IPFS SIG - Local Offline Collaboration, 18 Mar 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Local Offline Collaboration Monthly 2020-03-18: Disaster Response, Medical Imaging, NFC, Web Hosting

Description

On our March 18 call we discussed offline and decentralized approaches to data collection and disaster response (0:16), making medical imaging available on IPFS (4:40), Near Field Communication (NFC) for mobile (8:23), improved offline web hosting via DNSLink on IPFS (14:20), and whether one can build a searchable database on top of IPFS (25:36).

Learn more about the monthly calls of the IPFS Local Offline Collaboration Special Interest Group: https://github.com/ipfs/local-offline-collab/issues/25

For more information on IPFS
- visit the project website: https://ipfs.io
- or follow IPFS on Twitter: https://twitter.com/IPFS

Sign up to get IPFS news, including releases, ecosystem updates, and community announcements in your inbox, each Tuesday: http://eepurl.com/gL2Pi5

A

Hey folks, this is Terry we're here with the local and offline collaboration call of ipfs, not as not as super strict agenda this week, but here to chat about whatever is on people's minds in this space. So I threw a couple of links in our agenda to articles that come to mind.

A

These are all from the offline camp, medium publication, so one that was just posted yesterday is from RJ Steiner and it's about tangerine and they they basically replaced a process and I think in Africa, so in state a collection for surveys from an educational serial, and there were literal, like truckloads of paper surveys being dealt with originally and they used CouchDB and pouch DB, which are more traditional offline, first tools to transform that.

A

But at the end of the article he mentions that this works really well up to a point like collecting collecting data offline and then, when you have a connection syncing it up so that everybody can see the same data is what most offline first apps do and tools like how she the impose TV are built to do, but when they have people in the same school as each other on the ground, trying to get everyone set up with the technology to start with as problematic, and that's where a decentralized tool where I could get my get my version of the app installed from my tools version who's sitting next to me, instead of from a network connection, so he kind of highlights that gap like so much you can do with traditional offline for solutions, but there are little pieces where the decentralized web approach would be more helpful.

A

So that's one in kind of the data collection vein. Certainly, medical data collection is hugely important here too. So the article that's up above on fighting Ebola with JavaScript's. This is also a couch TV and pop TV. One I'm not sure that there's a decentralized angle to that, but I, a lot of the people who have come to offline camp have used how she be empowered. She be when they were fighting the Ebola crisis, so certainly having health professionals able to quickly see data from cases and other places is really important.

A

Bet. You can't guess I'm thinking about that right now and then there's an article there about the response to there was some severe flooding in India. So this one does focus more on some decentralized solutions to communications. I think maybe there's various like peer to peer and mesh networking solutions that people have used to kind of replace cell phone service or how, whatever other messaging platforms they would.

A

They would normally use to let each other know they're safe, because even if there is cell phone service, it might get completely overwhelmed during an emergency when everyone's trying to use it um and then the other thing is just like huge data sets for research, for example, which I could see also relating to the medical field, so max Ogden. Who is one of the creators of that?

A

As in this podcast episode that we that we recorded at offlane camp talking about how how data is structured specifically in a way that was good for those scientific data sets.

A

Keeping them online in a place where you know research communities might not have them the money or whatever resources they need to keep everything up to date, but you have to like to verify other people's work. My understanding is, you need the original underlying data. You have access to it to verify it so, and those are just a few ideas that I threw in there some things that seem a little bit relevant right now and there are other things if you go to the offline camp.

A

Medium publication, there's a tab specifically about decentralized web stuff, where you can find more on mesh networking and things like that, there's one on developing worlds. That also tends to have a lot of the interesting articles about you know, Medical, Response, etc. Hospital run is another app, that's really cool in the medical front, so those are some of the things that I'm thinking about I.

A

Don't know why I don't know if you have like within the ipfs scope, if you've seen things being used for like medical data collection or Disaster Response, or any of that I.

B

B

C

So I'm from the medical imaging space been doing. Oh.

A

C

Yeah and so I know a lot about healthcare data and specifically medical imaging, but I, just parse in at the last 80 that comes from Craig marks when you're going handle it's much easier for everyone.

