IPFS SIG - Local Offline Collaboration, 13 Nov 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Local Offline Collaboration Monthly 2019-11-13: Cohosting

Description

On our November 13 call, Marcin Rataj (@lidel) gave a demo of the experimental IPFS co-hosting spec and its accompanying CLI tool. See:
https://github.com/ipfs-shipyard/cohosting
https://github.com/ipfs-shipyard/ipfs-cohost

We also discussed Google's new Web Bundles release, as well as use cases which are blocked by traditional DNS reliance.

Learn more about the monthly calls of the IPFS Local Offline Collaboration Special Interest Group: https://github.com/ipfs/local-offline-collab/issues?utf8=%E2%9C%93&q=+is%3Aissue+label%3A%22%F0%9F%93%9E+Monthly+Call%22+

A

Hey folks, this is Terry we're here with the November 13th edition of the ipfs local and online collaboration. Call, there are probably a few words I missed in there somewhere, but lovely to see everyone. So we are starting off today with a demo from Lyta, which I will let him describe what he's done so feel free to uh steal the screen.

B

You this will be a very brief and mostly.

C

Like apartment for discussion, we've been always talking about hosting website on idea fires, but what about co-hosting? If there's a website I want to contribute my bandwidth to to keep it alive or I want to sort of like us like follow a website? There's a new version, new update. How can I maybe precache those new versions, so those versions are available to me instantly because my local node already has them- or maybe other people in my local network may benefit from having access to those new versions.

C

So, instead of like going too deep too soon in low level parts of ipfs Park we've been thinking, what can we do with existing api's? So we have this concept of NFS, which is immutable. File system is basically an abstraction there's, a pretty good.

B

Pretty good tutorial on that, if anyone owns to learn.

B

A

At the other products, cool tutorial, yeah.

B

A

B

A

Made that it it's.

B

A

B

C

You want to learn on the details after the fest there's a tutorial, but basically we been thinking. Oh that's, a kind of the highest abstraction level API. We have right now, it's very familiar to people. It exposes a file system like abstractions and we already build this with file screen in our web UI.

C

We let you browse.

B

Those EMFs resources, so it looks like file system. There's a file, that's directory. um So that's.

C

When we we started figuring out, okay, so can we build something using existing primitives to simply make it easier for co-hosting website without the need for installing something bigger like ipfs cluster like ipfs cluster is great, but not everyone wants to run a cluster of ideas. Nodes most of people run just like IP, first companion and I default desktop, and that's it right. So the idea was oh. Can we create something much simpler, but.

B

Targeted at those users who go to some very good tutorial or.

C

Maybe Wikipedia mirror, or some blog they like, and they want to simply help contribute bandwidth and storage to that. So we came up with very simple spec for basically dots and if you go that lip the meta repositories in ipfs Shipyard under co-hosting people I'll post a link later in the notes. But basically the tech is very simple: it's a set of conventions for storing those websites on NFS and a set for basic operations such as adding removing listening updating snapshots, maybe changing co-hosting type.

C

And then, when you run out of space, you can prune older snapshots. And basically, if you look at this, it's like super super simple. It's just creating directories and removing directories and checking if directory exists. So if I create a directory for some site, ID and site, ID is basically either domain name or IP NS a key.

C

So when you create a directory for a specific side, the co-hosting simply will fetch that website and kind of import it to your local node and the there are commands for updating snapshots and there are two ways of those snapshot. The two types of those snapshots. So one type is lazy, which means you will simply add this websites to co-host Inc locally, but you don't fetch entire thing.

C

You just keep like a pointer and if you go to that directory here, let's say: if I go to co-hosting as a fool, but let's say it's lazy or the purpose of this and there's some other stuff here. Only the stuff, I actually Traverse will be fetched. That's why it's called lazy. It does not present I think an option.

C

The second type is full it just the moment. You add it to co-hosting it fetches entire thing. Why we have those two types of hosting in this initial proposal, because I want to emphasize it's an experiment, so why we have two types: one, not every it's targeted to regular users and basically consumer-grade like laptops. Lady phones, don't have that much space. The space is cheaper, but not that much of the space is available in local machine.

