Hashing It Out
Hashing It Out

Episode 108 · 2 months ago

Hashing It Out - Pranav Maheshwari - TheGraph

ABOUT THIS EPISODE

Today Corey talks with Pranav Maheshwari from The Graph, an indexing protocol for querying networks like Ethereum and IPFS. We'll dive deep into how The Graph works, what it aims to solve, what complications it has faced as it has scaled out to its current state, and where it expects to go in the future. 

Tune in an enjoy!

Link: 

- Website: https://thegraph.com

- TheGraph Twitter: https://https://twitter.com/graphprot...

- Pranav's Twitter: https://twitter.com/impranavm_

Welcome to hashing it out a podcast,for we talk to the tech, innovators, behind blocked in infrastructure anddecentralized networks. We dive into the weeds to get at why and how peoplebuild this technology. The problems they face along the way come, listenand learn from the best in the business. You can join our ranks. Bolcom everybody episode of the Hashingit out on your host hard core petty today, would be talking with Prana fromthe grass run off. I don't you do the normal thing give everybodyintroduction as to who you are, and maybe like ten thousand foot few ofwhat the graph is sure of some. So I'm Cena, I'm a solution, Liline at a Gennote, which is the initial team behind the Gran Protocol and now one of thethree core deves building indexing and quarrying on the web, three Procol,which is the graph. That's the first. You know aspect of what I do as a partof the graph procol in general. What the graft to call is, you can say it isin simple, as we like you, don't take it step by step. Explain it! Letexplain it to me, like I'm three, when you can say it is like Google, but for the wet three world forthe blockchain world, Google indexes the web to Google Index is whatever ispresent in the traditional web. Similarly, the graph index is theblockchain, which is the future of you, know Web, which is the web three pointto what we call us explain it to me in a little moredetail, as in like Comin, seventh or standard you can say it is an indexingand quoting protocol or a different different kind of clock chains. What isindexing and queering? You can say that whatever information, that is, you knowstowed inside a particular data structure that has to be bought out ina like. You know systematic fashion and it has to be displayed before it can beshown to you in the front end of a different application right and to makethat happen to get that information which is scribbled through in the the desalinate base in a desirablerebel formant. You have to have some kind of thing. That's what Google doesfor the web to world it screams data, and then you know whenever you aresearching, it shows whatever it has already scraped. Similarly, the grabprotocol. As for the blockchain, whatever data is being Potema, youinteract with the decent les application. You interact with unison,you interact with the rearable interact with your own. All these differentdifferent awesome. Recent Lage applications. You are alwaysinteracting with smart contracts on the back end, which are entering data allthat data. As for your actions, make changes and have to be displayed in adesirable format to the front in how does that happen? That is done. Why arecreating different subgroups, which may be going in detail, but these subgraphsmake the entirety of displaying information in a decent life passion tothe different decent lives. You know applications to the front end. So it'slike it's. You can say it is the you know it. It is the layer which connectsthe blockchain data to the front. An you can say that. Then I get tin question from there likethe obvious naive question, which I'm sure you get a lot of, or you might even have fs that addressthis- is why not just use the block chain? Why do you o? Why do we needmiddleware to go from the data embedded in the blockchain to to get to the front of right? This isthis. Is this is a very like you know, cool question and you can put it in acategory like why the graph? WHAT DID...

PEOPLE DO? The graphs you know startedbuilding to turn seventeen? What blocking existed before that? What didpeople do before the graph protocol? What how was the information from theblock Chin produced to the decentlie application? You can say that that youknow before this particular you know wethe economy of Witte tack, biliteralproduced from you, know the the block chaining to the front and was therepeople used to do two approaches? The first or the most, like you know,famous, but not so much in the ethos of the wit. Three world was having acentralized server whenever people interacted with a smart contract thatinformation was listened to by a wet. You know web socket or a listener. Youknow stored in a centralized data base and whenever people requested it basedon you know, crea based on the data that was stored was produced to them onthe Front Red Right. This creates a single point of failure on. You knowthat particular decent Lizedata be on on that particular recon lifeapplication in the former centralized data base, because, whateverinformation you are saying in a decent play, Apurito might have been templedat that central point of great abuse that was created. This is the biggestissue that the graph actually identified. You know we wanted to change the way people are actually usedto interact. Second, second, you know second approach that you were alsotalking about. Why not directly use the blockchain to get the data you have tofirst of all understand the fundamentals of blockchain. It iscreated in a way to be sensor, ship resistance, as well as you know, kindof DB as or a Web, which cannot be censored, which can never be hanged, soit was created with an assumption to have sequati aretes. It has securityguarantees, while you know some other things like getting information insidethat particular block chain or getting that information outside is not thatmuch tackel, it's main Ushas or the main frontier of a blockchain is tohave security guarantees, for whatever information is, you know beinginteracted with that particular apprit, and for that reason the design patternof the blockchain is such that that it, you know, has linguist structure right.So whatever information you are interacting with is stowed in a lingisstructure in form of, like you know, boxes. If you are a literal technical,you understand that there are always boxes which are connected to oneanother. That's what can exiling the tacticals you want here so have youhave had it? Is it's a rose technical audience? So if I feel it so TURPIbring us up a some awesome, so you can say that there is lingle structuresthat make this block. Shame possible right. So if you want to get theinformation directly from the block, shame you have to first a pall havecaptivated one of the full notes. That's supposed to fall verychallenging task. If you have been running on a tedium Pul, not you willget my pleas if you're running even faster note like brs or either fallingon, do like you know, you might strike to start crying because it's so so likeyou know, it is very computer intensive right and you have to first of followthem that full node and secondly, every time some user wants to get. Theinformation that full node, which has the linkless structure of whateverblockchain is being made, has to be quarie from the genesis still thecurrent date. Just imagine I'm going to the Infortunia page today and despiteall the computation power that I have it opens tomorrow. Why? Because etedium was Pete in two thousand and fifteen every set seven seconds. A newblock is generated every seven seconds somebody is interacting with Info or toany swap so whatever input or in shop page has to load the the he the traverser or the pointer,which gets the data has to go through every block of the tedium log in to getthat particular information outside in...