C

Yes, I just met him for the first time at the desktop meeting prior this one but yeah. You know I been in medical building 20 years and left my position as CTO back in December to find the next thing and kind of stumble upon ipfs I think when it comes to medical imaging, taking a constant addressable model, it solves actually every problem. We have there's a lot of problems. We have within medical, imaging I. Think what interesting things is that medical data is inherently immutable.

C

You really want to destroy anything that you you capture and, in fact, from medical legal point of view. You often need to to keep track, for example, medical images you need to keep in the US. You need to keep track of them for eight years at a minimum. If it's an oncology, patient or pediatric patient need to keep it the life, the patient, so the data really lasts a long time and there it's also quite large, Lisa medical, imaging use.

C

It think a lot of the other healthcare data isn't as big but with a medical imaging. It's huge benefits if we can employ take advantage of the deduplication capabilities in so so yeah I know a lot about it, be happy to share with anyone and I've, been personally working on a project to take medical images in sort of an IP FS.

C

Actually, I have an IP LD data model on building industry, an IP, FS and I'm able to import data, export it back out in the standard format and view it using a web browser using JSP, FS so doing a lot of cool stuff there. But when it comes to offline I know that this is the meeting here.

C

You know, because healthcare is intrinsically immutable. It does make sense to be able to collect it on a local device and even cache. It is probably another important thing having the most recent version of something is obviously always ideal, but you know I'd say they kind of the rate at which the data changes is fairly slow and so even having access to something. That's a little bit old few seconds minutes or even days can be hugely helpful.

C

You know to a delivering patient care and, of course, the peer-to-peer aspect of this is also really really valuable, because, especially in rural communities through country, it's just be not have the infrastructure and so being able to continue to operate. Having some data is better than no data, so I think there's a lot of interesting things into healthcare, and now it's our current crisis, you know, is something I think this is kind of a good time to take this technology and move forward quickly. To me.

A

Oh yeah, that's really cool. Do you have? Is the project that you're working on now in a spot? That's public like a repo people, could check out if they're interested or there any links we throw. In the note, sir, it's fine if.

C

A

Not just asking it's.

C

Not my goals that actually make it all open-source, but it's currently private how's it trying to be I'm just trying to figure out how to do this right.

A

Cool that sounds like a really neat project.

A

um Why don't tell us about this link, you added or anything you want to talk about web NFC for mobile near-field communication right. It's.

B

D

From the other call so I'm I'm a conduit for sharing links, someone has sent to me so.

B

A

B

So, like Chrome 81 shaped an original trial of this I believe so original trial is basically a like an alphabet of three. You always like behind sort of like a flag. You need to either register and your website needs to send additional header with tokens of the browser knows that you are fine with enabling this experimental technology on your website or as a user you're fine with enabling it globally. It depends, depending on the techno, the new web api stat, they add or experiment with in this particular case, it's weapon.

B

Fc I did not like actually read the entire article, because I just jumped into this call, but why it's interesting for offline and local? Oh well, you may have two phones without any connection to each other. There may be no internet, no mobile reception. You know that both of you are using IP FS or the p2p no nodes in your apps or you have it installed on on the system. But still how do you like connect to each other? You you both can alike an able do no tooth, but still you don't know clear.

B

Id period is that this, like long string of characters, so this could be a way for the p2p signaling without any like centralized service, we've been thinking about, maybe web being the transport or like Wi-Fi direct like setting ad-hoc hotspot for situation when you want to like connect to each other directly in in those situations, still there's the remains the problem of discovery we may have the transport we may have means of connecting to each other, but unless we know the address and the identity of the other person, we are not able to do that in security.

B

This is something that could I'm, not sure if it will like involuntary transfer, but I feel it could at least serve as an option for signaling, so one up or one phone could expose a manifest with its own addresses, including local ones, maybe local hotspot, if nothing else is available and the other person would simply like scan that phone and be able to with significant degree of confidence to connect to that specific node and and get bootstrap from there. So it's still unclear probably requires going through the API, similar in similar fashion.

B

As we did for web forward lot of, but it's really exciting, that this capability could be available on regular website at some point, this means progressive web apps, which are designed to work in offline mode, could also support this. Like ad hoc pairing or discovery between people, it will I'm sure people will come up with very exciting ideas for like the easiest ones, are like multiplayer games when you've got quickly or when you want to like quickly share each other's like details. Somehow, but I'm really ecstatic about this like distributed decentralized.