C

So let's say you go to Wikipedia Mueller and you want to co-host that here's the edge case, the English Wikipedia mirror on ipfs, takes 650 gigabyte and a lot of people don't have disk of size. That would be able to fetch it to keep basically contain that amount of data. However, still people don't really care about. Maybe people don't care about like entire. You should speaking of people mostly care about pages, they visit and that's perfectly aligned with how ipfs works the blocks.

C

The pages you visited our cache in a local activist, not already that's built into ipfs, so the.

B

C

To simply leverage the fact that things that you absolutely fast, our implicitly pimp, which means those things won't be garbage collected when you you know- starts running out of space. So the only thing that lazy co-hosting does it creates those pointers in amethyst to assure if you visit a page on Wikipedia, that page will be implicitly kept around until you decide to manually, remove it from our house and that's the difference. The full full snapshot is when your snapshot entire thing and it's all the blocks are fresh.

C

The moment you add a page to your co-hosting setup. So I've been talking a little bit about specs and the high-level concept, but we actually have a common line tool can.

A

I, just can I just interrupt with a quick question, so the person who's making a directory called co-hosting is the person who wants to co-host a website or the person who makes the website originally has to say I'm setting this up so that it will be easy for people. The co-host.

C

Oh, the person that is setting up the website does not have to do anything. Okay, it's just people who want to contribute, can use this set of conventions and the command line tool. I will demo shortly to simply country to make to make it easier for them to contribute to that website. You don't need to even know who created that website. As long as website is on ipfs and is using either IP NS or dns link, you will be able to co-host it and yep.

C

So we, the over over arching idea, is to kinda to use this spec for experimentation and then maybe start with implementation. So the first implementation is the command-line tool which is easier to play with and easier to change the future. But at some point we would like to introduce those integrations into our browser extension, which is a diva's companion, which means when you go to distribute Wikipedia. There would be like user interface element which lets you simply add Wikipedia to your co-hosting website or any other website.

C

Another thing is how to make sure those websites are updated kept up to date. In the background we are thinking, maybe we could make ipfs desktop periodically check each site. You have in your co-hosts Inc roster other than update and if it's like a full snapshot website which you want to do full snapshots of it would like prefetch that website for you. But at this point we only have this command line tool which I want to demo shortly. You eat a node app. You can install it from from NPM. It's.

B

Called ipfs co-host and I believe I should be able to demo it here, yep, so you've.

C

Seen nothing so the idea is: okay. I want to add website dog's IP of SiO, 2 micro, hosting websites to micro, hosted website, so maybe I'll show you available commands yep. So basically comments are you. You can add website you can remove website. You can lift website. You are co-hosting.

C

You can change the type of co-hosting between this lazy type and full type. If you let's say you have small discs, then you move to a bigger one. You you may want to switch to full, or maybe you decide. Oh this website is not that important to me. I'll start doing the I snapshot. You can change the tag here. That's a sync command which goes through every website and check, see, there's a newer version and if so it stretches it and update your co-hosting directories, you can add it to like cron or some periodical schedule.

C

Your machine is doing so. It happens in the background and also things like IP us that could have that as an option at some point in the future and running out of space. You can remove all snapshots apart from like that last one is.

A

That syncing to the most recent version, a thing that's new to this tool or a thing that if something is on IP, NS or DNS link right now, that's the thing that already happens.

C

That, if we think we check ipfs either we check TNS or we check DNS link. If there is an updated version and by updated version we mean oh, if the latest snapshot different than the one we have right now, it may be better if I like illustrators, so it's active as co-hosts, say: I want to add dogs at FS, IO and, let's say I added this as lazy.

C

There are pretty emojis here, but not.

B

C

During mind stuff, all right, so I've added what happened here. I have ipfs daemon running on my machine. That is like I, give a desktop guide. If s whatever you have, you can see that the tool fetch entire dogs, ipfs IO. It results it to this content. Id and entire website takes seven megabytes in size, but it did not start and I think cause.

C

It was like a lazy version and if I go to web UI files, you can see I have in files hosting directory created and if I go there, I have full and legit sub directories and if I go to lazy, I have a domain name of the website, I'm co-hosting and if I go into dogs, IP FSI, oh I, have a timestamp of the latest snapshot.

C

That's only one snapshot because I just added it and if I go inside that snapshot, it's basically the website and I can even see the HTML even more I can copy hash and probably probably browse it, but more or less. The idea is that you have access to snapshot using web UI and, at the same time, you are contributing to co-hosting that website.

