IPFS Learning Resources, 8 Jan 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: How IPFS Works - Steven Allen

Description

In his introduction to how IPFS works (approximately) Steven Allen gives us an overview of IPFS and what you need to know. As a core developer of IPFS, Steven has been a long-time contributor to the project.

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

Okay, so this is gonna, be about approximately how it would just work so say approximately, because nothing.

A

Okay, so, first of all your reviews, I give us you probably see in this screen. Let's say: you're a pilot. A professor, just runs command, get very, very end. You see this. They call to see I need to contact points to file you've, probably seen it in a path with this we're gonna gate, we're URL, you didn't get files by reading again command.

A

Okay, this hawk is going to cover how we import files, name files, find files and then fetch files. The network, when your important file is there a bunch of steps. One is to chunk the file in smaller pieces. One is then take these pieces and like actually compiling files and directories and then find that we encode these files and Directory objects in a data storage system called IP LD. Then, when naming.

B

A

Use GNU and continent needs to actually refer to the pieces of data. We use paths to describe extra metadata about this, about these user data, we're dressing as if we have another simplified DNS for immutable names. So a cover this bit later, but you know because all content is addressed by cache, and that means you can't change the content. So we could system for energy like we.

B

Do do a new piece.

A

Of content and everyone change, the contents would usually go copy up right next, you actually find the content in this involves some overcrowding or content routing. We use it the used to eat specifically academia. Finally, you need to fetch it and for that to be cycled, that's wrong actually, before we move on using these three logos at the bottom. Basically, everything on this summing so on your left side of the screen is ideal. The areas like that's like data structure system, we use everything on the right side.

A

You just look a tepee, that's the networking system and then the bottom. That's the whole perform logo. All these system, these multi parts and glue themselves together point out how these things work, presentation, progresses. Okay, the first thing we do is we chunk the edge, so they start with a continuous final and a very small pieces. We did this for a deduplication to allow for piecewise transfer and Duloc we're seeking. Oh.

A

And also each chunk is occurs. Okay, first of all, T duplication. You can import a file. We may end up seeing that paly two pieces are the same piece, its case that you want to be super same with there, because we took this like this. We can actually just not store the second piece unless or the Emergencies at once.

A

Second whole piece wise transfer, I can keep on fetching pieces and then immediately notice, but that's the entire thing that this piece is back, so they can go, goes up here and actually find the new copy that he's the correct piece if I just have to like fetch the entire thing then affecting everything else. Oh, this is not for a thing, throw it away. That's another entire large file, your side! No, that's not the right thing, throw it away this case.

A

I can give them like each little piece and it's like learn early, whether or not the times. The right thing finally chunking allows is to seek in a file. So it means instead of us having to download the entire file use a teaspoon want. We can just as if, like hassle these two don't care about and find a piece. We want okay, UNIX effects. So now we have this file and it's been chomped into little pieces, but that doesn't really give us a file system.

A

That's the first thing we need to do is create and actually continue, this file a lot of these pieces that you've just broken up for that we create a single unit, those file, it's actually it's a purple tree where you have nodes in Basin, zelich, well, bytes 0 through 20, are on this side, in this case the bytes here through 200 on the side 0 through 200 to 350 on the side and then again, there's your 100 100 through 200, etc.

A

A miracle tree is just the tree of hashes, pointed at that point to other ideas. I use Akash's really be more with them, but probably using the word here: Ian, because that's how their was built on trees, I.

A

Actually like them, these are trivial, tubercle, jagged. It's imperfect in a cyclic graph. So remember how many can you duplicate data by chunking? We don't actually store the tree like this. Instead, we just like store one copy of the large blog and then like the intermediate.

A

This one over here actually points back to the original worth blog for the boys.

A

You'll see this data go over the place, it's confusing to people but like we do call this out, because a tree can't have it have two two pointers at a point, the same thing or two I. It just pointed same thing, a dagger to write this graph account.