B

That's that's, probably the most important part that it's like a way of doing the decentralized signaling on mobile with just a pro web browser. So that's why I additive to the list, mostly to put it out there? It's just like any original trial and it's unclear what's fully possible with those API spots. It's pretty encouraging that pretty encouraging fact that he it will at some point landing chromium and chromium, has over 80 percent of market. So the moment that lands most of people will be able to use it, especially on mobile right.

A

Thanks I, just completely spaced out on taking notes so feel free to add. Oh.

B

A

A

Neat yeah: what else do people want to chat about? What's on your minds in the offline space other stuff we've seen in ipfs lands, that's interesting for local collaboration or things we're seeing people able to do in general for local collaboration that we can't yet do in ipfs. Why not? We have to change to make it work. I.

B

Have a sort of related thing for an offline, so people are using ipfs for website hosting and they set up DNS link. So so they set up DNS link and basically the DNS name points at some HTTP server, that HTTP server happens to be a diva's gateway and that way they get a pretty nice hostname in the location bar. So, for example, if you go to English, Wikipedia mirror.

B

You can load it as a like a regular website. You can also load it from HTTP gateway. The fact is, it's. The same gateway is the same server. It's just like the way it's representative location bar, so the ongoing problem was it's. Nice I can load it from some HTTP server, which is somewhere, but the moment I load it from my local node. It's no longer like this pretty domain in front of URL. It looks more like this.

B

It's like my local, my local, IP right and then I P&S and then there's this net domain name, someone picked or peer ID, which, if you don't want to rely on DNS. So what will land in the next version of Goa TFS is something we call subdomain gateways.

B

There are subdomain gateways right now, but those are using custom configuration at the enjoy next level. Now we will support.

B

Gateway subdomain gateways and HTTP proxy mode directly in Goa TFS, and what that means there will be a blog post, but for now, and specifically for this- for this call is the way the nesting websites are handled bar by actavis, compile and browser extension are bit tricky. We redirect them to local a a qfs node, so they work in offline mode.

B

However, you'll lose this security isolation right now and people are not able to use more advanced ap web api is because they are hopefully that there's single sandbox for all websites loaded that way and what will change it will be that with this next release, we will be able to do something like this, so the original name or content, identifier or pure ID will be here and then there will be a pianist or ipfs sub domain on the localhost. So it's a bit better for user experience. You get like the name of the content root.

B

You are actually interested in up front and from that this entire host name will build an origin, isolation, power content root, so each website will be isolated and having that we will be able to like redirect all the websites which have DNS linked to local node without decreasing security isolation, and that, hopefully, will make those DNS link websites much more useful in offline environments.

B

First right now, if people are using a website which is backed by DNS link, and they want to do something more than just like read text, they usually need to disable that redirect so that they don't get a different security.

B

Current is right, so I believe that will be a pretty critical long story short. We will increase the surface of cases when we can load content locally from my local node. Instead of fetching it from arbitrary servers on the internet, which is like that plus added value for offline use cases.

D

There will be a blog post, I promise, yeah I explained.

B

A

Feel free to edit my notes on that section, which I'm sure a little gobbledygook Oh.

A

What else we can either chat about stuff now that people are thinking about or if there's something you'd really like to hear about on a future call a use case or company you've heard about that. We should try to get a speaker from you can do some pre-planning.

B

In sort of discussing website publishing and content delivery networks, we have people from terminal terminal that code.

A

B

I know they're in super early stage, but some spaces they are exploring our distributed. Content like distributed CD ends, which sort of like leverage- let's say they leverage JSI PFS running in the browser like a swarm of such nodes, running the browser, instead of something like cloud flower where or other CDN provider which it's still like it's the century, it's like federated within that organization, but it's still under the control of a single entity which decides which content is blocked or not.

B

Probably too soon right now, but at some point it will be super useful to invite them to this code. I believe they were recently on I give us All, Hands call and they gave a demo of integration with github. So, just like you've got to keep your pages, you would simply part push your website and gets published.