C

Another cool thing is that you don't need to use command line tool. If you decide that you don't care about this website, you can simply remove it here and.

B

It's no longer lazy, co-hosted.

A

Is there a reason that this needs to be split at the folder level between full and lazy, because it feels like, as an end user I would want to see the domains right here, not a full yep.

C

That's a that's a very good question, a question and kind of open one yeah Richard I.

D

Also am interested in that question, but I think I wore in teresting in answering it from a different perspective, which is I, don't think an end user should ever see this screen. Oh yeah I'm super excited about this work because I think it lays the foundation for higher level user flows like you're in a browser, and you hit a button that says I want to co-host this website. Oh yeah and.

E

D

You can unstart co-host and a user never has to understand that ipfs is even involved at all or care necessarily, and we have maybe higher level visual design guidelines for what the repercussions are of your browser being a server that is serving content, whether in site, BFS or any other distributed protocol. So I think I think this isn't I'm super excited about this work because it lays the foundation for other actual end user tools, of which I think I PFS web UI is an IP FS user tool, not necessarily a web user tool.

D

At this point, I think, oh, you went once these kind of conventions are tested out and we have a better understanding of what does lazy actually make sense. It's actually saying this will be destroyed at some point. It's not really lazy, so there's a whole bunch of other things that are probably need to be communicated through there as well. But this is an early experiment and trying to figure out what is the right? What maybe what's the right language? What's the right pattern?

D

Do we need lower-level protocol support for things like this or is a convention at the mmm fest level? Okay, is that is that a coarse enough tool for us to experiment with, but I would say from an end-user standpoint. This is all still very very far from oh.

C

Yeah I kind of I started with the most boring part, especially to highlight the fact that it's like very, very unusable to end user, because you should think about this as an early experimentation of the simplifying our API or like simplifying those concepts to the point we can make a library from them and then we can build a user interface around them and to illustrate that there's a full and lazy subdirectory here mostly to make it easier for people who started playing with this command-line tool to quickly understand. Oh those snapshots are full snapshots.

C

Those snapshots are part partial. However, the end user, the end user goes to this website right and the end user can click in like ipfs, companion or click here, and they see option to pin this. This is how it looks today right you go to a website bit and you can pin this website. What does that mean and that's the entry point which we will eventually want to replace with co-hosting, but all those details, I just mentioned in the past, like 10 minutes- are differences between full and lazy conventions for directories. Those are low-level things.

C

Those are high-level things given existing API stuff. Those are super low level things from the user perspective. People won't ever interact with them unless they really care about. Oh I've seen some picture on that website. I think I'm, co-hosting it I'll go to web UI and traverse the three three and see that picture again right. The idea is to replace this element here: pin ipfs resource right now. If I click on this, it will pin entire website recursively and that's it.

C

There's actually no UI for browsing all the websites I spin and if it's a Wikipedia, which is 650 gigabyte, it will be either file in the background or it will I constantly try to pin it.

C

You don't know, because we don't even have an API for tracking that- and that's the mostly that's the key purpose of this experimentation to even to identify those missing pieces to create user interfaces that we want, because if we have here, option contribute or like co-host this website or something like that, and when you click that you should not be prompted to big between full and lazy. There should be at the seat like safe default or some heuristic.

C

We can tell that your node has this store like this limit of storage, and if website is over the limit, it should either pick lazy by default or inform user. Hey. You don't have storage to host a full snapshot, but you can host away it like the opportunistic or latest action. Even the language. We are trying to invent here, lazy and full. It's just like temporary language. We invented, so we kind of start a discussion about those concepts here.

C

I fully fully fully agree that it's not end at end users and this tool is mostly to illustrate the idea, but the goal is to eventually either take this set of conventions of keeping stuff on amethyst, or maybe we will create a new coral PI's. But this way we can iterate faster without qlik. The long process of changing correctly is we can like use NFS called build conventional top of NFS, see what works see. What does not you?

C

Just bacteria you just identified that Oh lazy wives there I would expect those domains yes totally, and that was like one of those discussions we had. Should we should we like put domains here and have like a flag inside of a directory. So let's say: if I have like lazy, oh I will move it I need to add it again.

