Mango - Git Completely Decentralized - Alex Beregszaszi

I and my name is also xik as most people know me on github so I'm going to talk about mango which is making it fully decentralized and what does that means it puts get repositories onto athyrium swarm and I PFS I just a quick note about what get is get is a version control software where most developers keep their source code and they make changes to the source code other people can track these changes and contribute changes so the first thing is about get as it was created roughly ten years ago and it was created with decentralization and distributed systems in mind however ten years ago decentralized systems weren't ready so you couldn't do this you had to use a central part to get however today we have decentralized systems in place the etherium I PFS and swarm so this is the time to accomplish the goal of get the first thing is why do I care about mango and what do I care of it sorry why do I care about mango and why you should also care about mango there a couple of reasons you don't really want to trust the central repository with all your data they could take it down and you want more security than that this is a very important fact that you get by default if you use mango you get proof of existence of your code so whenever you push your changes or your code is a better name so whenever you push changes to your code and it will be stored on the block chain as well and because of that you get the timestamp according to the block chain and you can prove that your code with all its content existed at that point of time you can see that's very important because you can prove in case of legal issues that you were the first implementing that and you have the prior art and that's really important in the open source software and I don't really used for mango which get cannot really do is storing large pies so get just towards large files as a single block and whenever other people copy your gate or they make changes to that or they store other files or they change the file it will be a brand-new entire blob so if you have a file of like 10 megabytes which is the solidity compiler for every new compiler version you have the full file currently the solidity repository is roughly 100 megabytes in yet just for the compiler versions which using IP FS or sworn you have deduplication and whenever you store these files and use it a mango only the differences are stored this is something many people don't know but get and the last really important point is it here iam as a cause is building decentralized systems the whole source code of this cost should be of label two distinct rely systems the last photo so what is get itself get is a content addressable file system what is form it is the same what is I PFS it is the same there's a tiny bit of difference though it is working on your local system what other two are working in a global decentralized way so you see it makes sense to move your git repos to either swarm or I BFS that this is something many people have seen you know get looks fairly complicated people mostly know a few commands and they get afraid they delete stuff and they start from scratch I can assure you a few tricks or you know just try to explain how it works internally and that will help you to understand how mango works and don't be afraid it won't be long so you get everything is an object and every object has an identifier that this identifier is a sha-1 hash so in this example the first one example I'm just putting putting some data in to get and you get back the identifier and of course once you notice identifier you can print the details so there I just loaded a file which at the content hello material hello Shanghai on the file system all these data blobs are stalled stored in a compressed form ok so now I have data and get and the second thing I need is the file name for that data and probably permissions in some other details so the second object think it is called trees and creating a new tree is simple use the identifier of the data and you set the file name next to it and as you see printing this object works the same way as previously ok at this stage I know what my data is I know what the fine aim is but I don't know who created that file and when it was created a white was created so the third object we have is called the comet and it's another simple command which you can use and you get back another simple hash similar to before you can print that hash and you get the familiar display which you already see in get logs so I'm not seeing all of you should use these comments to manage get because you do have the higher level systems to do that but because of these three objects you can represent get in a tree and the leaf nodes in ingots are the data blobs then you associate those data blobs with a fine name and finally with the comet and subsequent comets can refer to previous comments etc and you see the tree is just a flat structure if you want to have multiple directories you can just change trees together when you create a merge or a pull request multiple comments will be referenced in the next comment ok so that was the hard part a little bit what kind of objects we haven't yet but you can see we have all these simple bits of data and for a single comment and actual you know high-level commitment you put files into your git repo you have at least four five objects so my first implementation of mango was just building all these objects putting all these objects into any theorem contract that works but probably it's not very useful and for a simple repo of a couple of files we just a hello world in them it takes roughly 50 million gas like developers earn a lot but probably they don't have enough money to keep up with typos changes and pushing attempt to an atrium contract so better implementation is realizing the fact that we can use IVFs and swarm so this is what I did and I put all these objects into IP offense and swarm but as I've mentioned get uses