B

They showed a variation which is super sleek and is publishing type device with DNS link, so that, like connects nicely to what I mentioned before by default, there's like a DNS link and if someone has like, if its companion and note it will upgrade people to to that fashion, nothing else comes to mind, I'm, afraid, okay,.

A

Chris, how about you anything in local offline collaboration space that you wish you knew more about sort would like to chat more about I.

C

Don't have a specific question right now: alright,.

A

um Yeah, if you get to a point with your project, where you're interested in giving us a demo of it, we have these calls monthly, so feel free to reach out. We could certainly schedule to do that. Okay, awesome um cool! If we don't have other specific stuff to talk about I'm, happy to call it a day, um let alone gonna leave these notes here for a minute for you to clean up my my butchering and then I will share the notes and video and such with everyone later on.

A

But thanks thanks for showing up I hope everybody is healthy and hunkered down as appropriate in their region. A sense stayed sane amidst a little chaos and an unknown here. So it's nice to see you.

A

C

Terms of this meeting, like what would you say.

C

So offline club, how does I TFS kind of view that or does it deserve a team on that or has a weave into ipfs exactly.

A

So this is the meaning of a significant interest group about that topic, um but I think I think it's people who are coming some of the people who come to this meetings, who are more engaged in, like our collaboration, space, which includes Lytle and Dietrich, etc. Are the folks within ipfs who can kind of grab the ideas and think about how they apply and how we can improve ipfs for certain use cases, I'm, afraid I'm, not the best.

A

At doing that, you might be able to give a better a better answer to the question than I. Can it's.

B

A pretty good question, good answer, pretty good question. Everything is right, yeah I think. The main purpose here is to like exchange notes. Some people are more technical and have like very low-level understanding of ipfs and hippie-dippy stock.

B

Other folks are very, they resonate with needs of people in offline environment or places where the internet connectivity is not that the best and actually it's a very important to have a place like this poor folks meet and exchange notes even to exchange links, just be aware that there's one need and that there is or will be a technical possibility to do something with that. So that would be that. That's how I understand yeah.

A

And we have so that the notes that are a gen RN has all of our meeting notes and we have these recorded calls online. So if anyone is seeing us for the first time on recording this or Chris, if you haven't caught past past meetings, you can catch up on those.

A

If you want do to see the kind of variety of topics that we've had will often have a demo of an interesting new app or someone talking about a particular particular use case where they're not using IP FS right now, but we've had people chatting about. You know needs of Native, American communities or various groups like that.

A

So it's mostly a discussion space, but I certainly hope that it is sparking actionable ideas in some folks and that they'll come back and show us the things that they've made so yeah yeah.

C

I think this kind of high-level brainstorm right yeah, but you know for my project- are not with, is just taking the data and kind of burning and IPL D I best Astoria yeah, then kind of next thing would be like how do I actually find it? You, usually you have some kind of database that you can search to find data and that led me to orbit EB, which kind of has some offline capabilities. It looks like and some it's a lightweight database.

C

It only scales so much, but the thing that's kind of current to me is that seems like in it's almost like with immutable data, and you kind of have almost everything is offline to begin with, because you're going to store it to one node and then you know it's immutable, so it's like someone can change it. So really, it seems like the whole ipfs world is offline first, because there is no central database, there's no central thing to store, and so this kind of has a broad crust cross-cutting.

C

You know aspect we respect its inherent in ipfs and immutable storage of the game. It's everything kind of begins on one node then spreads out from there, and there is no central thing. Yeah.

A

I should just maybe clarify that the term offline first as coined by the folks at neighborhoodie. Originally it doesn't mean that I guess technically all of our data originates offline. We type it somewhere on some device, but it doesn't mean that it.

A

It doesn't mean that we first access. The thing offline, like the most common use case, is that we have that we do have a central database, but that we're making the vast functionality the most important functionality of our app work offline on the individual devices.

A

So in an example like pouch, DB and CouchDB, pouch TB is your in browser database and you can be writing data write a native writing data offline and then, once you have a connection, you would sync it to CouchDB in the cloud where you have something central that people are sharing and sync it up again. So that's it's there's a bit of a misnomer there in the first part, I think of it as putting the user experience first by making them have access what they the most important stuff offline.

A

There's. Also, a term floating around now called local first, which I take to be a bit more in the specifically decentralized scope. The centralization is a very as a small niche portion of the broader offline first audience so for whatever that's worth.