D

I would I would yeah I think looking at looking at that list is more like looking at when you open your browser it actually or like so. Your IP fest repo is in a hidden directory in your user directory, or your browser profile is also you know in the application, support directory or in hidden directory in your operating system and I. I think that this is really what we're looking at is kind of the contents of that hidden directory, and that's probably the best corollary to think about and.

C

Also, keep in mind that right now we just dropped it into co-hosting directory in web UI. However, if it becomes a thing this that could be like a special director which is not on this regularly. It could have a different user interface, which is more around the concept of co-hosting, and things like that. So it's just like a conversation starter more than actually usable thing, and once again it's like experiment in mfsb. Well,.

D

I think I think it's cool is. Is that really displace the strength and power of having that filesystem virtual file system so and to some extent it talks about the strength and power and actually just using a file system instead of a repo?

D

And that's one of the when people contrast dat an IP, FS ipfs seems a little opaque from a developer standpoint, even because everything is hidden away inside this repo, even in a git repo you're, still talking about a repo but really working directly with file system and ipfs is really different in that way, because that repo is that black box and you're not working directly with the contents of the files you're taking them and putting them in the fridge. And then you close the fridge door.

D

And then you open up the fridge and you pull something out. Then you close the fridge door again. So your visibility as a developer or designer of the sum total of things that are in the fridge is always obscured by that door and that's how kind of how I see this type of issue so the power of having that file system view? What m FS provides that mutable thought what we call medium file system? Other people just call daily reality.

D

What they have all the time is is that it gets us a little bit closer to be able to do this type of experimentation, and the fact that that you know Lytle and and hak-kan and oli can use that MF s abstraction be able to come up with these types of conventions that allow us to play with higher level application patterns. Like would ya a little bit closer, at least to what a user just wants to do. I want I want a Wikipedia extension that I know.

D

If I have that extension installed, it just auto saves the Wikipedia pages I've been it serves. Those like that. That's really what I want to do. Is user I? Don't care about any of this stuff right. So this gets us a little bit closer to the point and be able to say do something like that. Will you take the core of I'm envisioning?

D

You know broad strokes again, a future where we have maybe a share or a send, or a save or republish extension, not an ipfs extension, so something that translates to what user needs are a little bit more directly and the experiments like this are fantastic because they help us kind of envision. What what those opportunities and those permutations might be.

E

Adding to that I think it's really I love the fact that it's path based so stuff like that when you you're not only talking about like cross language. At that point, you can potentially have like cross application, utilizing the same data as well. So when you're talking about like, send and receive and stuff like that, the in my mind, that's that's where I'm going of like. Oh, it's really cool that we can just have this convention that is shared across multiple applications like that. That's really cool.

A

If someone figured out how to host on ipfs a website that was more dynamic like a single page application or something and dealt with whatever you have to deal with with hash rubbing and all that jazz, would that then be able to be co-hosted like this too, or would it run into extra bumps?

A

The reason I asked is like, if you're thinking about an application, where, like you put whatever you want in the URL, and then it was sucking in data from an API or something. How are you predicting all of the possible permutations? You know what I mean. Oh yeah,.

C

Yep, so that's by going back to the like general question: okay, so stuff on IPF ifs is great for immutable, static websites. How do you introduce mutability in general, I? Think that's sort of like separate.

C

It's like not specific to co-hosting, but like co-hosting, is mostly about about keeping those initial assets. You need to fetch to bootstrap the dynamic web app the dynamic web app could be like jeaious, which uses CRD T's or some other magic to introduce, like mutability and persistence of mutated start state. Somehow, however, to boots to like boot that web app, you need to fetch those static assets and that's what you would be co-hosting.

C

So the next person that wants to visit that website for the first time needs to fetch those static assets, and you will be one of peers that provide them. Yeah I think that's separate kind of separate discussion from hosting my sort of like the default use case. I have is most like Wikipedia cuz, that's a very good example of each case when you have a very valuable like universally valuable website to a lot of people, but at the same time, it's too big too. We start on a single website.

C

You may have subsets of people who care about different subsets of this website and that how can we create UI or some conventions for people seamlessly contributing a fund, wave and and storage without actually making any decision so I totally agree that we, the decision between, like lazy and full, that should be like hidden from the user. Those are the cloud level primitives. The challenge here is to how to frame the language, the user interface and like entire, like interactions or maybe even like the lifecycle of the data.

C

So someone published with a single snapshot of Wikipedia how that single snapshot can be co-hosted by a group of people who actually have no central coordination, people just care about subsets on that website.

