this is not so much a technical question or observation as.

E

N

About privacy and I I'm completely aware that when you say, for example, you propose to resolve this by saying that the authoritative function can be managed either by the Rooter or by an off Network Service, and that you're leaving the technical options open.

N

My concern here is that I can see that the vendors will deploy this in a way that requires the off network service to be used. In other words, that requires data to go from your home net to a third party in order to get the naming service, but in doing so that introduces some privacy concerns for me right.

E

N

Would like it yeah personally, my preference would be that if it's my own, that I would like the data to stay on my.

E

Home there, thank you very much.

N

D

N

It's just something to be aware of, and perhaps to mention the privacy considerations, yeah.

D

That's that's I think it needs to be more than in the privacy considerations. I think that we need to document technically how that works, because if we don't you're right, it's gonna. It's going to leave open an opportunity for, for you know, snoopers to snoop in but yeah the basically we've talked about this at length before and in fact, there's some text in the old home that naming architecture the talks about this, which we can probably use. So thank you. Okay next slide multiple provisioning domains.

D

This is a little bit scattershot I apologize for that I've had to prepare kind of a lot of presentations. This idea for some reason, so multiple revision domains, I, don't know how broad the familiarity is with the concept of multiple provisioning domains in here. Basically, the ideas call nets can be multihomed, and that means that you're getting configuration information from two different ISPs. Also you have the configuration information for your home net itself and those are really three separate administrative domains with potentially three separate answers to sometimes the same question.

D

If you query the DNS say say you have a connection through Comcast and you have another connection through Time Warner and you query the DNS for Netflix com. You might get a different answer and if you use the answer from Comcast over the link to Time Warner, you might get really not very happy results.

D

So the idea with multiple provisioning domains is to essentially carry that information to the point of consumption so that the consumer of the name lookup knows what the context of the name lookup was and can use it in that context and not in other contexts. I think that's about all I'm going to say about this. If that, if that's not enough and you need to know more there's a document, I can remember the RFC number off the top of my head. That explains the problem and quite a bit of detail, question seven: five: five: six!

D

Thank you! Yeah.

O

I feel I'm a random open, wrt basher, amongst other things, sweet sorry using the multihoming to provide redundant service or a load balancing or what.

D

We don't actually I mean so I have I, have a list of open issues and that's an open issue that ought to be on my list and isn't thank you for saying that. So the answer that question is both right.

D

You may want multihoming, because you want to get more performance, you may want multihoming, and so, in other words, you want to do NP tcp using both up links or you may be doing it because, for example, you have a home security system and you, like the home security system, to be difficult to bring offline by cutting a wire.

O

Flaky rural ISP, yes.

D

O

Fall back on 4G or 5g Springs yeah.

D

Right so so all those are options and and they're a little bit complicated in the sense that that, depending on what sort of connection you have to your ISP, some things are probably desirable and other things aren't so when I'm talking about multi, multiple provisioning domains here, I'm not talking about solving all the problems that exist with respect to multiple ridging domains, I'm just talking about basic plumbing like if you do that. However, what, for whatever reason you did it?

D

If you do that, we need to say how it works, and so that's what I'm talking about here. I think that you know if you, if you had a setup where you wanted to have, for example, fallback to 3G 4G whatever, but you didn't want all of your Netflix traffic going over that link. That's a problem that I'm not proposing to solve I, don't know how to solve that problem. I think that you could solve it by.

D

You know actually I think that I think that the npvd are a document does try to solve that, and so in a sense, I'm, not saying that it's not solved I'm. Just saying that I'm not trying to solve it, but that would be the place I would look. That's that's a that's a topic that we've discussed quite a bit.

D

We discussed it in the myth working group back when there was a myth, working group and it's still being discussed in the MP BD document, so the MP vdr a document so yeah it's it's it's a it's an issue.

O

In your last slide, you're talking about multiple links, hmm all right are you talking about multiple, multiple physical, subnets or number of subnets? Are you talking about multi humming, in this case right.

D

So hum that's our multi subnet in principle, I mean you don't have to be. You could have one subnet, but but in principle they're multi subnet, they might have. You might buy two or three home gateways, your connection to the tea right. You might have one ISP, but your connection at the ISP might be through several home gateways on the way to the ISP home net is trying to solve the problem of making that network a good network.