A

Okay, now that we've create the file, we need to action, video directory. We do this by this mapping file names, two files, again music actions, verbal links so appreciate. For. Finally, we need to actually look at Steve greatest logical structure using hashes and bobs of data and data structures. We can actually encoded somehow transfer Network sort on disk talk about it, so my god for that musical ability.

A

I know these come about me, metaphor, understand a encoding marketing data structures we use in that BFS, but we also have a can use I, peeled a to understand. It gets interfering because it also murkland data structures, basically I, believe it's a verb, typing or hadn't been a metadata to perfectly the type of is like Haley. This is a gift time. This is an this. Is the nithy ring Doug and like, as you use, I really didn't walk these data structures. You remember, though, I'm like this is the tears are for me.

A

This is the context of in so you can like remember how to and find the next piece or understand the different pieces. So just a diamond a bit more to what level he is. The name comes from in Japan greatly data. So if you're familiar with massive linked data is actually originally a case and vertically.

A

But the original concept was to use the URLs like you'd, have some decent object and it completing some of the j-stop.

B

There's no URL.

A

And that other you could put another just are confusing the problem with this is that URLs have an authority and then authority in this case a comm can change what it wants to. You don't want that. We don't want to trust anyone so ideally uses fashions. So, instead of having to be rel usually equals an action hash points into the Roger, it's also a mutable.

A

Okay. Now we finished up with internal data structures banana chocolate. How do we actually gain things? The system? So we now have all these like objects that are restored. My ability I may need to be able to talk about them and pass the round the network, but if you sign called cm IDs for content, identifiers, loosing, two examples of see.

B

Enemies there, the first one is a.

A

Cmu public used to the second one is the sort of new style of cid that we aren't using very frequently yet yeah. Actually, we use.

B

It for content dressing that.

A

A

Basically, it's just enhanced with a bit of extra minute, and this is how many every single piece of okay. First of all, why do we use dashes? Oh you heard: why do you see if he's in half is like they're, a verifiable immutable and trustless permanent so like if I haven't see it in point? Somebody it'll always point to that same thing. You might have know people who find it.

A

So this is whenever I say permanent I don't mean like the data is always gonna, be there I just mean it like my name for that data is always gonna point to the same data and like the data can't change someone.

B

Else can't just like replace it's like. Oh, this is the thing you wanted. No.

A

I can tell exactly what I wanted if I got back the birthday. So that means like this is a picture and if I'm, a pretty good picture, weird and good by the same picture, I don't get division. My other cat, who's, less friendly, more friendly, okay person coming back is content dressing, person's location, aggressive, okay, so content identifiers address content by content, not by where it is, is it genders, but what it is not where it is. This is important for ifs, because then it could be any more networking fashion for men.

A

Here's you don't to say trust them yeah both now we're gonna do what exactly. That is what the distinction is. Okay, sure dressing, I can say: I can't cause. He isn't there get. Someone happened say go up into my account content dressing. This is my god. That's really difference. We're computer system describes thing you're. Looking for okay, she doesn't tell us what we're finding okay.

A

So what was the problem occasion? Rest well like I, can say, go here if I might have. But what if you find this guy, you don't know, that's not the right cap, because it will define the cap right there. You found the cat. You give me the cat. You gave me the wrong cap important. You can't know that. So that's what we use.

B

Look at your content dressing because, like if.

A

I just give you the picture of my camera. Then you can look at the picture. The cab of the car give me the credit cap. You know that you're actually giving you the red cap.

A

Second aggression a bit of metadata this one we will skip through pretty quickly because it's it has a lot of details. This is multiple answer. So remember how I said that CI neocon identifier is 1/2. Wasn't that metadata?

A

This is section working discuss little bit. First of all, we have two types of schedules. We have C 0, C, 2, D 1. You don't have to get too much into this, but basically CDP's there is the old form of cid so that when they actually use most frequently at the moment see you baby, one is new form of CID. It has even formatted. It could really didn't. Have nothing to give me 0. Db 0 has only one piece of in it and that is C.

