Fluence Labs Wasm in Web3, 14 Oct 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Steven Allen, Protocol Labs – IPLD data in Wasm

Description

Speaker: Steven Allen, Protocol Labs, IPLD
Topic: Operating on IPLD data in WASM

IPLD https://ipld.io/
Protocol Labs https://protocol.ai/
Wasm in Web3 workshop https://avive.github.io/wasm_on_the_blockchain_2021/#/

A

uh um uh So yeah uh I'm currently working on trying to match up ipld, which is the interplanetary link, data system and awesome. uh Basically, ipld is a system for uh I guess it's abstraction over merkel, trees or working like data. uh They like basically allows you to like work with multiple different types of data like jet or ethereum or bitcoin blocks, or even like uh just arbitrary dogs of json.

A

um You can work together by a hash um in a standard way, uh so we use this to like uh have standard transports to operate over where they can transfer and sync uh trace between local computers. um uh We also uh use this to like uh write little selectors over data, where you can actually query video language. We also have languages for um uh sorry. We also have languages for describing different resources and types and out there that kind of stuff. Okay, so that's what I feel deals.

A

The core things you need to know is that there are cids or project ids.

A

A

um Yeah, so I what we're trying to do here is basically uh mash up, but then hyperled and puzzle. uh So instead of just working over like just like links and stuff like that with you usually function, types that you have.

A

Instead, you operate over ipl data, so it's actually giving these structure types, but it's cool because it actually lets you sort of like look to an entire like tree of data, um which is not something that you using otherwise uh and basically, you take the function to take an input cid, which is a quantum type, uh and you have an output, cid um or an output.

A

uh If you want to have a method or define a method, uh you would take in an input state and an input parameters as again objects, and then you spit out a new state and a new output. So, for example, we're planning on using this uh to define a virtual machine for um um yeah we're also planning having epm support uh yeah. Can you actually hear me where is it yeah, but there is some like beeping sound on the background and it's quickly walking down the street. I'm really sorry, no worries.

A

That's a digital! Nomad live, we know it yeah. Sorry, this is going to be live. Okay show us some uh some of the views of the city. Okay, uh so hopefully it's a little bit better. Let me go I'm running okay.

A

Hopefully that's a bit better! Now, uh construction. Okay, um again, I'm really sorry about this uh yeah. So so, basically, yeah like our goal, is to build a something uh vm on top of this, but our goal is also to be able to use this for like like happy fest in general.

A

um uh So like we already have this problem and I think that's where, like um we have to like you kind of have to bake everything into the protocol um where instead, we would actually shift code to people, uh so our goal is to be able to basically ship wasn't modules as I build the objects, exactly truths um uh and transfer them uh with function or dressing, um uh and they can just buy things later. That's not entirely the point of this talk, which is uh next slide. Please.

A

uh Yeah, so this is trying to cover like why we want to do this um again. What I talk about here is like with ipld data, uh it's all network transparent, because everything is always serialized, it's a bit of a problem as well, uh but it means like you, can have these functions and, like your methods like like at every point, you always have the serialized data. You can always translate between nodes.

A

uh It's also recently efficient in certain cases um where, for example, because like you can link to other objects, only the change data needs to be transferred. So, like you can have you can point to a very large things same with ethereum. You can point to your state tree. um uh If you only actually like move around to change pieces, um the other cool thing is like.

A

If you have proper support in your vm, you could laserly load code uh so, for example, a module that makes another module by cid and it can lastly load the code on demand as necessary.

A

um Finally, uh yeah you can pass on large databases about copying, so whatever hopes here is like um I can have my smart contract. My smart contract can call someone else's smart contract and like pass in some sub dag in the state have that smart contract operate on the subject of the state uh and then return like the transform tag.

A

um The evm supports right now, where, like I, can call someone with a delegate call, but then I give them full access to everything where I can send them data and then receive the data back, um but then, like I have like pay for sending everything, but in this case I can kind of say, look, there's a reference to all of my data. Do what you want for it?

A

I can get it back, then I can verify they did it correctly and then link it back um yeah, it's secure because everything's content addressed, uh I see id everything is also structured. Unlike unstructured data like um uh like digital arbitrary bytes, so you can actually explore the data understand. It depends on what you're calling draw. That's, that's really why next slide please, uh but why not is inefficiencies? uh Everything is always serialized and always hashed. um uh The impact of this depends on the grain of literature functions.

A

If you have like triangle functions, it's not going to work if you have to sort of like big things like contracts, it works better, um but this is also optimized away in many cases. uh So the hope really is that we'll be able to find ways to do that, um but yeah. That is really what I want to talk.