...the infudit minis of page verychallenging. I would say this would make the user adoption something closeto zero and that's why we require something like the graphs, the firstapproach centralized the approach is not the correct way to make a daptright. It is a good way to make an application which works on a Thelu,don't call it the decent lie. Obsita then right. It should not have tocentralize the application that use a decentralized data I entail, so you can say that, but ifyou want to use a decent lie, application and call it a two dat todecent lies that, then, the information that it is getting from the blockchainto the world has to happen in the decent lives fashion itself and that'swhat the craft does they want to. I want to add on a little bit too. Yousaid, like I think, people misunderstand how inefficient it is to to pull datafrom the blockchain, especially like I e, if you on, I feel like like asmart contract for that instance and we'll get into kind of how how thegraph forms a layer above this to solve it. So specifically ere like in, likethe decentralization of how it does that, but like, like you, said earlier that the way this is optimized force is forsecurity and to add onto that, it's the link list structure, the block thanstructure that we use it. We'll just talk about a theorem for now isspecifically made such that you can perform distributed consensus proof of workvery efficiently and store blocks such that they can be passed around reallyquickly and then store them in reference them. So like how you're,storing these things on on your hard drive and accessing them isfor your for your full notes that are participating in consensus or even justpassing, passing information around so that so you can. So everyoneunderstands that this is the canonical blockchain and we're happy about it,and these are the transactions is only for his optimized for specifically forconsensus right, not right, hulling information out and so how you packagethese things together and then store them, and then optimize for accessingthem on your hard drive has nothing to do with what transaction you did a year ago onUrsua and optimizing for getting that information right and so the because ofthat, the the cost of asking the blockchain for some types ofdata is incredibly inefficient. It specifically the further back you goor asking like astec asking sin I have the blockchain. I run. I run a fullnote archive note. If I want to ask it: here's an address. Give me all thetransactions associated with it right it's on. Okay. How am I going todo that? I need to understand every transaction. That's ever happenedacross the history of this thing and I need to pull that. That means that,like you said I terus built a long time ago, it is to go through every singleblock and then go through every transaction in that block to them passfor that specific address and in some cases, based on the way thetransactions are formed with things like internal transactions like it'snot obvious, that it includes this address from a two or from at, like tis either money said to or from or I mean to in the contract. So I have to aI have to look at the contract and see if addressin said one of thosetransactions, so there's a myriad of ways that this address could beinvolved in some type of transaction across the entire history of theblockchain and to you think about that. computationally like I have to go to myheart drive and then pull all of these things. Are that query all these thingsto see if it exists and there's...

...interesting ways to do this, but, like,like, we said beforehand, the way that the distributed consensus works, whichis what these notes Ropier has nothing to do with asking that question, whichis why it efficient and so like solutions like the graft, were likeokay, well synthesized for that, let's build something that is optimized forthat and the ill get into kind of how you've done that in a language as youchose to like make it so that I can ask a question and only get the informationback that I care about and nothing more right. Different instance. Most of thequeries we ask at the blockchain or in the form of like a filter right, but- and that, like Iwant a transaction that looks like this, and it gives me back a bunch of thingsthat look like that, and then I have to go through that. It's a bloom filtersare kind of the general thing for this, like I gave it a pattern and say giveme all the transactions that Matos match this pattern and, depending uponwhat that pattern is, I can get way way away more stuff than I want or exactlywhat I want right walt. Not It's a way more, and so I have to pass through allthat stuff and then and then get it, and so what the graph has done, whichwe'll talk about her a bit is like basically me just like. I want this andit's like okay, and here it is that's it that's. This is all that that, andonly that, but if you want to talk about kind of the graphic graph q o Kino in general, how the grab functionsyou can say that, as like you know, you very nicely explained it. It does looklike you know that Doctin is a lingle structure, which is a data base initself, and it has the data all you need. All you like have to do is useand make a decent the application. But that's not the case. You have to pass every you knowspecific block to get the data, and for that reason the graphs was created as aprotocol back into thounded. When the founder, you know, got back understood,dadivae and a blockchain in general is the next web. It is the web three pointto for me. Definition of a three point to you can say that we point one onepoint or which is like you know. Definitely I have found from a tweet.Very famous tweet was that you know providers create information, we createcontent and they get paid, but two PONTOS. Even worse, you as a personcreate content and the providers get all the Monde and were three point O,on the other hand, is when the jowsers create, you know, content and theythemselves get paid. It is a way to make decent lived where it is. Away to you know, Have Censorship Resistance Web, which is going to rulethe world. If we have censored web ruling the world, then we have thesetech giants cooling, the world. That's why we need the next step in the weband that's what block came was, and that was the importance of making. Youknow this particular deal a De Breaker Web. Three Point: two is the de Breakerto the tech advancements. I don't you know, I believe every technology isgreat and everybody is doing that and doing their job, but I would always saythat you know machine learning was an improvement into the current web. Whileblockchain is the new webbite machine learning will be used in thisparticular web as well, and we are going to provera that- and the Graf isgoing to play a very important role. Why? Because you yourself like, as youwere talking you told about, we need a very, very synteche data. We need avery specific data that models are. You know made upon and based on that, theinformation is processed, that particular data has markets right now,Faloo only sells trata as targeted at Youtube only cells trata as targetedeggs and that their economy, but the at...

