Hacker News new | ask | show | jobs
by jszymborski 3214 days ago
I've been very curious about IPFS, but I can't find any descriptions of the limitations and guarantees of the protocol.

For instance, what is the durability of the files I store there? How long do they exist? How much can I store? What is the availability of the files stored? What about latency?

2 comments

Great questions, and not to be unnecessarily pedantic, but IPFS is not a method of storage. IPFS is a replacement for HTTP and has a lot of attractive qualities and works in production today.

Filecoin is a method of incentivizing decentralized storage on IPFS and many of the questions you've asked are either totally or partially unsolved.

Those questions will ostensibly be solved by groups (possibly Protocol Labs, the creators of IPFS and Filecoin, possibly others) over time. That's kind of the bet you have to make if you buy Filecoin, but it's not a bet you have to make to use IPFS today or build a solution on IPFS that works at least for your ends but not all ends.

Hope that makes sense - it's worth digging in more on the forums themselves as to which of those problems are more or less solved and what the possible solutions are.

What is to stop it from just being used for storage though? I see no reason to just use IPFS as a place to dump static files, besides ethical ones.
Sorry just to be clear - you can literally not use IPFS for storage, it doesn't "have" memory or compute or anything. It's a protocol.
Ipfs doesn't provide a storage service as such, so:

> what is the durability of the files I store there?

You don't get the files distributed automatically. They're stored permanently only on clients which pinned them. (Either via user action, or because it's a storage service)

> How much can I store?

As much as you want. But you don't get mirrors by default.

> What is the availability of the files stored? What about latency?

Depends which nodes copy / cache your files.

Exactly right. Especially the last two points are important: no mirrors by default and depends which nodes copy and cache your files. The big challenge ahead for them is how to get various implementations (Filecoin ideally) to make it simple to get dependable nodes to copy/cache/return/etc your files when needed. I'm very excited about it and think they will get there, just important to understand the present vs. future of what works and may work.