D

So if you plug a bunch of routers together, you don't get like three layers of NAT or you know you, don't you don't get a network that where each little link is isolated from the other links- and you don't have to solve the problem of getting the network to work by bridging everything together right, so the way that we normally solve the problem nowadays is we just? We just have a bridge.

D

So if you have a bunch of routers that are plugged together, essentially all of the routers are repeating everything to each other, and so you just have one giant subnet with many routers in it and that kind of sucks. So the idea with home notice to actually have a fully routed network with service discovery that works between links. So if you have a printer on link a and your on link, B that works so sure and by the way I'm using open wrt for prototyping this, so we should talk yeah, okay, so anyway.

D

So the point is that that I think we we have in hand the technology that we need to use the draft that Eric think and I. Think David. Are you on that draft, yeah and I'm, not sure the other authors are, but anyway, that draft probably provides us with the lubber that we need to make progress on this and I think we just need to write this up. I think we know how to do it. We just to document it. So that's! That's another work item, so reverse mappings, there's actually two different problems here.

D

First of all, do we care about we're sorry? Do we care about reverse mappings at all I'm, not claiming to know the answer to that question: I care about reverse mappings, just cuz I find them useful when I'm doing traceroute. Really, that's the only reason I care about them, but that's a good reason to care about them.

D

Do we care about global reverse? Mappings?

D

That's a whole separate question, because if we care about global reverse mappings, then we have to do delegation.

K

Phil hummock I, just like to point out. You just said you want reverse mappings, because I'd like to capture we've, got to make the m-net debuggable yeah and you know that needs to be captured as a high level requirement, because I can't debug my network.

D

Right yeah, that's very good point: I'm, not sure that the home, that architecture includes that one of the probable future working items might be updating the home network architecture.

O

Reverse mappings are also useful for name based ACLs.

D

You should say your name when you come to the linkage.

O

Sure I'm still flow pretty well I.

D

Know but yeah yeah yeah, so I mean there are lots of ways that you can use reverse mappings if you want, but but the bottom line is that there's a there is kind of a big debate in the I in any IETF about whether they're actually a good idea, and there are many people who come come down on the side of they're a bad idea. First I think they're a good idea I'm, not sure that we really need them to be that we really need a global delegation.

D

But if we do, we need to figure out how to do that and I know. Mark said that he had a draft about how to do the delegations using prefix delegation and I know that Daniel Daniels draft also talks about a way to do that. So we have technology to do this. If we want to do it.

D

I, don't know that we need to specify this in the home that naming architecture but I'm mentioning it here, because we need to specify enough in the home that naming architecture that if this functionality is deployed, we haven't said something that would prevent it from working. That's all so we need to think about that.

D

Let's see yeah.

B

D

Also, there's the there's the point that Stuart and I think we're discussing a couple of days ago that actually I'm not sure what Stuart but I, think of steward, because Stuart's mister mdns, and so the issue is that you can. Actually, if you, if you do a query with mdns, for you know, 10.1 68, that when I 2 dot in edited opera, it will try to answer that question. I think is now right like like. If you send an M DNS query, it will it will, and somebody out there has that address.

D

It will reply with an answer yeah. So so that's one way to do reverse lookups.

D

But if we want that to work with discovery proxy, we need to make sure that we actually say how it works, because I don't think that it's obvious the reason I say that it's not obvious is because, since the query is going to come in with DNS, DNS has a way of answering that question and it's not necessarily the same way that that mdns has of answering the question.

D

So we just need to make sure that we plumb that correctly I'm not saying that it's a huge problem, I'm, not saying it's a big can of worms. We just need to specify it.

O

Using static assignments, you want to be able to resolve that name, even if the host that has or resolved that addressing of the host that has that address, isn't turned on at the time, so I think DNS SD would be a fail. In that case,.

D

Sorry, DNS SD, yeah, yeah or mdns yeah.

N

D

Really important to know the distinction between DNS SDM DNS right now it seems like they're the same thing but part of part of the work that Stuart and I have been doing. The last year has been making that not be true anymore. Ok,.

O

D

O

Where the host that nominally owns that address has to respond on its own right, yeah yeah.