...a wet Le Porto believe in the opennessof data welly. At the graph belief in the openness of data, that's why? Ifyou create a subgraph using open smart contracts, you get open apis that youcan use what anybody else who is interested in your particular subway ofcan lingereth from it is not like a tragedy or a goal that you say foryourself and then sell it for your. You know intention purposes. So I t Ireally Yehoi sorry con, I'm sorry, but but but that was the main intention ofcreating and being so much involved in the web tree Pontou as a culture andalso you know, being so much importance, giving so much importance to thisparticular port of protocol for getting information to the world using openapis in a decent life passion yeah. I often tried to convince people that the current structure of the Internetis and like the problems that weassociated with it, mefres that the problems do weassociate with the Internet with, like you know, big tech and censorship and kind of privacy violations is a directconsequence of the infrastructure of how the how the Internet is made andlike an in this kind of Weben, spoke centralized fashion. When you feellocations that way, what naturally happens the emerging process? Is Yougain users that depend on your infrastructure that you own to serve them data, and so what ends uphappening is that you, just a mass, more and more and more data that youthat you have studium ship of and come to find out that in that bigcomes to be really really really really valuable. When someone says hey I'dlike to access a specific group of people with these attributes and seeand sell the Mans, because who do they go to, they go to the tech companieswho have the infrastructure? That's holding all of this data and curated insuch a way were like. Oh, I know who those people are where you go: you'rethinking for the money right and so like that is a jet like what weexperience today in social media and getting in incredibly direct serf ads,which are and sometimes convenient, but I'm incredible for like like violationsto my privacy and I'm not getting any money for it. A is because of the waywe built the the Internet, but three is a direct attempt to try and change thatin a way we're like it. The way you build applications still has reallyreally really good quality data right. You can do machine larding, eventuallyon this data, and then the blockchain is perfectly structured in a sense, itfollows a specific set of rules because it has to otherwise they wouldn't bevalid transactions. So, like doing machine learning on this fast of Matadata will be a thing. I actually wrote a tweet about the other day asking likewhat people have been thinking about doing this, because I'm very muchinterested in it, but at the same time like accessing those people based onthe attributes, you'd like to access them from will probably end up being like a formof consent. In some cases in case is not, but more often than not, when youwant to give them something, they're going to be they're on probably goingto benefit from it or right. At the very least, like youhave this sense of there's value, a cruel at the at the day, the layer andso accessing that information oftransactions from various people based on attributes and then doing somethingwith it to make money. It will probably have value accrued atthe path of base layer and assuming that these people are sendingtransactions, they havee the underlying...

...asset at the base layer, they're,probably going to benefit from it, because that, like there's potential ofthat value goes out. Right. Also, like you know just to add to that particularpoint that you're talking about this is more so, like you know, I believewhatever progressions we do in life, whatever innovations that happen inlife are somewhat related to philosophical structures that werecreated back in the days and that's how you know we we progress and make futuretech that's cold by how we evangelize about the future, and there is thisvery you know specific quote. I forgot the person to to told about it, butthis is very specific to human lives that if you are in a particularecosystem or an organization where people are expected based on theinformation they have not on the world they are doing, then you are probablyin a wrong GECO system in the wrong. You know world and that's what the webtwo point over doing. It was respecting. It was you know, giving a lot of moneyand increasing the share prizes of companies which had information alltheir economy is all their. You know, structures and all their what we callas the money, Legos or the money. You know, aspect of generating revenue wasvery much related to the data that they had of fuses, which was not even there.It was generated where you know interacting with those decent ties-application, not I mean CENTAINE application fit. So that was actuallyyou know the reason they made so much progress, but that's not how theequisite should work. Ecosystem should be working with open data, open sourceinformation, open source, softwares, open information and people who shouldbe valued really valued are the people who are innovating on the data that iscreated, and that is what the graft does. The graph opens the doors to thebirds next GEN application data in open source fashion in a decent life fashionwith open. I guys all you have to do it is make use of these. You know datapoints very specific letus. If you want specific data points, you can createyour own sub grabs. If you want, like you, know, suck information from thesubgraph created by somebody else. That's also open in the open markets,all as a user or as a person you have to do. You know incur value in thisecosystem of the three with the graph is used, that data to do innovation andthat's what the ecosystem. I WANT TO BE WHY? Because, again, with thephilosophical line that you are being respected, not for the information thatyou have in that particular ecosystem or all narration, but rather becauseyou as a person or as an organization, are innovating. You know you get inrespect and revenue because you haven't done something which you know. Progressis the world. So that's what you know was the philosophy of creating thegraph as a project, and you know that's that's how it all started funny. Thus Iwas in Best Berlin right when the graph got started. Ibelieve it was just I remember walking by and for some reason or another. Iwas really into graph Cuba Graph Co, like just grass, as as a technology for creating a s, a single specificationthat both a front and back in can use, and so you kind of married that at that previous divorce, that that people hadexperienced of like who designs the API and who follows who went between backin in front of development. We can rely get into that here, a little bit butlike first, I want to. I want to ask like that. I Eve approached to this: Isthat cool? I have a really. I have a really interesting way of pullinginformation from the blockchain and indexing it in a specific way such thatpeople can ask the types of questions...