C

It's interesting to think about. You know, I, think, as this offline space grows, I think I suspect we're likely gonna find ourself thinking. Let's write all of our applications this way, or at least as much as we can as opposed to right now, I think, as you start a new web app you and we have a database app server and then your front-end and your databases guide your centralization piece, I. Think as we develop more knowledge and success in this offline space, I think it'll shift to be. Let's do them again, offline first and I.

C

Think, like I said as I came to it, I was originally planning on using I've kept us with some kind of database and where I'm at right now I was like. Is there some way? I can use ipfs as the database I? Could you know use that for everything, so it's completely decentralized but completely work offline and you could sync up afterwards. It seems like it'll, be a specific to my use case. In fact, it's kind of foreign to fine use case for the most part, but it just seems like a good architectural strategy.

C

If you can't do it.

A

Lytle, what's your initial response to to someone wanting to use ipfs like a database I feel like there would be like a gut reaction from super IP FSC people.

B

It's better if you're friendly, it's a key/value star, ski and.

B

No I I feel it's it's yeah. It's said that the base in that sense and a lot of people just need a way of putting something and getting like references to it. It's effectively glorified. He value store for many people who you sacrifice in production and I've, seen people who simply expose just like put put like, add and cut or put and get API methods as soon as they would exposed some end point of CouchDB or something like that.

B

I just feel it's more it's more database and like framing it with just just putting it that one box sort of like blocks people from leveraging the fact that you don't need to care about date. The way data is moved. It's super super important for folks, even like people who care about like docker images in in companies. Where was super painful process of making sure those images are synchronized and available when you need them and if I DFS we just referred to that using CID, and it's it just fetched from the network.

B

You don't really care from where. Similarly, from with databases, you don't no longer likely to care about like maintenance by direction. Multi-Directional synchronization, I years back I worked with CouchDB and it was like super useful. That's it had bi-directional synchronization in ipfs you effectively, you don't really like. You've got like just in time. Synchronization. If you want. You just share those like references to that data share a substantial data, so it's like much more much more flexible than the regular data base.

B

That's what while it's like grabs me, but mostly because it may close some some potential use cases if we start talking about IP fastest of the database replacement than just like my.

E

Personal pet peeve yeah. Well, let me clarify so.

C

I think ipfs right now is primarily just I. Think for storing files, replicating I. Think when you go to something like IPL D, which allows you to build. Essentially a data model, a Content restful data model. You can actually model a database, you know with it. If you think about a database, you have a bunch of everything, begins with an empty databases and there's a lot of transactions that go against it, that are private. Each transaction gets written out to ipfs. It's a separate block.

C

You, you know you have the same, coming up the structure available as a database. The biggest issue is: how do you kind of find the head the head note of what the current state database is or the head head note of the transaction tomorrow and you know I, think I'll orbit DB does use thoughts of the kind of broadcast I mean IPS could kind of, but I don't think it's probably not designed to be done to be update that quickly.

C

But the point is, is you can store data and you can get it and you can build a log, a chain of things? It's all in the trinsic, so yeah I give. This is not a database file to say you could build a database on top of it using ipfs right, kill me as your primitive okay.

B

Totally and with stuff like stuff, that's on top of that you, if you like leverage things like see our duties, you could get that conflict resolution without like centralized points of failure, which is usually present in most of stacks. People have in one form or another.

C

Yeah I think it. You know, you're right, I mean how many of us have dealt with building highly available solutions at the database level in the application layer in key, because it's kind of built in to not even be out of date. If you're not connected, it's kind of a little disconnected comes in but start thinking about, can I go offline first and with your application, it can solve a lot of those problems. I think.

B

Yeah, totally and I feel also like nothing like black and white context. Even if context where you have some connectivity, but you it might be very spotty or you don't have it all the time IP fuss with stuff like I'm, not sure if graph sync landed, maybe I should not mention it, but but being able to selectively sink just like a subset of a graph that, with with the whole thing that we just mentioned absolutely ifs, yeah totally.

C

Yeah yeah I've been looking at a graph sink and I've actually been more engaged in the IPL D communities, so I've spun up on the selectors and schemas and graph sink so yeah I think there's a lot of a lot of power in there about it as a modeling, your what you normally model a database, whether it's mando or a document or a database or a sequel based database. You think about that as an IPL media structure.