C

So that's what what I find it really exciting? And it's even if it's pretty simple, it already opened a lot of new venues for thinking and also probably it identifies problems with our api's. Even the fact that we are not able to track preload of huge websites such as Wikipedia, you have 650 gigabytes and we are not able to track progress.

C

We may be wherever of that problem, but it's highlighted by by this experiment that if we want to build user interfaces, we would have to add api's that give you give us a better insight of blocks tour. So people would want to people need to know what part of Wikipedia I have stored on my local disk right.

C

What percentage of that 650 gigabytes? I already have cached locally.

C

Those are like hard problems not only like technically, but also like how to communicate that some prior arts, like some prior art, can be found in like BitTorrent or other peer-to-peer sharing apps.

C

They have some they're a bit much better on showing what's happening, how much data you already have locally who's fetching from you and things like that, but yeah, that's a co-host expect I, don't mostly wanted to put it there to mention that we started experimenting. We put a pulse on this, so it stays like this at least until 2020, but I believe it's at the point when it's worth playing. We got this common line to really still recently. That's why I wanted to lemon cool.

A

Thank you. Yes,.

A

Dietrich is this web bundles when you.

D

E

D

So I was at chrome, dev summit yesterday and the day before was the first day and they announced web bundles. Previously, Google's been talking about the web packaging specification and the set of related technologies.

D

They were using it initially as a standardization approach towards their amp efforts, which are somewhat controversial but also are, interestingly, they have a characteristic with regards to the web security model that is kind of closer to what ipfs is doing in terms of the separation of content from location. The way web packaging was initially set up was the idea that you could or communicate.

D

It was the idea that you can take say portions of your website as as replayed HTTP requests bundle, those up sign that bundle of HTTP requests and the assets that the server would respond with with these certificates that you have for your domain, your SSL cert and then take that bundle and put it on Google servers and then Google could serve that set of pages from their way from their servers to your users very very quickly, and you would still have the nice little security UI in the browser that says that you are at that actual website say Terry's website, even though that website that is Terry's website came from a zip file that was signed and sitting on Google servers.

D

So it is a implementation of a web technology that actually does separate content from location to some extent, but the encryption pathway that it uses in the and the security model that it assumed at the end I mean a rendering of that website. The browser it still works perfectly under the security model of the web today.

D

So, in some ways in broad strokes, similar the separation of content from location, but in the actual emulation, still stays within the paradigm of the web security model, as we understand it, web bundles is the kind of next follow on follow. On step from the original specification, where you can package all the different assets into one bundle and lital actually has been kind of tracking him and Jim pick been tracking looking this technology for a really long time, but because it's a offline, friendly technology that was announced in its current form.

D

Yesterday I thought it might be worth bringing up here ensuring with people and getting people's thoughts on or I'm, very least introducing people to the fact that it exists and may be implemented at some point as a standard by default in Chrome or a technology by default in Chrome not standard into some other bells mechanism.

A

A swain again how it's nothing friendly. You.

D

A

Talking about like normal web sites on the north, normal internet rate, normal.

D

Web sites on the normal internet, you can bundle up and in there on their web page, they actually share a demo with I'll share this here's the web page for it getting started with bundles on their fancy new website web dev and leave here's a visual aid. All of your web resources, bundled up into one fancy, looks looking looks like an installable package just interesting.

D

However, this visual representation is not not helpful in that it doesn't portray the fact that these are essentially HTTP resources, not just static assets. So it's almost it's like a bundle that contains a tape and that tape is a log of the HP exchange back and forth between a client or server, and you replay those.

D

So it's not actually as opposed as this visualization, and this shows, but it's the high level it kind of works that way, but the demo they have here is interesting in that it is a web bundle between a game shared between two phones that are sharing it over I think web Bluetooth in this one, but basically any you can do it over a local subnet or you could.

D

If there's a way that you could actually get these two chrome browsers that have web mentals enabled and implemented, you can share a website, basically offline, the entirety of it.

D

But I'm not gonna play that video right now. Well, your it looks like Michael already put the URL on the notes. So if you want to check it out speaking.

A

Of the notes, I've been attempting to take notes, but anyone who has actually shared something should go back and fix the notes so they're accurate, because I tend to butcher them a little cool and then was this one new Dietrich, the recipes. Yes,.