...that they care about and getinformation really quickly and succinctly. But that Eve way of doing that is like Ibuilt the service, and now you query my service, and now I have folgate overall that data and I can maybe choose to either not serve you. What I think Ican change that information or I could. I could fall off a cliff andit all over works anymore. Right, how is the graph democratizing theprocess of of indexing this information and in serving it to users right? This?Is this is a pretty awesome question and again, I would take you through along part of how we are approaching the process of decentralization, and youknow how we have made products and protocols in a way which, actually, youknow, dwells into that Etos and actually, youknow, make make make the next move in that particular direction itself sobackin two thousand a yet like you know, eighteen you're talking about when youwere in a the Berlin and went through that particular graph protocol thing atthat point in time we had like you know, or our first prove timebeing created, which was the grass hosted service, how the things movearound. In that particular space, you can say that we always took theapproach of progressive decentia on what that means. That ment is in thethree one. Some people are, like you know, just create a white paper raisesome funding and boom three years gone after three years. They're, like youknow, this protocol is totally decent lives. This is not how I work. This isnot how something what of product can be based. I don't you now reallybelieve in that it has to be in a progressive day. You cannot, you know,liberate the world. You cannot have like from dictatorship to everybodygoverning themselves. It has to be in a progressive way right. That's that'show the world has to work, that's how people have to go on and be disciplined,and that's what the God, because you know, if you leave people you know inthe free world, they will, you know make lies because they are not trainedto be decentlie. It has to be in the progress of way right. People have tofirst of all understand the government's process. They have tounderstand the disciplinary process. They have to understand how thingsshould function right and that's how they will progressively you know, adoptthis decentlie society and that's what the graph as a product itself. You knowinculcated in its values that initially, with this query, launch with theproduct launch, the first move that we are going to do is, we are surely goingTaou know, have a product which is the host of service, which is the firstmove towards decent, live indexing and quiring. This will give the productmarket fit as well. As you know, I have adoption to different different depthswho are currently having these centralized servers. They willunderstand this protocol. They will understand what sub graph is how tocreate. You know this particular thing how to index using the protocol insteadof their centralized, so called Servis and you know, create API which can beused by their particular application, as well as some other applications infor and behind using that particular information through the hostage servicewas, you can say, a very large indexer indexing different different. You know some graphs. We are definitely want togo very be to what a subgraph is, but in a brief you can see that contractsthat a particular application has. You know they have certain kind of eventsand they need to be tracked for getting information outside. So those kind ofevents are, you know correctly tracked by the sub grabs and the sub graft ismade in a way that, in the end from the contract events, you get apis which canbe used to you know, process or produce...

...data that is directly served in thegraph cual format in API to your application, to the Beta to directly tothe people right. So the host of service was the first approach inmaking the decent lives indexing and clearinga reality. The first move towards decent lives, indexing and quiring, andwe had a great you know. Success in that particular approach andpeople really love this kind of thing we go. Ah, it was a big pain pointright. You can have decent lives. Applications you can can have lendingyou can have borrowing, you can have F, you can have social tokens, you canhave social media on the block say, but you need information that has to bePoterat to the front end to the users in a decent life in up protocol basemanners and people don't want to lun their own server on their own full nodsand then get that information and that's what the graph actually solvedand people really loved it June slap synthetics. All these big decent lie,applications build, I of graphs, interacted with us and that's how itall but started. But to answer your question about the place we want toreach the decent lies in vexing and Quin. That is what the graph let workis. The graph network, you can say, is not just one indexes, but a group ofindexes permission, less anybody can come and index on the graph network andlike, for example, there is a subgraph right and there are hundred indexes intexting that of graft right. So, instead of one indexes on the hostedservice. Now there are a hundred cus indexless that particular informationusing the same subgraph process and serving that information in a decentlife passion by the decent trice. Because now we don't have one INDEAN,we have a network which has participants in the form of indexes whoare indexing, the sub Graf and then storing that information, and they knowproducing that information, life and e. also to your extend that thatinformation can be protected in the form of proof of indexing, that datacan be verified in United Stata is true or not. Why? Because they have these,you know network. We have part disciplines in the net. Work are theparticipants, let's go a little deep in the network and how the decent lizeindex, including actually functions in the way? First, one is the indexer inthe graph network right. I already told you about the host: If service at thefirst product Nateral, the second one, which is the decent, lies more, youknow, decentlie version of indexing inquiring with proofs is the graphnetwork and currently being run only for a tedium network. The vose serviceserves about twenty two plus networks, and you know soon you'regoing to do that for the net or who else yeah love lot of work, I would say-and we can talk about like in a war, the future process of indexing and howtough it is to yeah and get get the data in a decent light fashion, as wellas still make make it happen. But, coming back to the topic, the graphnetwork, it has five participants or we can say right now. One is the graffindexer, which is indexing. The information specific war subgraph right,that's the indexer. Second one is the curator, since the graph of theprotocol is a market for different difference of raps, there are fewsubgraphs, which might be more important. You know there or, as in ina way, might be quarried more than the rest of them and there might be somewhich are like query less so base. How do you find out as an indexer to getrewards on the network based on the economics, the more equating the morethe revenue egen rate, how the indexes find out, which particular sub Gab theyshould index and which should they should not, but that is baton. curationsignally to the curators are signaling on different difference of groups thatthis is important. This is not. This is important, but and that's how you knowthe next step of curating works and...

