Hacker News Reader

Comment by jmeyer2k | Hacker News Reader

I've worked with IPFS and I completely agree with you here. There are so many basic bugs that haven't been fixed because of the over abstraction of LibP2P.

Currently, it connects to bootstrap nodes and tries to discover peers asynchronously. Obviously, it doesn't handshake with the bootstrap nodes in time for it to start discovering peers so it waits another 1-5 minutes depending on the setting to try discovering again.

Startup usually takes a few minutes due to this trivial problem and an issue has been open for it since May: https://github.com/ipfs/go-ipfs/issues/5953

View on HN

IPFS person here - there are a ton of improvements to the libp2p DHT in the works (many which also relate to bootstrapping efficiency), but a key component of shipping these to an active 100K+ node network like IPFS is testing and validation, so it's taking us longer to ship these improvements while we build a testground able to simulate network conditions to increase network _reliability_ as well as performance. IPFS is still actively WIP and directly focused on improving performance, reliability, and usability - so feedback and collaborators welcome to help us address these issues head-on. =]

I like where IPFS is going and I think it's an important bit of software, so it's good to hear you're working on it.

I do have a question though on the immutability, how do you address problems where people contribute harmful content, such as revenge porn pictures, or doxing someone? If the content is immutable, it means these things would always exist and never be removable? Is there any moderation that can stop this kind of stuff?

Just wondering, I still like the idea, it's just a question I have been thinking about.

What IPFS means by 'immutability' is that the mapping from id-->content will never change. That doesn't mean that the content will always be available on the network and accessible to any viewer. If you remove the content from the network (aka all nodes stop hosting the content), then no one will be able to find and access it.

For example, if bad person A adds something illegal to a node they host on IPFS, then person B can legally compel them to stop providing that content (ex by suing them). While the id-->content mapping would still be immutable, the content can no longer be found/viewed because bad person A is no longer hosting it. There is no "automatic persistence" in the IPFS network, so only nodes that choose to host a piece of content will become providers (aka no automatic process will keep bad content around if all hosts have taken it down).

This is actually where I feel like there’s only trivial purpose to IPFS. Any content can still be trivially taken down simply by going after anyone that’s hosting it. Yes, there are legal (and moral) benefits here, but it no longer represents a technology to free information that is restricted for personal or political reasons.

Immutability on ipfs just means a given address will always point to the same content. It doesn't mean that content is or will always be available. Nodes who host harmful content are responsible for hosting it whether it's on a peer-to-peer network or a server/client network.

Unfortunately if you want an uncensorable network you won't be able to censor the nasty stuff either.

This is great to hear, I'd be extremely happy to see IPFS become fast and reliable, and I think it could be a hugely important piece of software.

Absolutely. In fact, I've been following those developments very closely. I hope to see (and possibly contribute to) some significant improvements to IPFS, especially the DHT.

That's awesome! See active work in progress on DHT bootstrapping here: https://github.com/libp2p/go-libp2p-kad-dht/issues/387

I've been working with Dat, which I think is similar to IPFS and really enjoying it. Being able to modify your site after you publish it is hugely useful (previous versions are always accessible - edits are stored in an append only log). It has its own problems (editing from multiple computers is risky but that's being worked on), but in my opinion that's essential for a decentralized web

Been working with libp2p, IPLD, etc (IPFS stuff). It's improved a lot. Still rough on the edges but definitely improved.

Fundamentally, the spec are good. These guys might not have the most mature development practices but it's a completely different category than tor. The project are broken down in a very [theoretically] elegant way and one project that truely exemplifies this is IPLD: https://github.com/ipld/specs

IPLD (among other protocol labs projects) is applicable to so much more than just IPFS!

Reply on news.ycombinator.com