D

So, for example, if we have a stateful, authoritative server to which you send registrations, then that's authoritative server can remember this information. Also Apple has a something called a sleep proxy that will answer on behalf of something that registers with asleep proxy. It's actually an EM DNS service. That's a DNS SD service, that's advertised using DNS SD. So if you're a device, that's going to go to sleep before you go to sleep, you can go out and say: hi! Is there anybody out here?

D

Who can remember me and then, if there is, then you send that information to them and they remember it and right now, that's generally just stored in memory, it's just cached! So if the device- that's remembering it goes away. Well then it's forgotten, but you know.

P

How h2o from Apple I'm, not here with a question or a statement, it's a request to the note-taker I think this is a really interesting question. I, don't know the answer to it. I think it'd be really good for the working group to solicit input on. Do we care about reverse mappings? What are they used for and to make that list? Because debugging is one example.

G

P

The wide internet sending email is another thing that sometimes breaks. If you don't have a reverse mapping that probably doesn't apply here, I think capturing those arguments in the document would be good and on the other side, there's the privacy concern that being able to send your machine a query and ask it to divulge. What may be personal information is something we should also weigh up. Yep.

G

P

Possibly the answer to that is that these reverse mappings can exist in some repository somewhere, but not publicly accessible. So friends.

G

P

You as that you, as the administrator doing debugging, have the authorization to look those things up, but a random IOT device on your network doesn't and I. Don't know the answer to any of these things, but I think if we can gather use cases from.

D

P

Will help us make an informed decision so.

D

In other words, one of the things that you're saying is that we need to make sure that even devices on the local link that might leak information, because there are not really owned by you in a sort of a technical.

P

Sense, I'll give a little example that happened this week when you and I would be broken code at the hackathon. Our I noticed that when I was advertising printing service from my Mac, something else on the network discovered it and created a dynamic translator to translate.

P

It's basically offer a print service which is a particular subset of IPP and there's a piece of software, which is very useful.

P

If you have an old USB printer that doesn't do air print, you can run the software on your Mac and it now it's reexpose that on the network as an AirPrint compatible service, I guess whoever was running this on the laptop didn't mean to be, and with relatively little sleuthing I identify to the person who it was and found them in the corridor and said you probably want to turn this off and the fact that it was that easy to identify the human individual, I think would make some of us quite worried. Yes,.

D

E

David's Ganassi Apple, just on that topic, shameless plug again for the DNS services country working group tomorrow morning, one of our main work items is the privacy of DNS as deal as to it was saying so, if you're interested to work on that or see and come tell us if we're doing it wrong. Please come tomorrow morning, Thanks! Oh, look, there's a clock right.

D

A

D

What it's 30.

A

N

D

All right, so, let's move on from a reverse mapping. I think we got that one link naming this is actually we had kind of glossed over this before and I realized that actually, my I was very pessimistic. In fact, I was completely dismissive of the possibility of being able to present a name to a user that they would recognize and understand and I realized. Actually, that's not really the case.

D

We probably can construct link names that make sense, and we probably need to talk about that somewhere, not necessarily in the home that naming architecture. The reason the link names are important by the way is, if you're using discovery, proxies discovery proxies have to answer. Em DNS queries, you a subdomain that includes the name of the link, because that allows you to avoid having to deal with name collisions across links. So so we have to have these link names. The end user might never see the link name we'd like it.

D

If the end user never sees the link name, but if they do see the link name, it would be kind of awesome if it actually made sense and as I was sitting there writing or reviewing the document I realized. Well, actually we could come up with names that make sense. For example, I happen to have Netgear Aarti 17900 I.

E

D

What I run open wrt on and it has ports nature, the ports has a label, one of them is labelled to win and the others are labeled LAN, one two three and four. So if I were to come up with a network name, I would probably give it a name like party, twenty or RT, seventy nine hundred when for RT seventy nine hundred, LAN or RT seventy nine hundred Wi-Fi, which then led me to think Oh.

D

What about SS, IDs and so I went I, looked in H and CP document and discovered that the HNC p document doesn't say how we select SS IDs. We really need to answer these questions, I think so. This is another thing that I think should be a work item, I'm, not convinced that this goes in the home, that naming architecture, but it does. It is relevant to the home that naming architecture. So if it doesn't go in the home at naming architecture, we need to at least figure out.

