IPFS IPFS þing 2022, 10 Aug 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Auspinner - @danieln - IPFS Implementations

Description

Auspinner - presented by @danieln at IPFS 2022 - IPFS Implementations - https://2022.ipfs-thing.io

A

It's uh good to meet and see all of these uh great implementations of ipfs uh daniel, I'm pretty new to uh pl and ibfs. I mean I worked with it on the other side some years ago, as a developer and uh now sort of coming onto this side.

A

As a developer advocate, um I have a, I guess, a fresh and naive perspective, and so uh the goal with ospina was to really um learn about the ipfs stack, the protocols, the specs bit, swap some of the broader specs that we have in the ecosystem, like the pinning api, spec and the go libraries and to explore the space of trade-offs.

A

um I think uh you know uh dietrich's talk uh demonstrated how we want to get ipfs into so many different runtimes, and uh you know it's always a game of trade-offs, um and so I wanted to get a better feel for that and maybe uh be able to then bring that back to also the broader ipfs community.

A

um And, of course, I wanted to build something, useful um and explore this idea of like a short-lived light, ipfs node, so something that sort of has a short lifetime um and so to sort of move on into what actually or spinner is. I think it's a good uh starting point to start with this question of what do all of these have in common? Does anyone can anyone guess maybe yeah something a bit more specific.

A

Yes, pinning api server um spec, so yeah we have this great uh pending service, api spec, and the great thing about this is that it offers great interoperability between the different services. So you know they might actually implement a slightly different http api, but because they adhere to the spec.

A

The http api that adheres to the spec is the same across all implementations, and we have actually discovered that there's some deviations, there's great work by russell in actually checking the compliance of each of these services.

A

But now that we have this compliance check and we have a swagger api, we can build all of these great tooling. That works with all of these different services and give ipfs users more options. And so how does earth spinner work and what does it actually do? Essentially, it allows you to take a car file and then pin that onto um a pinning service, so it takes the car files. Input makes an http request.

A

Hopefully the pinning service will give back uh some delegates to connect to, and then it basically starts a leap. P2P host connects to those delegates um and makes the blocks available over bitswap.

A

There's a little demo here, it's uh one minute so I'll just just to get a feel for what it's like. So you can list um the pins that you have and here I'm actually making a pin request that made the http request. The status is queued uh it connected to the delegates and then it sort of tracked the state.

A

And now we can see that we have this rocky road pin um and then now I'm just getting the uh token for pinata and I'm basically doing the same thing with pinata, so pinning the same file pinata and estuary they're both returning um delegates and spoken to some of the uh dot storage team- and I know you're uh uh going to return the delegates soon so um and there we have it.

A

So it's uh waiting for the blocks to be transferred, and hopefully that will complete yeah there we go, um and so that was it and you can obviously also delete uh pins.

A

So that was it and um those were some of the libraries that I used for this. um And finally, I so this was obviously a learning journey for me and uh you know, as I mentioned you know, not all of the pinning services returns, delegate return delegates which makes it tricky, because the only alternative you have is to pass origins and if you have a short-lived ospina instance that just starts a new lib p2p host, it's probably behind that or a firewall. So that means it's not publicly reachable.

A

So you have to do upnp or you have to do hall punching with circuit relay, and so I got upnp working, but then that sort of breaks this idea of this being a short-lived instance, because well you have to wait for the router to give you uh to do the port mapping. Another thing that was a bit tricky was tracking the pinning progress.

A

So, basically, you have a status that you get from the http uh api and you can just pull the api and wait to see that the status changed um that felt a bit coarse, but that's how it's doing it and if you're obviously doing a big file, then uh you know that's not ideal, because you can't know how far along you are, um and so I did an experiment using the bandwidth counter to basically measure the uh uploaded bytes uh over the basically the lib p2p uh host um and yeah there's some sort of different ideas uh about how to sort of make this maybe a bit more versatile.

A

I did a session with uh jaropo on basically taking a file and then unix effects signing it. I guess that's the word yeah, basically chunking it and turning it into a uxfs um representation or maybe using go ipld, prime and, as I mentioned, better progress tracking and if you have any ideas for how this could be a bit more uh versatile and usable um I'd love to hear your ideas.

A

So that's pretty much it. uh It's been a great learning journey and uh thank you to everyone who helped. uh I think we have a lot of different ipfs implementations and it's really like cool to see this uh uh the discussion around what even is ipfs kind of like expanding as we explore this space. So thank you.