IPFS, CoinList, and the Filecoin ICO [audio] | Hacker News Reader

IPFS, CoinList, and the Filecoin ICO [audio]

blog.ycombinator.com

170 pointscraigcannon26 comments

I'm excited to see that the Filecoin project is getting an update. It long seemed to be stale, with IPFS being the main priority. While a new shiny website and this podcast is up, the new promised Filecoin paper isn't yet public.

It will be interesting to see how it compares to Maidsafe. A notable difference is that Filecoin uses a transaction ledger, while Maidsafe promises to use groups of nodes 'close by' to maintain a consensus: https://blog.maidsafe.net/2015/01/29/consensus-without-a-blo...

A lot of the early discussion about P2P dark periods is largely ignoring the usage for internal networks rather than being the basis of customer facing applications. Many large companies (fb, twitter) use P2P protocols for internal distribution of resources. They can make far more assumptions about network topology when they control the network thus these aspects of P2P algorithms don't get applied to the general internet. I'd argue this period was when people really figured out the value of BitTorrent and decided it was often too difficult to expose to the end user.

FWIW, I built a BitTorrent tracker "framework" (chihaya.io) that a few orgs use to get the advantages of P2P with whatever extra middleware they need for their use cases.

Nice. By the way, it would be interesting to see a Bittorrent tracker ported to WASM, so it can run in the browser.

Have any of these storage coins figured out how to deal with the fact that it's easy to collude with yourself and efficiently pull off a Sybil attack to make it look like you're hosting lots of data?

That's precisely one of the innovations in the new Filecoin paper :) -- we call it Proof-of-Replication. Papers coming soon.

On Sia, contracts are direct been renters and hosts. Hosts can collude with themselves to make it look like they are storing lots of data, but to what end? Renters on Sia only pay for the data they are directly storing.

For the Sybil attack, Sia uses proof of burn. Hosts provably burn money to demonstrate legitimacy. A Sybil attacker would have to spend real resources to appear as many hosts, and as the network grows, the required burn grows as well.

Then why use the Sia coin instead of just paying someone to (provably) host your stuff using Bitcoin? Not being snide, I just don't know what the platform offers if you're just doing direct contracts anyway.

In Sia both the renter and the host pay up front (host is putting up collateral). The payment is held for the duration of the contract (typically 12 weeks), and then the host only receives the full payment if they complete a proof of storage on the blockchain.

This means the host is guaranteed to get paid even if the renter disappears forever on week 2.

You can't set that up on Bitcoin.

Sia is quite volatile. If the payment is held for 12 weeks, the host has no idea what that payment will be worth. Do you have a mechanism preventing this issue?

You know what it is worth at the start of the contract, just like when working with foreign currencies. But yes, volatility is an issue that I presume will be less of an issue as it grows.

One thing I don't fully understand about Filecoin (or siacoin for that matter): can I utilize this as a service provider to directly host and deliver content to my users, or do I need to sit in the path of the traffic, and retrieve the content for them? Or does this not fit into the traditional hosting model at all? Filecoin mentions the retrieval market, but I'm struggling to find more details.

A concrete example would be a file or image hosting service. I can pay to store my content using Filecoin, but am I required to retrieve the content and transfer it to my users?

Sia has not finished this part of the code, but in Sia it's possible to operate like a CDN. The user asks you for content, and then you point them to the host that they download from.

The user still needs to be told where the file is, but this info is a few kb, and the file can be any size. The download is direct from host to user.

Very cool. That means you could have your webserver grab the URL from sia, and render it into the website (assuming users can download the content via normal http).

Is there documentation on this anywhere yet?

You won't have to sit on the retrieval path with Filecoin. Nodes participating in the retrieval market will probably simply expose a Filecoin-enabled version of the HTTP-to-IPFS gateway. The v2 paper will have more details very soon.

Interesting to note is also that, if you want to deliver content yourself, you can just run IPFS nodes on their own, without Filecoin. IPFS has been working fine for the past two years.

I can't wait to try this out. Sounds very interesting.

IPFS works but its not really fine - trying to use it on a small scale (home internet connections) has proven completely impossible for me. It reliably saturates my upload despite not actually serving any content - the DHT connection overhead in every mode is just too high.

This would be fine, but the whole point is decentralized internet - if I can't have a node running at home, or more specifically a node running on pretty much every device I have, its not practical.

Yeah I feel you, the management overhead is a big pain for many people at the moment, ourselves included.

There's a couple of improvements to this coming in the next two months: 1) a new transport (QUIC) which has less overhead, 2) smarter connection handling which closes unneccessary connections earlier, 3) more efficient announcements of who-has-what.

How can Filecoin compete with Amazon?

- Not having ridiculous profit margins on bandwidth

- Having the profit margin of a commodity (storage), instead of the profit margin of company in monopolistic competition (GCP, AWS, Azure)

Why would you imagine it hard to compete with Amazon?

How would availability of the data compare? Will the stored files be available at all times & under high loads?

There are tiers of hosts/miners with different options.

Thanks Craig and YC. Ton of fun.

Out of curiosity, can you clarify how Filecoin is different than Sia or will I need to wait for your new whitepaper?

There are a ton of technology differences which the upcoming Filecoin v2 paper will address (very soon!).

On the conceptual layer there's a difference in that Sia and Storj seem to be a bit more geared towards consumers and file storage, while IPFS+Filecoin's aim is to be infrastructure in the commons.

(Disclosure: IPFS & libp2p dev)

I thought Sia is focused on the enterprise?

Either way, thanks for replying. I'm looking forward to reading the new whitepaper soon.

same here. thanks for stopping by!

Reply on news.ycombinator.com