D

We need to specify something that allows us to solve this problem in a nice way. I think.

O

So, are you considering VLANs and guest networks in this topology also right.

D

Obviously that that would have to work yeah. If you configure a VLAN, though that's something that the user decided to do right, so at least they have they probably you could probably ask them to name it, and then it would have a name that would mean something to them. Well,.

O

It could just be checking a box on the set that says: do you want to have a guests along Network, for instance, and you don't.

D

O

That, under the covers, that's implemented using VLANs for sure.

D

But but then, but then from the perspective of what we're talking about here, if you forgave that a human-readable name, the name would probably be something like RT, 7900, guest, Network right.

O

Accommodating VLANs would be extremely definitely.

D

Useful okay, thank you. I.

N

Ted Robin Wilson again, but I'm not sure that selecting our societies is necessarily as straightforward as you might think. So, for example, a few years ago after the Google snarfing publicity, one of the solutions that was proposed was that if you added underscore no map to your SSID, that data might be collected, but it would be discarded or not used that first. That means I have slightly less freedom in selecting an SSID than I had before.

N

Second, when, when I then added a wireless repeater to my home network, the wireless repeater expected to be able to add a suffix to the SSID, so underscore no map was no longer the end of the SSID and I have no idea whether it still worked as a as a way of suppressing use of that data. So there are a couple of reasons why that might not be quite as simple I suppose. Another one from a privacy perspective is that I don't necessarily have to broadcast my SSID.

N

There will still be one better than necessarily have to make it public and in privacy terms. I might want to do that, but on the other hand, if someone can't tell which, which Wi-Fi network they are seeing and trying to join, they might actually end up breaking a law. I mean, maybe not in my case, because I said it boss read on him, but if it was an insecure wife like network with a non broadcast SSID and someone connected to it and, strictly speaking in some countries, that's illegal.

N

mmm So it's it's a slightly weird area. Yeah.

D

Yeah I mean I, don't think that we can perfectly solve this problem, but I think we need to at least specify some default behavior that that users can can use, and you know we we already see default behavior and the default behavior typically is. The name of the network is Linksys, which is actually really bad, because it means that whenever I wander by a Linksys box, that's the same SSID, if I ever if I ever enable links.

D

This is a default, join Network, I'm, I'm screwed, so you know anyway, yeah I agree, that's a that's a deep problem, but we don't have to one of the things that I've been learning through this process is that best is the enemy of good enough, which I already knew, but but it's amazing how many cases that applies and I think this is one of them.

D

Ok, so this is the other side of the state full versus stateless, authoritative, naming thing, which is that if we have stateful, authoritative naming then the way we do service discovery is different than if we have state less, authoritative, naming state, less authoritative, naming would be discovery proxies so so, and we'd like it to be the case that services that support stateful to service discovery, which would be new services, because this is a new protocol. People don't know about it.

D

Yet we like for them to be able to discover reliably that this functionality is available so that they can use it. So we need- and this isn't there's nothing hard about this- we just need to have it in the document. I wanted to call it out, because the document doesn't currently say how to do this. One thing that's a little bit controversial is that in the DNS SD Service registration protocol document- and this hasn't been- this hasn't been discussed by the DNS SDO working group.

D

Yet so they may also poopoo this idea: I just co-opted the domain search list from DHCP before and Ra and said that's going to be one domain name, and that domain name is the domain name under which you will figure out whether or not you have service, because search lists are a terrible security idea and we don't need them, so we might as well use them for something else. Another way to solve that problem would be to create a new RA option and a new DHCP option. That might be a better way to do it.

D

In fact, I am not even sure that I'm on the fence about that, but the problem with that is, it creates more work. So we need to decide what we're doing about that and yeah. So if we don't have stateful service discovery than services that are trying to publish their information have to support mdns. This is something we really already knew so I'm not telling you anything new, but we just need to specify in more detail than is currently in the document. How that all works?

D

Name, resolution: ok, so name resolution we unless the home that has a global name, which 99% of the time I, think it's not going to. We can't use DNS delegation to solve the problem of how whatever cache it is that you're talking to decides where to get answers to questions about. What's on your local network and in fact, even if we could use it that way, we probably wouldn't get the right answers because of the privacy issues that we were just talking about.