...that's how this thing progresses andfunctions in the right processes. So the indexes are signaled by the ratorthat this particular subgraph is something to Peter to this particularsubgraph is something you should be indexing right. Why? Because, if youindex all the Sagra, you have limited a location. You have limited resources,you cannot be Google for the block came in that no one is a shot about, so theyare like in no small small googles in the name of indexes, while indexingcertain subgraphs. The second one is, as you told trators third ones are thedelegated they just you know, are you can see passive part sepents in thenetwork? We just sing, you know delicate to a particular indexer, andthey are only here for the Neto economics and make sure things areright, and then there are arbitrators who are actually making sure that theindexes are doing the thing in the right, with the thing on as proof ofindexing, which makes your at the data that the index is serving to theinformation to the to the application who has created the subgraph is correctand is right that is run wi a proof of indexing, and that is verified byarbitrators in general and that's how the graph protocol in the network sensefunctions it. So you have a particular subgraph. It has one fifty plus Googleof its own, which are the Indians right, who are indexing that particular uptrap, making short that the information is correct and there are arbitratorsmaking short that the indexes are doing their great job as Youll as curators.Very fine, like this particular aprop that they have is the correct one. Itis. It is the right you know one which is going to be used in futureapplication. This particular sub graph is going to be quarried in the future.I, how actinal guess about how a specific arbitrator just take figures out whether or notsomeone is indexing correctly, I'm assuming it's some type of interactiveproof or like they need to understand what a specific subgraph does and thetypes of things you can do within it and the Times of data that arerequested and they need to know that they ned they didn't know the rightanswer and they ask a given for person who's indexing. One of theseputting these self many googles question there's a random question withthe SUP graft and he see I think if they have. The right answer. Is thatbasically the case and so like? Assuming that they give you the rightanswer, you can a certain number of times. You have a certain M of a loveof confidence that they have all the data right. So it is more so, like youknow, making sure if a is somebody, it's optimistic proof right, okay, soin a way it is understood that things are correct. Everything is going right,but if somebody you know asks that, I make sure that this is right. That'swhat is proof of indexing. That is the proof that is given by the indexer thatin generating you know, I m you not really indexing this particular up Gruband I'm getting rewards for this particular subgraph, and there arecertain kinds of topes generated which actually make sure that arbitrators canverify. Again, you know, generating prove, is easy and very fine should bemuch much more easier, so that you know the next book is very very fast to makesure that to for indexing is correct, and that's why hanks currently havingon a trilium, because move is generated on the the of net work itself right.Okay, that's how I'll be greater stuntin and that's how this works, howthe axis for that like how to? How does that like why? Why would someone dothat right, like so like I'm assuming there's some type of stake associatedwith being an indexer, and if I proven, if you know, if there's proof thatyou're not doing it correctly, you get slashed or you removed. I I the work ina corect. That's so that's how you have indexer allocations, and that is thereason you have delegated allocations...

...right. That's why we have likeprocurators. It's like subgraph is important, just Tigart, that a Rapleywill make sure that things are going right now for the index ers to makesure they have the rights take in the LETO. They are doing the right things.You know producing the right stuff indexes have to stake on the network,allocate to a particular subgraph right disputant, for example, to be anindexer on the graph network. You at least need to have hundred k gate stakeon the network. To be a part of that, you know, I'm the Delegato can sixtimes the index has allocated to a particular subgraph six times make make that the thing right soindexes allocated to a particular subgraphs right and make sure thingsare right. Abit traitors, then verifying cross question that that thisis right or not, and if it is wrong, then there is flashing that happens onthe network. What is the the reason people should have more allocation onthe graff network? You can say an indexes is incentivized to like youknow, have more stake in the network, because the more the stake you have,the more is the reward which is a mix of the quaries plus the three percent.You know inflation in the network that is givento different different in the exerts on to the more allocation you have as anindexer. The more you know more is the more is the you know, the body you willget because kayaks more sense to me, because I was curious about how much a Colei assume that people whoare asking questions like the ind users getting information or the applicationdevelopers who are who are queering, your self grass they're not paying forqueries, because you want that. You want that information to be. You knowas freely accessible as possible. So I was curious what what's the incentivefor someone to actually run an indexer because, like you said pretending,depending upon how how much they're inning it's a lot of work and it'sright in to have a phone in so having having that Ben said. Having that beingsaid, there is a combination. First is a three percent inflation rate on thenetwork that is distributed among index El. We on what the percentage of ourlocation they have on the network and the second one is the Quadi face right.So every sub graph has apis right. Anybody who wants to access thatparticular sub graphs API, so graph is let's say you know you Nisa has theirown subgraph right and they have open apis. Now. All you have to do is justgive a subscribe to that API and generate queries and that query has tobe paid in Git as a token, so you have to pay pay, pay, take particular API. IGat right and you're, not paying the graph or a nodality you're, payingrather the protocol the indexes, because it is the combination of thequarry face as well as the you know, the the indexed, a location plan, thethree pretention okay, so you are paying quaries whenever you do themright and how does that? My a e price for those is that is that the actuallyindexer who said the price for quaries right? That's that's also, like youknow, you can say, that's that's like abargain between the indexes and the person who is trying to get it. Thereis something on a scale, her TATEGAWA, the block. Again, I just want to likenot take some Tumut and talk about that. Like the immense work that is beingrequired to make a decent ly than Xing and Quadi the reality before the graph, there was nothinglike indexing and Quari. When we dwell into this, then we made the hostedservice and wanted to make the graph network we had to run round to everyeconomic spot. We had two researchers totally working on the economicstandards for making this happen and...