D

A

D

A couple of people that I that you know that being being somebody who worked in web stuff for a long time did ever all for a long time, people reach out to me like aphrodite FS, to meet my needs I, and so one person in particular reached out, and he commented on one of the local, not fine issues, I think with some use cases around work he's doing in Southeast Asia.

D

One of these has worked for a UNICEF at a Bangladeshi displaced people's camp, where everybody's got a smartphone in this in this camp, but the internet access is very poor, the quality is low, the bandwidth is low, the network goes up and down, but within the camp there's a lot of interpersonal sharing of resources, so they have things like they've tried to set up like local media web servers, and things like that.

D

But he's so he's is it's almost like a a a partial offline use case, and he said the number one is because a lot of the people in this camp are also not literate.

D

The number one use case they have is for local distribution of media photos and videos, and there are not a lot of applications that will support that, and he said you know, he's he's not a programmer, but he's technical enough and he's been running open source workshops for years, like I, met him in Hanoi in 2010, where he was spending a Saturdays running who Mbutu accessibility trainings for just that, because he thought it was great and people need to know it.

D

Just really awesome person, who's long long, long done this kind of work and knows that the number-one problem he has for these contracts for this work that he's doing for yourself is DNS.

D

Everything in the web is tied to this remote resolution. They were quite like ah there's all kinds of web stack technologies that work in a vacuum offline, just fine, but ultimately it comes down to the fact that he needs that resolution in order for any of this stuff to work, even the stuff that says it works offline, it doesn't work offline out of the gate out of the box. It needs to make that initial connection, even so everything for him as somebody who's in the field trying to build the practical offline applications.

D

Apks are a million times better than the web, because you can pretty apks on an SD card and it's somebody. The Android phone will load them up and install them, and the web is ultimately problematic as long as it depends on DNS. Sorry,.

A

Did you say apk apks.

D

A

D

Which is the package of an android application? Oh, oh, you can take an Android native application. It is a single file that you can put on an SD card, put that in your phone and your phone will be able to open that up and install it as an application, so I again harkening back to the kind of installable web applications that that web bundles, that we just talked about or trying to eventually maybe get to apks already have that form out kind of form and function.

D

Standpoint very, very easy and seamless for people in that regards where it's the web. As long as it has this location based model that it will always have that constraint, it depends on the network coming up from a philosophical standpoint. There is a network dependency from step 0 of initiating web request. Ipfs does circumvent this to some extent, but of course we are in that awkward user experience, standpoint of trying to say well open up the browser.

D

The thing that requires this thing online that general purpose tool that you use be able to get to the Internet at large and then put in this long long string of things. So there we have other problems that we need to solve, but this person asked for basically higher-level recipes for using IP fess to do things like media and image distribution or build applications that might work offline. So I thought this would be the right place, maybe to ask that can be a community of people around where we can find those higher-level application recipes as.

F

Sounds really close to something to the date hope South I have presented some earlier point in time. We could. We could discuss about that. I would be interested to hear more talk to this person if yeah.

D

Thanks I will make the introduction.

C

Something I know notice over and over is how all projects in be web space and also like in the crypto space, maybe crypto notes, because they have like Ian s and solutions, but eventually DNS becomes this single point of failure in in ended up. You look at to let let's say you've got like the distributed music player which keeps stuff on dog chain and stuff like that to actually load the web app. You need to hit DNS first and so that's sort of like been a recurring theme. We are solving all the like.

C

We are sort of like solving problems along the way, but this, like the first thing, remains the problem and sort of like we've been having those like those discussions about.

C

Is there something we could build on top of Lippe to p or maybe like bit additional service exposed by ipfs nodes to harden or DNS or make it like more robust? Somehow, basically, the idea I often go back is what tor project is doing.

C

There is a way to basically make your tor demon to expose dns service on some poor local port, and if you set your localhost as a DNS ever then all the DNS requests, instead of going to your local ISP or someone else, go to tor network and come out at random exit node and that exit node does the lookup and then the response goes through tor network back to you. So it's obviously super slow.

C

But if you are like caching, those record results records locally you only the first time you request for domain will be so and then I wonder. If there's anything, we could do better in IP Affairs, given the fact that, instead of relying on a single node to make a lookup for you, perhaps we could ask a subset of peers that we are connected to hey. Can you resolve this?