D

So the HomeNet Rotter is need to be able to do bypass for local zones and there's nothing really hard about this. We just need to specify it. It's not specified in the document right now. So that's another work item Janicek right now the document doesn't say how to do: DNS SEC for home darpa, because DNS SEC depends on delegation of trust and a locally serve zone, can't have a delegation of trust from the root zone. So how do you do that?

D

Well, the only way that I can think of to do that is some kind of tofu thing or some kind of enrollment process, or you know, tofu by default. Enrollment is better. If you can do it so we've had some discussions about this. I can try I think it was Tokyo who was talking about this and wanted to work on it.

D

This is a cool problem, I'm, not sure that it belongs in the simple home that naming architecture document right now. The template on that naming architecture document actually says two different things that are contradictory about how we support DNS. Second, we should resolve that one, the other.

K

Phil Han Bay I can't remember any of them off my head, but in the days when, when we're trying to deploy, DNS, SEC and I can was not wanting to deploy DNS SEC, we did develop TLB Dennis look aside or something like that. Well, actually, the zeal there was a whole family of them.

D

K

Know TLB was the one that was deployed, but there are other ways that may I mean the other way that was proposed was to stick a certain. The top level government domain and there are others, are being developed since so yeah. This is a soluble problem. Yeah.

D

Definitely it just you have to configure the trust anchor somehow, so what I kind of think would work well enough to be sensible would be that when you first connect to a home net, if so this this absolutely requires. There's, there's no way to do this without some kind of special behavior for home. That's the behavior could be that you configure a trust anchor manually in your in your resolver configuration on your host.

D

It could be that there's some way that the host can discover the trust anchor for the local home net and having discovered it, then, whenever it goes to that home net, it can know that it's on that home net and not some other home yet and therefore disambiguate so and I think that the way to do that is actually just to generate a dsds record and use that as a as an ID or Jerry a zone, signing key put it in a DNS record and use that as the ID for the network.

D

So we could do this in the home that nating architecture. I don't think that it would be a big hold up on sure it would be more work than not doing it, so we should decide whether we want to do it.

D

Stephen made a proposal, I, don't know if you proposed this publicly or privately to me, okay yeah, so the proposal was that we just like get through this. Just have have a bunch of bi-weekly working group, whether they call them interim meetings, yeah a bunch of interim meetings. This apart. This is something that the various working groups do and it could be quite effective for getting work done. I could just like be in.

D

You know, background mode and grind on this for a while, but I think that we would actually make progress faster if other people are interested in working on this with me, I say me, because I've been doing most of the work on this, there are other interested parties like Stuart, has interested and is on the document and has made a lot of useful, Corrections and comments.

D

But if there is interest in the working group and actually getting this thing done and out, I think that one way to make that happen would be to just do like you know five or six bi-weekly meetings and just like get through it. So.

L

Dangerous again, UCD is really is okay,.

N

D

Mean I think we can just put it to the mailing list. Right I, don't know.

B

Quick ask for um if we suggested bi-weekly meetings by show of hands of the people in the room who would be interested?

B

Okay, so I'm, seeing about half a dozen okay Julie? Yes, I'm gonna! Let you in you behave this time.

I

Okay, so I just like to quickly mention something that I mentioned in London and that I mentioned on the mailing list. So Barbara will probably new to me because she's fed up with new way in.

D

Fact you have just been muted I, don't think by Barbara Julius I.

Q

Think it might be a record cancellation, we need.

D

To put on your, you need to put on a headset or something.

D

Julius we're not seeing you anymore, so things have gone from bad to worse.

A

D

Yeah: okay, good he's, jabbering.

A

So it's alright one of the things I thought about that process and whether it's why weekly or some other date, it was not necessarily to do that and at the end of it that we just published this document. But maybe we were at the end of that. We want to document that's kind of editorially finished yeah.

E

A

Then hopefully we can get feedback from implementers on and then that's what we're not really proposing this together to publication requests for the document, but more just to get the document in the stage where yeah.

D

A

It's people can start implementing and giving feedback yeah.

D

Sounds good jabber.

Q

Away from Julia spellcheck, we need to prototype to play with all of these ideas. Yes,.