sha-1 heshes and probably all of you know that I PFS doesn't so shavon isn't really good anymore we could change get do not use shavon but unfortunately it's really tied in which have one and we want to keep a backwards compatibility with every tool you have today as well as solutions like github so we cannot change that well a simple solution is to map I PFS identifiers to get identifiers so what this means is I push every single object to I PFS then i create a snapshot file of the mapping of these identifiers and i push this snapshot 52 I PFS as well okay so why do you need it here IAM right well you don't actually need a theorem at least to this level when you push all your content to I PFS you get this top-level snapshot file and you can give this snapshot identifier to anyone you want to give your coat too they can download your code that way however if they make changes or you make changes you're going to get a new snapshot identifier so then you have to give that identify to them they make changes they give a new one you make changes give a new one it's not really practical so this is where a theorem comes into play and it makes this a practical solution so at every gate push they push to details into a mango contract which is what was the latest comet which is the gate identifier and the location of this snapshot pie and today it is cost hundred forty thousand gas which is a big difference 250 million and this contract isn't optic so this could even go for the deck so what is mango and mango is actually a couple of things it defines the semantics had to store get objects in I PFS or swarm it also defines a contract interface for itanium it gives you an implementation for the get back end in able to communicate with each here iam and I BFS and it also gives you admin tools to create repos so every git repo has its own contract there's no point making this complex and having a contract for a number of repositories this has to be simple and you know secure as it can so what the etherium does on top of this it can manage access control and you're also free to implement whatever else you want add there's just a repo interface you have to conform to but in if anything else you do is up to you m for the actual repo contract which uses hundred forty thousand gasps there's an implementation in the mango admin tool chain is it this is an actual example how to use the tool chain and once you install it you can create repos this of course will need access to an idiom RPC node as well as to I PFS or swarm so once you create it you get any tea room address and that's the address of your repo then you can create new users you can delete users you can set administration rights and a couple of other things okay at this stage you have your repository so the next thing is to upload data into it assuming you have your repo already you just have to set up the endpoint which is your contract address and you can push data into this repo let's mention before we don't really store the data in the repo but we need to update the references in your repo okay now you have all your content a decentralized so it's time to share it with your friends and they can download it as well since you are able to create multiple accounts in your repo contract your friends are able to push changes to your repo at the same time while you know contract addresses work I think there will be a better way to do that and I hope ens or a similar system will come along soon and you will have symbolic names for your repos such as solidity that etherium that mango the thief I still need to convince the guys to register five letter names in eve but hopefully that works so where are we with mango is it a replacement yet of course well not entirely but there's certain use cases it's already capable off I would suggest that you use your tools you're familiar with today and your services you're familiar with today but once you have a version of your software or release you want to persist for and an indefinite length of time you could use mango and you could push that certain chainset the certain version of your software to mango but this should change in the future so we have to improve mango and we have to improve tools to provide full replacement full decent replacement to social coding so the first thing we have to do here is which is a tiny thing you know with each releasing it you can only set a version number we would be rakul to also set release notes what that version includes so that's actually possible with it add this feature called get note which you can use to put notes next to any object it's kind of hard to use so this has to be improved the second big thing is to have issue tracking we didn't get and google has a project to do that with get notes but it's a bit clumsy so there's still room of room for improvement here the really big work here is a supporting pull request or merge request and as well as like github start forking so if anyone has good ideas had to do it please talk to me or anyone involved is improving yet and of course we need user-friendly front end is yeah people some people are happy to use the comment line but others would be much more happy to have a nice web prompting to create manage and view and interact with all these repos one important aspect there with improving the issue tracking and the pull request would get this wouldn't be only for mango if you have all these things waiting it itself no matter where you take your repository or your data including your issues and your pull request will stay with you and I hope actually github and get lab and other bigger companies using it would be able to support some of these features so you could freely move between mango and the decentralized the website or additional eyes up and between these services without any hassle well thank you for listening I think now it's the time to check out mango and actually no one believes this but it works already so you could try it out and thank you thank you Alex