C

It's pretty amazing, so kind of blowing my mind of all the Bela, but you can do your. You know. They'll do the features this gets built out.

D

Yeah I always have this feeling that of like I can see the shape of things.

B

To come, but it's like always so much needs to happen before we are there on you just like lower level lower levels, so yeah we'll get there eventually, but it's super like like it. It's always super encouraging to find that people like share the vision.

A

B

Did you like in the context of IP and did I break or no you're fine, really go ahead? Okay, Chris. Did you like in the context of IP on the exploration? Did you use your URLs in any context like for the ideas or no yeah? We are like trying to understand potential use cases and semantic way like Semantic. Web is one of places where your ends are used and we probably want to register ipfs for those like mutable references and, like you are eyes to our ends.

B

Those are the standards that we need to follow if we want to increase adoption. So it's just curious. Okay,.

A

So sorry, will you just define uri nur n for people who haven't heard them.

B

Yeah, so your eyes are Universal resource identifiers, an example of that is URL, which is Universal resource location. So you else a subset of your eyes and this basically a way of addressing things, a formalized way of addressing things and your ends are used in context where, like location, do not matter.

A

What's the understand for.

B

Notation I believe one second I I had open.

B

Let me quickly share much because I got the examples. Yes, so like it's, a uniform resource, name I lied before, but it's a uniform resource name in the Uniform Resource identifier has ISBN number right. So if you want to refer to some neck, assuming there's like a namespace of other identifiers- and you want to address that, you need a way of addressing different namespaces in like uniform way. So that's why it has like uniform and yeah books, RFC's and other stuff.

B

We probably if people want to let's say, there's a scientific paper and someone wants to attach or like refer to a data set or the tools they used for creating that paper or doing the research if they use content, addressing that's pretty useful, but often there's a certain level of formalization needed, depending on the context. Some like Semantic Web projects require they simply standardize on your using humor and everywhere. So that way they can use all the sub namespaces in the URL and be more confident about what they are parsing.

B

If they let like user to enter that identifier, it's basically like a formal formalized way of addressing stuff. Without providing location in web browsers, you don't see your URLs, but you may see them in like more specialized applications and contexts.

B

When I hear IPL D and people trying to be like data structures, I've seen often people experimenting with the IDs and in the IDS I believe they either have a ID protocol scheme or they use your hands. That's.

E

C

It I was just like curious. Well, you know, I have not heard of users. I've only been going to the team meeting staff for about month, Oh what they did come up with this week's team meeting here, Volker BMX, something he some kind of open. Oh gee, see what is this open Geospatial some kind of mapping thing they're about to become their standard. It wants to get see IDs or at work with my PFS, and so that may be a good using you, Lorenz and stereos and see IDs.

C

Yes, want me to come up this week. That is related. What you're talking about oh yeah.

B

Totally yeah yeah, that's a very, very good example of that yeah.

C

Yeah he's like yeah they're, only updated every 15 years, and so now he wants to try and squeak it in somehow so we'll work with IP LD correctly yeah.

B

Totally yeah and that that's sort of like why it's important a lot a lot of time. Those standards either exist for decades or once something in standardized it will be present for decades. So it's important to make it easier for people to use ipfs, intox context, yeah super cool.

A

I'm, just looking at the calendar here for April, it looks like April 15th we're scheduled for a next call and I think we're good to not have it conflict with the GUI yep. We're good cool. Well, thanks for coming everybody as always happy to continue any chats in the repo. If you want to and I will once I get these notes added and the video up I will respond to there's an issue for the call itself so feel free to subscribe to that.

A

If you want to make sure you get updates on upcoming calls um I just paste the, since you heard about this elsewhere. Let me taste that for you Chris, but that's alright, you'll see updates about this Cup and that's the repo we use. So if you want to start a discussion about anything, feel free to do so in an issue there um yeah great to see everyone, and we will see you on April 15th. If you have discussion ideas, demo, ideas, someone, you think should be speaking on the call.

A

Certainly let us know in that thread and we can get someone scheduled. Alright stay safe. Everybody have a great month until we see you again, bye.