J

Q

Implementation status can we have an implementation to play with.

D

Okay, so this is a fine question, any other questions just kidding, so we've actually been working quite a bit on implementation. We have, although so, actually there's something that I didn't actually hit on in the slides, I sort of it's mention in the slides, but I think I didn't actually mention it. When I spoke about it, which is.

D

Yeah, so there's this one thing: when I first started working on the simple home that naming architecture I came up with this really cool idea, called discovery relay and I think that it is actually a really cool idea. I'm, not sure that it's actually, that hung. That is the place where it's most applicable, I think that it's actually more applicable in corporate networks.

D

D

I've done a whole bunch of work on implementing discovery relay and I've. We have a DNS SD SRP implementation that Tim Tim. What's there well anyway, Tim hi, Tim I can't read that from here.

D

So during the hackathon Tim Tim put together a client-side and we're gonna we're gonna work on a server, so we're going to work on a ACTU by 9 to so that will have a server-side version of that and I've also been working on discovery proxy, but I don't actually have a fully working discovery proxy. Yet so the implementation status of this is that we're not quite there yet and I agree that that is a very important thing.

D

The reason why why we don't have a discovery proxy yet is because I prioritize doing the discovery relay work, which is essentially done at this point on. The other issue, of course, is that the code that I'm working on I'm not sure quite how to get out. It's definitely going to be open source at some point, but that's still a work in progress. Oh I might even had a slide about this somewhere. Oh, you know what no it's in a different presentation yeah.

D

So the answer is that that we really do need to prototype this and I would be happy to work with you on that Julius. If you are interested and if you don't particularly want to work on it and just want to see it happen, I think that it's it's my belief based on where the work is going, that this is going to happen soon, but it is not done yet. So that's where it's at.

B

We will take that to the list and create a doodle poll to find well, and actually it would probably be good if you're interested, if you could email the chairs, that you're interested in what your time zone is. So we can at least have some idea of what sorts of times to pick.

A

B

It depends on whether we have people in the u.s. west coast. Where are you oh I'm,.

D

On the East Coast Stuart's on the west coast, Stewart and I have a standing meeting on Thursdays and we often pull in other things to that meeting, so we could potentially, if it's bi-weekly, we could just have that, be one of our meetings. You.

B

Know I, don't think I've seen anyone from Asia, but we should certainly allow that to happen to our list.

D

A

D

That's the end of the stack.

D

Okay, well, this is not long. I mean this where this could turn into a very long conversation, but I, don't think it needs to I just want to sort of point out that we ain't done yet.

A

D

Basically, when I was sitting down working on the home net naming architecture presentation, it occurred to me that there's actually a fair amount of stuff that we haven't done. Yet that is still on my mental to-do list, which doesn't mean that we have to do it, of course, but it is on my mental to-do list and so I figured that I would share my mental to-do list with you all so that you know you can laugh and throw tomatoes.

D

So I, don't think that we actually have a complete home net specification yet and I think that, even when we're done with a simple home that naming architecture, we will still have more work to do. I. Don't think that if you were to just ask some random router vendor to go, implement HomeNet right now that they would succeed.

D

So one of the things that I think is missing is something that's like the home net naming or sorry like the the whole net architecture, but with explicit details about how to put the whole thing together. I don't think we have that I. Don't think that the home, that architecture is that. So that's something it needs to happen, and then you know sort of a subsidiary.

D

Bit of that is that we need some stuff about how h n CP does the various things that need to be done, not just the things that the h n, CP spec does, but the other things we've come up with since then, and also how we operate home net. Naming so I, don't know if we need I think that the SNC know the symbol on that naming architecture could be the home net. Naming architecture.

D

Ops document I think that if we, if we succeed in doing the things that I laid out in the previous presentation, that that will actually have that.

D

But but we need to. We need to have explicit instructions on how to set all this stuff up or else, and it needs to be something that somebody who's, not an ITF person would be able to use to produce a home net router that would successfully connect to someone else's home net router and work right.

D

So I think that there are some more technical things which I'll discuss in later slides and also I. Think that we need to go through come through the home net architecture and make sure that we haven't missed anything because I actually haven't done that recently, I, don't think I've looked at the home net architecture itself in at least a year, so it's quite possible that there are things that we have forgotten about in there the we need to either talk about it or decide. We don't we're not doing so technical topics.