B

They deserve what.

A

They are suppose to stick with that, promote they actually just a special hash ever think of this type of hash them use. Then it wasn't have to digest and something about the most like to enact did no, it's the type of attachment used to produce an heir scientist, see adv. One is a multi base. Encoded I feel before mat. Molding are stable, ok for full of addition just covered this whole dance. It's all describing maps.

B

I, just basically.

A

And give you the hashed.

B

I, just and b5 output actually use.

A

That so shop sits out but I and then, like a number, it's a multi phone arc to indicate which function we use this way like I can cite you shot, 3 or shop, 3 or.

A

Stop whatever they want to in the future and I think you'll know which one I'm using because I concluded in the hash itself. But your talk was like a multi couldn't. Did you explain it is that a multi good at using non-metric network so often are based on these new magic numbers did like are used to indicate various things. Various file types like you ever sorry, you're, trying to figure out what type of file something is on your desk windows uses extensions, but actually like Mac, OS and Linux.

A

These single magic numbers where there was like reefer she bites the file and see it like the bite unit. First bites match someone known prefix for some pop to the bottom, but fortunately there's no tipping sorry. There are tables these, but they aren't like decided to have time to just like sort of random earth.

A

People just hope that these magic number so much to each other and a lot of times. It's just like a lot.

B

A

Of all, basically multi codec is trying to avoid this, where basically, we've created this massive table of non magic numbers that we are pre allocating and like keeping track of so the legacy for every single concept we have, we can just assign a number. So in our case like we have numbers.

B

For multi hashes for multi matter or addressing systems, we.

A

Have codecs for our civilzation formats, everything so, basically, whatever we can with some new concept that we want to like give a short name to. We just create a number stick at the table. We're done okay, final piece here all day base to get deep into this.

A

So basically, if we go back to see any p0, if you look at Z to be 0 or always.

B

Starts with Q M, something find something. You also notice that in capitals.

A

And Lord is there. This is not level browsers and we realize that. Actually you know what certain contexts will need to use. One basically put in the context. We need to use another base encoding, but we need to actually able to tell which basically we're using. So this is revolt against Canada kind of like multinationals of the format's or all the moments. We use the prefix to tell you what type of basically music. So in this case, if you see a premise on you'll know, it is encoded using face 32.

A

If you see a Z you'll know to put it using Facebook da. If you see a few loads, you could use base 16. You also may notice that McBee is a valid character in basically 2z is about character and episode. Character convey 16. So basically, these are just social, describing basic code instruments.

A

Okay, actually, this is the final one. Finally, I Billy I say so remember in CID who wanted how I said that it's a multi based encoded. Now we use the boulder is prefix and then some basic code of thing to both my ability, format and multi hash. The ability for map tells us how we included the data that the function so the whole thing I told us like which data we're talking about this especially content.

A

They I believe format as part of this tells us that how to understand the D that we are wanting to that's it. It's like we have one for some objects. One get object over here blocks at your transactions by Kiehl's s files, all these different formats.

A

Okay, that was a lot of details. You fell asleep. Please bake back up, move on the only reason I brought. That up is like those are a lot of really cool part. Is you should look at okay, so you may notice that is use? Has we don't use.

B

Urls, most part of them say like something.

A

Colon slash slash something else or something : something else. We don't do that because, like URLs passed, a namespace.

B

But unlike hero's, transform.

A

Cursive so notice, then, on the right hand, side we have a / TX left, github, comm port number 5's state's, not kidding. This is a way to describe you get URL, for.

A

The second version is how you actually describe this thing normal that using New York dollar, but in case you use it kind of you have to give this a get + ssh, because you can't compose multiple verticals. That path was actually composed protocols. So this is why he is not. This has been a very contentious issues if I bring it up, but we use them because they want to go up like a bed protocol with the protocol protocol run protocol protocol protocol.

A

But to do this with a URL you'd have to like create these long strings like X, plus y plus Z, and then like I, visited forum for every single combination.