...making sure like the inflation iscorrect. It's the right way right next. Well, the indexes should get benefitsnow they can eventually make their businesses all of indexing right so andthat's how a protocol should be. It is not a business, it is a protocolprotocol generated businesses right. That's why indexer like should be. Ifthey have this much a location there for small indexes. This is the teamthey can have. This is a business they they can have. So this was the firstpart like you know, making show everything is correct and you knowwaiting on that and the second one is, while you are indexing or decent lives. Data base, which is inthe form of link list, is very, very tough to make it in a decent lifefashion in itself, and that's why this particular specific question that youasked and many other required innovation on their own and the graphas a protocol and the three co devs, the you know the first facial teambeing Agen Rod. Never ever ever, you know, negator or, like you know,subsidized on that they were always if there is no mission required. If thereis like you know, a protocol to be changed. If there is a lot of Cote tobe written, make sure but ill be written, but we'll never ever. You know like compromise with the decent lizeand indexing and crating part of it. We will. Never. You know make like hundred.Only hundred validators are a lot of something like that. You know if image,usually so, just to be clear, the most more often than not the there can onlybe a hundred validators is a circumstance of the of the distribution.Sins is process like a after after specific mat, specific amabilites insome types of like old, pvt, tishue consensus. I think it very inefficientand not useful right. Can the right head and message of the passman passaround, but for Familie that, as as more isolated as a single index, or doing alike pulling a specific amount of information from the blockchain andindexing and serving it you're not limited to that type of thing? Right,that's that's! That's that's why we can have as many indexes I as are there solike it's. The protocol is as Decentlie as it can be, but having that been said,there is a limitation. You know for some networks AVONLEA. Similarly, therewas this limitation that you know if you are doing this Guarin, what shouldbe the quilty piece? What should be rated? It should not be like you know.Okay, this is the rat. You make it happen and, like you know, every query,tough or easy eath call is a little more expensive e get log is moreexpensive. It get blong by data is easy. Why do people like you know if it isfixed, then then you know it's no fun. It's not the right way. It is not thewatry way. It has to be a bar game between the indexer and the person whois trying to be at it. It has to be to and fro and thought that the graph, asI told you, is the machine for innovation in a waited something knownas scaler, which is like a state channel of the roach, but weep in a waythat you know it functions in the most callable passion she state channels are great, but wehave very specific limitations that the graph sold made something known asSkiller as a product and made sure this payment process is also happening. Soagain, just to talk to you about that, how scalon you can say indexer, youknow, and the person who's trying to get the money. They have like anagreement that this is the amount that we want to pay then pool or indexingsummited. And then you know the whole payment happens, and this is recurredon currently on the Polygon Lovcen and we have some starts. For example, likepickle finance till I saw it was in like eight quales for second and theypaid something around the grt a week...

...that was US looking for. I was looking to see how I looking I'mwhen I'm looking to the right. I have your web page pulled up in dockdocument, so I'm looking to see where scaler is because I haven't heard aboutthis process and how like, what's the different, like we've done a lot ofshows episodes on the show about statechannels and various other scaling technology. What did you do to like? We can assume the audience knows whatthe State Channel is and how it Jim, like generally operates right. You lockup funds. You Pass Menus past signed messages around that basically changesthe account balance of both people, and then you say all right. We're done andyou close it right. So you have, you know, block up trend like lock uptransactions and you have finalization transactions. What did you do forscaler? That different shates from that so so utentis Kale, is a product of anext last. It's in not coming together and making this you know, approach ofdecent lives, invexing and quiting Wi state channels. That is, there is onevery right, another quarry they are made into a receipt that is transferredinto the vector nod, which is again I'm going very technical. If you don'tunderstand it, like you know, we can. I will definitely link to to the blocks,which is the Iowa. But but the thing is there is a query, then Terry, secondquarry their median to a deceit that received is made in transfer to victoryou, not transferrer, which has ter out. Now, it's not like. Like the state channel, you are oneperson in one person and we make a strait channel we to the transaction.It's like one query. Second, wary make the receipt, and then you know why thevector note they are transferred to a particular consumer or the statechannel for doing the payment. This is this scales, the scaler or what we callit. The state channel approach in a much much more better fashion and makesindexy and quiting a reality wine, because you are doing it, for example, something like tiny swap is doing about.You know a good number of queries for a second right and if you waste so muchtime by state Chanel making the Strait Channel- and then you know doing thisparticular approach, then finalizing the query, mint and then serving thedata then bear back to the first problem that we talked about, which is which is like you know, the Dexter.That is I so this is. This is a podcast adoin video for, but I'm going to sharemy screen with Hol with wish the blog that came up here and use this visualization as a way tokind of help right figure this out right. So it's, but so is this. Is thiswas a difference right that we introduce the vector part of it heredown the line, consumer and doubter coming together to make it happen right.So there was a quarry waiting to a receipt, acte transfer back to transfer,and then the payment happens. That makes sure that the scale is infiniteand there is a lot of stale happening in the network and making sure like thepayments that are happening are in the correct formats. That's that's how youknow the skiller works. Okay, so the way you cold say it's an innovation on how you like the messages that arebeing passed around so you have like. I was call like the the atomictransaction right, the primitive which is going to be the query right rightand in front queries. You then make a receipt and then it so like it. It'sdifferent people passing around different things, and so you get to thewhat like? What data structure, if you will, what proofof all these things actually gets embedded into? What would be the statechannel and then o you to scale with a bunch ofdifferent parties, as opposed to like what people traditionally think ofwould stay channel as me, and you have a negotiation, we lock funds up. Wepass around signed messages of moving just those funds and then close it whenwe're done. This is kind of like changing those messages for passingaround to allowing a bunch of other...