D

We could discover it could discuss in no particular order and in no particular degree of importance the DNS SEC on the home. That thing that I was just talking about. That's that line item hung that identifier, which is probably just DNS SEC on the home net. Key sharing. That's how to set up public keys using H, NCP I. Don't think that we actually have a spec for how to do that. Yet we need to do that. I, don't think it's hard, but we need to do it.

D

One of the things that home that to do when your ipv4 uplink goes away is that ipv4 stops working on the network which can interrupt communications on the network. I think we need to talk about that.

D

Mabel keying, Barbara actually already talked about link naming I talked about SSID naming I talked about so configuration. It would be nice if there was a way to write an app for your home net. That would configure your home net that isn't proprietary because I don't know how many people here have worked with the various IOT devices. You can buy for your home now, but it seems like each one of them has an app, and this is not good.

D

You don't want every single device to have its own app and the only way that there can that it can be true that every single device doesn't have to have its own app as if we have a standard way of configuring home nets, so it might be worth writing a document that explains how to do that.

D

It might also be boiling the ocean I, don't know, but I think it's worth talking about some point enrollment. How do we establish trust this I think make it was? David was alluding to this problem. We don't actually have an answer for that. We need to have an answer for that delegation of zones. We don't actually have aspect that we've selected, although we've certainly talked about it, and what Daniel presented is something and I know that mark also worked on something. So we have potentially answers for those, but we need to actually make those happen.

D

I think it's actually kind of important that we consider what the home that working group should say about port control protocol, because the ITF put a lot of work into port control protocol. It's actually a really great protocol for doing things. You might want to do on your home net and it would be nice if you could do them.

D

So we need to have the discussion about multiple provisioning domains. I, don't know that. That's necessarily a simple hum that naming architecture thing, which is why I mention it here and I think that it might be interesting to talk about how you use devices on your home that how you connected devices on your home, that, when you are at home.

E

D

That should work without you having to let everybody else get onto your home net, so I don't think we have a document that explains that right now and I think would be worth writing one. So these are the things that are on my list: I'm sure that there are things that are on your list that aren't on this list and I suspect. There are also things that are on my list that aren't on this list, but this is what I came up with so.

O

The external IP before failure trickles inwards, it's.

D

In the hn c, hn CP spec, that, if your, if your ipv4 uplink goes away, you d configure IP on all internal interfaces, and so I think the reasoning for that is that you wanted to avoid having devices try to use ipv4 when it's not working and in principle, if you have ipv6 working on your home net and you're, not using ipv4 for connecting for intra HomeNet connections. That should be okay.

D

But if that's the way we want to solve that problem that we need to be a lot more explicit about that than we are right now, because, right now, my experience when I first started using hn CP was that I was happily using my network, and I did something and about ten minutes later. My whole network just stopped working, because I didn't actually have ipv4 working our IP v6 working on on on the whole routed network at that time, and so when my ipv4 uplink went away, I just couldn't use my network anymore.

D

This is this is not desirable.

O

Very least unreachable yeah.

D

I mean: do you think that's really a problem at this point, I mean you. You have happy eyeballs running on most things at this point and so in principle, if you have working ipv4 and non-working ipv6, you should just wind up with an ipv4 connection and vice versa. So I mean what you just said actually applies equally and is the motivation behind the specification hn CP that if ipv4 isn't available, you drop my the ipv4, because you don't want people picking, ipv4 and not being able to get off the network.

O

D

So the reasoning was that, in fact, ipv6 would always be working, because you know that is this is the future and ipv4 might not be working.

O

D

No you're actually in the future, I don't know if you've noticed it's just that it's not quite the future. We expected, and it's also kind of unevenly distributed.

O

D

B

Anyway, is that it that's.

D

B

Okay, so I think we, the chairs, have an action item to set up some doodle polls and things like that for people who want to do the bi-weekly meetings with I'm Ted to try and get this to a real implementable state.

B

Did we have other action items.

B

Okay and I'm going to be pushing Julius really to get the Babel profile. We're completely done done in the editor skew.

B

We might do a collection Kitty to get Julius a new, headphone, yeah and I. Think that's it. Thank you, everybody. It was fun.

M