A

Okay, the final part of naming is I minutes. This is actually our naming system. Ie an S stands for the interplanetary news. It's so I Venus allows you to map the public key towpath and then way this works is basically taking your the packet right amount to, and you take your public key and a private key. And then you use your your private key to sign the Matthew. What you do, then these publish the little fella hey. This is like this is the kind of this publicly currently wants to.

A

Then, if you want to update it so I guess updating you just signing the record and Lisa. You actually put additional sequence numbers and like this is version 2. You know you can say this is 3 per support, etc. Finally, this initial point of arbitrate paths so I can use my pianist point that I made us I. Think if you're a user to point to swarm I could use it to point to you think anytime they wanted. So it's not assisting specific as Jennifer was cool.

A

If you've ever used or copies a Guinness needs it it's slow, but it doesn't work I'm sorry. This is actually one with a DHT. It's a system we currently use for resolving I can't as bad as we're working on it, proving that it's not an inherent issue. Happiness itself.

A

Ok, now that we've covered how to import files, how to name files finally gets talked about the.

B

Design equation.

A

Now I'm going to actually find the file and then download zip'. First of all, probably finding the file we use a sudden, isn't called gravity content routing. Basically, we take a CIE and into translocation risk. So we talked all about a highlight.

B

We don't use locate your asses. Well, that's not actually true. We actually have to use chicken in case you know stairs.

A

Find the content, but we start with a content, address and then use a grounding system to do what to work, to actually find out the set of people that currently have so because.

B

We're using content dresses, it means that this set can change. It also needs to know what the trust.

A

Agents so we'll find is that people that claim to have a file and they'll pass them to the file and we can verify they give it back by 5:00. Okay. So how do we actually do this? Well, we keep running it. It's quite simple: we just have a.

B

Massive table of.

A

This pure has this thing: the spirit, business paintbrushes. This thing is held by this set of years. This thing is held by interviews so I'm trying to find. In this case you have art, takes the ID. I was right.

B

Kid, let's say: okay, who has it and it says.

A

It's a bit of a problem in this up: you're, pretty big, a single peer can't easily hold this Brown table, maybe at the moment they could buy in the future. They won't be able to do that, so we have to figure out like how do we do this Master Brown, cable them, where we have to store to do that, we're using all those distributed hash table and you take the router table and distribute it amongst all peers and efforts. So, instead of copying, like everyone know, who has one body or my handsome no.

A

Problem here, how do you know who's? What piece around it so Mike? Yes, I can take this right. A bit looking breaking pieces. Give it up so.

B

A

But now, like you, have the same problem where you don't know where to.

B

A

Sir, before you know, who do that are content.

B

A

You don't know who the grass, before the beast around you, we need to find that piece of pair of the continent, so we solve this using CAD Emily. Now this is pretty common in your business systems.

A

Basically is tremendously she should be driving people, so dude was like giving up one piece: one person, what meets another person piece of the person you like, based on the graphing record, like based on the piece of content, you're trying to perfect they're trying to member who has what we supported based on the content.

A

Some pure honest, you find the others responsible for remembering who has that content, and you tell that peer is like.

A

Okay, so, like you have some piece of content, if you want to you find the peer there's responsible, remembering who asked you who you tell them, I have killed them, they've tried money, kill them foo, you find, if you're responsible for every us, computer and as Cuba, and they tell you some or usually a lot of people. Ok, now we're looking into torrent. How he'll know works because it's a bit complicated and you give me paper but I'm gonna, give it a quick overview of the general idea.

A

Gemma is based on two key features: a distance metric and a query algorithm. First of all, this semester is: let's.

B

Utility is this peer closer, or is this your closer to some Iza project.

A

Yes, these competency, you have a px and wife tells you, which peer is closer to the content. The second piece of content layout is occurring over. The query. Algorithm allowed me to use a distance vendor to get so the peers that are closest to the content.