...people to pass messages around provethat they did it and then give it to us so that we can embed it and what wouldbe our state channel right and if you like, you know want me, I can alsoshare Shamus Green. If that's sile and give you even more, like you know,reaper dive or, like the user flow of how the scaler approach our sides I yeah, can you see so this is basically like you know. Iwould again not be able to go through the whole thing in a podcast. It doesrequire a podcast the for of its own. If you want to go on, I ll to anotherone today to kind of go and more deferent to kind of how the stuff works.I'm fascinated by by like how you have strong guarantees around like valueflow without hindering the actual userexperience. So yeah right. So it's like you, know a client and the indexerincorporated, and then the other. You know proof and the solution comes in sot some part is off chain and Tom Part is on share to make sure the disceateas security guarantees in gone. If you talk about so this is basically likeyou know, as I tell you, how scaler functions definitely suggest you tostrong recommendations to check the blog definitely check this particulardiagram that I created using my drawing bad and writing. I know, but you knowthis is a what that, on a description for people who want to read it right.This is this: is the level of innovation that the Graf had to do tomake sure that there is indexing and quitting in a decent life fashion tothe decent lies application to call it a decent lives application? Let's givesomeone a o dive in a little bit to just the concept of a sup grafts rightis because we were running e e. it's many times we want, but usually thesethings go about an hour so yeah, but like my interpretation of a sub graph, isthere's data on the block, chackright, everyone who's using a therium sending transactions. We talked aboutkind of how it gets compacted and then put into a hard drive for nodes to dopasific consensus and all that right and me will just use the standardmethod or the standard kind of perspective. I'm an applicationdeveloper that I and I'm making a decentralized application. I have smartcontracts associated with this Nesera applications that take transactions from users and thenadmit events based on whatever these contracts. To so like me as anapplication, and then I may be interested in some of the events from other smartcontracts for my application. So what I wouldwant to do is like say like these are. This is the bucket of information that I'minterested in, that is on the watching right Goe, and this is how I would naively pull it from transaction data of the theriumblock chain. So I'm going to take that and thentransform it. Using Graf Anin define an API of all of the things. Any end userwould want to know about that data. That allows them to ask very, veryconcise questions about any any of that information or subscribe to any eventwhatsoever. Right, correct, and so that's that's a subgraph. That's that's!That's pretty. You know awesome definition of sub grab that you havegiven, but in genere you can say in a more simplified sip, simplified fashion.You can say a subgraph like the graph is a protocol which can index events ison the contract that you give it subgraph is the repository which yougive the protocol so that it knows which particular information to index and is, do you make the holy padiumlooked in, but doesn't why things...

...automotive or does it need like manualspecification of this is where you get the information, and this is what youdo with it. So so, like you know, we can definitely will go people into howa subgraphs and what on the composition, but in Gerda you can say it is arepository which is collection of files. You know made globed into a repositoryto give information to the graph protocol that this is the informationwe need and in return you get graph. API is which you can use to differently.You now make your application should now we are going deeper. We go likewhat is this repository? What of the files does it have? The first one isthe subter of Vimala. This is the file which defines that this is theinformation we need. This is like all. We need to create a subgraph right.This is the net book will be indexing between Dein lets, etherium, so nameman let. Secondly, what is the contract you want to index? What is the datasource you want to index to the contract? Address third thing: What arethe different deferent? Even if you want to index so the number of evils,the name of the events all of them sufficed to one, so you can say thesubgraph. Why mepil is the description of the sub graph? It is you know, butwhich is the most important like you can say, the really may not envy of theSagram you know, so it gives you all the information to the graft and allthat this is the information that I want to index done with the subgraphtio five. The second one is like what information in what particular pressionyou read. That is why the the scheme- Adora Il fine fine, so youjust tell like this is a schemer that I want. This is how it should look, and Ijust define it. So it's like the scheme of definition right. Similarly, you dodo in a bet to point over it, so it's misty. The third one is the mappingfire, which is the most important one. So you get the information from theground using graph like in a protocol, so it gives you the information. It haslistened from the events that had been specified in the subgraph VIMALA. Ithas the events now it has to be made in a fashion the data that the events havegot have to be made in a fashion so that they attest to the Schemer thatyou have defined in the scheme. Adoracion fine getting me so the bimefile has the evens. The graph protocol listens to that Puente for you right.You have define your scheme now. You have to convert this information thatthe grafter recall has for you in the scheme of format, and that happens inthe mapping file. So it is an Assey skip format to make like some things inthat, so that the scheme or suffixes to the the event information. That isthere right very easy. Not, I would say it's like you know: It's not very tough.If you dont basic JASC, definitely SODUSD you to give it to try strongrecommendations to go through the grafham flash dogs, but this is howthings progress. This is our things work, and you know this is how youdefine a a graph, very easy, just a Wie file, a scheme at file and the mappingfile to make sure that you have the information of the contract stored. Soeasy, so could you do you have any intuition on? Is a computational resources requiredto run an annex or based on a given Soutra like what is? What is the what scales the worst so like? If Imake a sub graft, what should I be worried about in terms of amount ofresources that I need to maintain an indexer that that services thatsubgraph right? So I would definitely again giveyou a link for that. You know there are certain things, but you can say that ifyou want to run a small indexer which...