C

For me, this DNS thing and then you get let's say 20 responses, 50 responses, and then you can tell oh, like 10% of those responses, give me different IP. Why is that? That could be because they get like closer closer IP because of Joe location.

C

For that specific peer DNS may return a closer IP, but also it could be an effect of like blocking at the DNS level, like censorship stuff like that, and then, if you gather those like, let's say, 20 50 responses, you can tell okay I need like a quorum like 50 plus 1%, to actually successfully resolve a DNS address.

C

So just like a an idea. What is the if there is because I feel there is a need to actually maybe not even we are not able to like immediately replace DNS, because it it's been around for a long long time and will be it's a very well understood. Spec, there are clients which speak it well, both tcp/udp and like HTTP.

C

So there's the idea that if we expose DNS service the card and DNS service as a feature, maybe of the p2p or IP house, could that be useful in the context that should describe the trick, because in that environment, perhaps connectivity to DNS was limited. But maybe if that person was connected to like 50 peers in that specific country or place, someone could have that record in their cache. Just like just just might mean using around yeah.

D

I really like that, from a from a rola like robustness and reliability like I'm thinking about the worst possible Network scenario. That seems to be a a good characteristic of a reliable using using p2p to do something like bolster the reliability and robustness of a of a DNS like technology, I think it it's really different, the type the type of kind of like blue sky. What? What are the characteristics of? How do we design a new internet protocol from scratch, with characteristics like that built in I?

D

Think we still wouldn't end up with something like DNS. We might end up with some some kind of fallback mechanism like that, but when you talk about kind of like what the what the like golden the ideal path through would be like there, there are ultimately just better ways of doing that, but as far as a way of bolting on some more reliability onto a system that is deeply ingrained in every almost every layer of a lot higher parts of application.

D

Development yeah, like even the diffuse ipfs based music, app I, downloaded the desktop version and, of course, requires me to make an HTTP connection. Even for the ipfs part. It only works if I can actually get cores to their website and.

C

Just just to be clear, even if we don't expose this as a service to other apps, we lacked all is doing effectively that you can point any the app that wants to resolve DNS at this port and it acts as a DNS server.

C

Even if we don't expose it as an external service, we still could use it to harden a DNS link websites because, instead of basically relying on the DNS that is returned by your operating system, we could add this additional readability layer to DNS, linked websites and say I need this like sort of like a quorum or because, in case of DNS link, we don't care about a record. So this case, when DNS returns, you IP that's closer to you. It's no longer a case.

C

The only type of records we care for DNS linked websites is txt and that one should be universal. And if someone is sending you a different DNS link and everyone else is sending as something else that may be, because that one has more up-to-date version, but it's more likely that someone is doing man in the middle at the DNS level. And if you ask more and then a single DNS server, then you are able to mitigate those types of attacks.

C

Or DNS acquire sites work in an offline environment because someone you are already connected to has record cached, even if uplink is down.

A

When I saw that issue that was posted, we have had a number of discussions at offline camp about either stuff in refugee camps or disaster recovery, a lot of it involving like mesh networking.

A

So there are articles both existing on the offline, can't medium of publication and supposedly coming soon from some of the folks who were just there like. We had a refugee camp discussion and we had a building for a load of middle income countries discussion. So there may be some new stuff coming out with those angles.

A

So we have seven minutes. Is there anything else? People likes general today.

A

To play the jeopardy theme song for seven minutes.

A

No all right. Well, it's very cool. Well, thank you for the Domino's very interesting to see. So if we were to like right now, we are off of our schedule because we did the second Wednesday instead of the third. If we were to go back to our normal schedule, it would be Wednesday December 18th, which is one week before Christmas I, don't know how many folks are taking super long vacations and how many people would still be around on the 18th I'd probably still be here. Does that say you can't seem reasonable people yeah?

A

Okay, all right! We will get back on that schedule then and I think just for the sake of reducing admin, workload, I'm gonna switch. It's been like a different issue per meeting, but I'm just gonna switch it back to one issue and change the name, so it reflects the time of the meeting. So I will just make anyone that will keep forever and keep updating.

A

So if anybody has ideas for future events, as always, there will be some kind of issue there that you can drop them in and we can keep updating the same agenda here with the next time. So I will see everybody in a month ish or some of you next week. Some person see you later.