A

Basically, it's kind of like coming at compass, the weights of the content, and then how are we totally gonna walk to a book. Emily is specifically, he uses a symmetric XOR of a hash of the content and the actual appear too much, but if I really eat the number did it basically hold is probably the.

B

A

Piece is the query alpha, which is pretty simple. You a suppose appears to the quantity know about which meal to go through and make it. You said, appears of closer now. You know close to here to the content and you go ask them any.

B

A

Minutes it'll be how to chase need property. Really. Everyone remembers appears of next to them and then peers early early happy for the cap refer they're. Getting have to go there again. The cool thing about this. This is the cool thing about. This is like that means like when I'm trying to find content, I can always get halfway. Close reach times, like the first top logo, probably cross, never defined appears like a closer next time. Those beers will tell me about peers that are can have my closer. Then the peers I, like those next.

B

Beers will make and it appears the closer I get it I get it again.

A

Oh, if you want.

B

To learn more about this I highly suggest.

A

You from paper you deeper goal. Definitely a paper.

C

A

It's fine, it's one! This is what we have actually finished it yeah, so we've covered. How do we actually take the dated or they top you best? How do we talk about? The data might be fascinating. Okay, fine.

B

How can actually like a.

A

Spears, but it used to get so look at soft works is using want lists. The one list is literally just.

B

A

You want so in this case one of my fantasies. He wants cue a personal space, Gillen pervert I mean what am I cats, my other cat Ozzie, once yeah couples, blood.

A

And whether you is it big of house list- and they tell everyone else, Mel's listing certain things they want looks like all they're connected, and that is the step. So now this swamp the. What lists now no easy determines what the chance awesome determines what he has easy realized very.

C

Busy has blood.

A

Food, the food she's trying to meet the Aussies I, was killing.

A

And so then they send these files its pieces, the files to each other. Then they proxies up there. One list one interesting.

B

Property but let's watch does he don't actually like it's not a request.

A

A system you don't ask for something and then get a response back. It said you tell someone else what you're looking for and then I mentioned. Then they give it back to you so quasi. There's like easy. Why I actually remember cause he still wants cuddles, so the future makes you more cuddly with locally happening. Easy may finally give us he goes, but for now she doesn't have to haul so she's a bit same with Ozzy Ozzy, making eventually decided to give your personal.

B

Space is a good idea.

A

In which case, actually personally so that was this well now move on here to become important miles, they make them finding them and find a new person to swap and then making to spit them back out. That was how I think that works.

B

A

B

A

B

See in the glass station, that's what actually goes right.

A

Below them grabs think is the next protocol for fetching data so bit swap like it's not like I tell you this at them. You said things I want you tell me some things you want was prancing I, give you a description of the subset of some marbles there gurgle tree that I want and you give them that to give back to all ones. So, probably that's all I'm trying to like walk down a factory tree I would have to ask you for this piece. You'd.

A

Give me this beast, then it's a no actually I want the last block, so I want block in the back or a machine whatever you did D so I would ask me, for this means back to me, and they say: oh I want this last piece, so Tata scale is being affected. Unfortunate means they have to do one more trip for this is one round trip for this and the whole merger for this, and it's gonna be really slow with Network.

A

Let's grab saying I can actually just describe his entire roster to all months where I can just say: hey I want the last block. This file, or what imagine two makers, that recursively.

B

A

Want the last link of this node as Merkle.

B

Tree and then the last link of its child classic this travel to find the image.

A

The records of the the final charge: that's what gossip is next question.

A

Okay, so how does grass think we're like paralysing fetching mobile gears? Okay, at the moment, the critical asian? Doesn't it just like you tell it fetched in this one here in theory. The way this will work is like you start asking one gear for sorry, some up here for something like it's up set up, dag and then you can like send them update person, actually don't care about the subset, and then you go and send that subset another pair. This is a complicated.

A

So the next step in implanted graft sent will be the combining pieces of its.

B

Walls with pieces of ram singh.

A

You make like a system, you can W choose, but your senses.