...is just indexing, let's say one forgraphs. What two subgraphs, maybe five for Bass, not more than Lak, you canhave like four C pos opposes GB of PG progress. You know database of G, a disk with a Tera Bite, one TERRABAthing, four BIMS of the CPU and, like you cansay the memory of sixteen dbs. This is a small like thing again, you cancreate a Google with just a small thing because of the beauty of the graphprotocol, but I can give you the links to that or maybe like how do I sharehere? I chat yeah. What's wrong, one open up my other file here, I'll putthat in description for others to say you even just add it on here. If youwant a Harvard requirements on screen forthose who are watching podcast you just going to have to go to the docks in thedescription. Sorry part of that question was like what parts of the things that I'm looking for skill battle, because I given sub graphcan be large, like if I think about like a large dat like Una, swap like I'm thinking about how manycontracts is it given to the centralized application care about thelarger ones are going to have to track a bunch of different things and a bunchof different events in order to keep track of all the things they needto index, whereas, like I may want to just look at a single contract thatdoesn't do very much so the compotation Al Resources of those two things aredrastically different. What what what should I be worried about? Should I beworried more about events or should be more worried about like transactionthrough put for, or should I be worried about the number of smart contracts ingeneral? Or is it I mean, of course it's a combination of these things, but,like you, have any intuition on like this, a differing scale of them right.So so, like you know, I tell you: The graph is the innovation in terms ofQuadi in a decent life passion we have about as many number of googles in therafter to call right as the number of ILEXES, so everybody can innovate andunderstand their costing model to be beneficient as an indexer, but it is acombination. It's more like you know. Your question is directly like. If Ihave a sub graph, which has very less number of quaries. What is why will Ibe so motivated to be a part of the UNI so o graph which, as like hundredevents or something like that, so the reason is June sokoma will have morequaries the more quite you have, the more, not the more like as an a as a person as aSOGRA index or the more rewards you're getting why the economics very throatvery thoughtful. The graph is a protocol is created so beautifully.It's not just the allocation, but also the quarry feels that thing at Otetswhy the Ewe will right so the more because you Ni sa it is more peoplevisit in food or tales, but the more quiting happens. The more is like youknow, you can say the more people the quarry face will begenerated and given to the indexes on the part which is tough, like it'stough, to indict something like you need to have, while it might be easierto index a sub grap, which has like say just one event to something like that.That's true, but the craft believe me with you, know two more code, depthsbeing added is working on making it much much more efficient. That issomething known as the craft node which listens to these events, and that isbeing you know, within a part of its own. You know so that that thatparticular part is written in Rast me...

...graph, not partes, written and rust,and that is being like made much much more efficient. There are differentdifferent approaches that are being tackled may be one approaches tolistening to our PC. All the time you know there is any vent fired. Thesecond one that currently the streaming fort, which is the second code, haveintroduced the file house, which is storing the data in a very like BeryVII, milly quol format. And then you know making the graph yuill approachand then giving it out, which is the fire, was by pathos approach of doingthis, and there are like different pen thoughts that the graph is working onand making things more efficient. But, as I told you, the blockchain is thefuture made with the the consciousness of security and the graft to the callis the protocol with made with a consciousness that this blockchain,which, as security, should be as secure as it is. While we make the indexingpart and availability, the data belibe part of that blockchain a reality in adecent lass fashion. So all that an din world, the prone definitely check outthe Brafford to have more information there. But, yes, the graph is workingto make even uni swap graph indexing. Very, very easy. Thinking of thatparticular graph, very, very easy, and very very fast, and me many more things,maybe, like you know in the future, when you see we'll, have non medium,compatible chains also being indicted by the graphs, because the approachthat the Graf follows is that the graph is Internet of block say believes inthe culture of Inter of love. The other traveling. Is this wonderful languageof like being agnostic to Keith's? What that's? One of thereasons why I liked in the first place, to be honest, is was that I was allowedto mix and match various back end components, whether it be like RSSFeeds, creating my own DB and serving servicing. That party incorporating you know a bunch ofother stuff and and combining it and melting it in such a way where I have asingle, unified API, and I can change it and it's dynamic. I can change allthat stuff the same as long as it's servicing the exact same API right andso the front had never cared it just follows it just follows the API thatboth the back in front and agree upon, and so when you think about that in thecontext of the graph and web three like you, can you have this future, where itdoesn't really matter man it, of course it matters, but like you're able topull information from a bunch of different blockchain networks andvarious formats and whatever and unify that into a single API that at thatanyone can then ask very specific questions about for anyone, who's,never heard of graph the language. I definitely recommend you look into itso that you understand kind of the the benefits of being able to ask a veryspecific question and only getting the information you care about right and similarly, as I was talkingabout like grace in itself, is you know, application agnostic? Similarly, thegraph does not believe. Like you know, one particular Bob Chain is the winner.What three is the winner? The next progression in web is the Willer and wedon't want to decide what people want to build on as a block. Can we justwant to be an infant ature? If you want to build on so Lana, you want to getthe information to the decent side. Application is the graph. Do you wantto use near? Similarly, if you want to go for the tedium built up Wi ups andget the information, so that's what our approach is towards Internet of LockTams, and we believe that this wort three future with the web three inPastrucci in gender, is coming to light very very soon, like sooner than youcan think. I'm here for a reason, all right for of. I really appreciateyou coming on the show and helping both me and any of thelisteners better understand what the graph is doing and how it does it andwhy it'll work and...

...those are listening. I stronglyencourage you to go look into the links and the description. Anything else isanything else. You'd like to and Parnon the words of the Great Stajo, I wouldsay they have e Stat fulish and keep Elinora. I.

In-Stream Audio Search

NEW

Search across all episodes within this podcast

Episodes (109)