Seems quite promising! I heard it could use more maturing.
I tried it out a while back but wasn’t too impressed, it still has a ways to go. During testing what I did was write a script to pin (aka “seed”) any public torrent downloaded in qBittorrent into IPFS. My goal was to see if anyone ever found/downloaded my pinned content via the available IPFS search engines.
But the reality is that the available IPFS search engines were crap, they sound nice in theory but are so bad at finding/indexing anything. It was rare that anything I had pinned in IPFS would show up in a search engine let alone someone find it & attempt to download it themselves. There’s still a lot of work to do.
The IPFS software itself also had a lot of performance/memory leak issues, I never could get it to run long term before it crashed & I’d have to figure out how to restart it or wipe its data & start over.
The other issue is that the IPFS project sort of feels like it’s treading water. The IPFS devs went on to create Filecoin & seem more focused on that nowadays. Think of Filecoin as IPFS + cryptocurrency, so you have the privilege to pay people to pin/host (“seed”) your data. And to top it off the Filecoin version of the IPFS network is incompatible with original IPFS network. Funny since it’s the same devs but also is a bit illuminating that they purposely designed it that way.
I’m currently building a peer-to-peer search engine for it. I will be uncensorable and private, as all things should be. We have seen nothing of what can be done on IPFS yet
Copyright lawyers have already found IPFS. The old internet censorship works fine. IPFS produces a list of IP addresses that host content with a certain hash.
Download beemovie.mp4 or governmentsecrets.docx, hash it, and you find a list of all IP addresses that currently claim to host a copy. DMCA the lot of them, or go after their uplink through their ISP, and you’ll take down most of the network. If that fails, you can pick out the home users and start your run-of-the-mill copyright lawsuits.
The concept of IPFS is great, but it’s not uncensorable or private without the ability to run IPFS servers behind anonymization layers like Tor.
Just needs an anonymisation layer by the sounds, like dandelion for monero
This is what I was thinking about. My search engine will run over i2p and tor when available
Good for you but what about the search engine?
Besides being overhyped basic tech where way more useful and practical solutions existed for decades (Freenet existed since year 2000 btw, and Tahoe-LAFS since 2007), there is nothing private about IPFS. This is a dangerous message to purport.
IPFS is as practically useful as NFTs. No wonder the two crowds connected well!
iroh is an attempt to create a useful and practical IPFS. But none of the bigger practical features is implemented yet. And the design itself doesn’t appear to be finalized. I’m willing to give
iroha chance, although the close proximity to the IPFS crowd doesn’t fill one with confidence.You are right, notice the use of future when I talk about privacy. What’s great with IPFS is that it’s based on libp2p. Libp2p makes it easy to add support for any transport protocol you like. If your transport protocol is private (tor, i2p…), then the whole protocol on top of it (ipfs, my search engine) is private. I’m pretty sure it was a choice for them to refuse adding support for private transports, because they don’t want illegal activity on ipfs, thinking it’s too early. But it’s inevitable in the long run
There is no need to talk about an imaginary version of IPFS. GNUnet already exists. You can add that to the list of actually superior technologies that long predates IPFS.
As I mentioned, IPFS is nothing but very basic tech that got overhyped to junior/uninformed developers, and crypto scam victims.
In my experience: hosting a server takes tons of CPU for some weird reason, exposes your IP addresses along with a queryable list of stuff you host (or even have downloaded, depending on your settings), has huge latency issues, and is extremely limited in its throughput. Its peer-to-peer nature allows for much easier privacy invasions by nodes looking for people interested in certain content.
There are reasons why Brave defaults to an IPFS gateway rather than a direct IPFS client if you try to navigate to an IPFS URL. You can read them here.
That seems to be a pretty serious set of problems.
It depends on your use case. If you just want your website to be available in different means and don’t care about your IP address being exposed, IPFS is absolutely perfect. If your website ever gets popular, the network will act like a free cache and everyone has a better time. It’s also useful as a backup in case the web server ever goes down, because other nodes may have a copy even if your server bursts into flames. Slow access to web pages still beats no access to web pages!
If you want to browse the web privately and want quick, interactive, responsive websites, try a public resolver or don’t use IPFS. These are not things IPFS excels at right now.
As a user, I don’t want to share my downloaded images if people can use that to datamine my Lemmy browsing, so I wouldn’t use it.
Yes, instances could do the bulk of the sharing, but then that’s just downloading and rehosting images with extra steps.
Something like Veilid would be more interesting.
Veilid wasn’t released yet when I wrote that comment. It’s definitely interesting.
I agree that IPFS is rather suboptimal for social media and self hosted stuff. The current main use case, being a place to store the data attached to NFTs accessed almost exclusively through public gateways, works fine though.
I feel like some of the major scaling issues that lemmy has right now might be aided with IPFS in some capacity
It’s in a wierd place where unless you can guarantee that other people will be hosting your content, it’s worse than a direct download server. I’m in the process of building a system similar to torrent private trackers where you have to seed to learn CIDs of the content you want (yes, if you know them already I can’t stop you, I know.) And that helps distribute the load. We’ll see how it goes.
I’ve messed with it in the past, found it interesting but too complicated and slow. Not sure if it has any future whatsoever other than as a curiosity.
The concept to distribute files is nice on paper, but in practice it’s not that great.
File discoverability is poor, most people will not know how to act as a node and mirror files, and there’s no builtin privacy protection in place and it’s quite easy to figure out which IP addresses are hosting something.
That said there are some nice projects out that leverages IPFS, such as OrbitDB and Wikipedia on IPFS, which by its decentralized nature makes it harder to censor.
You could also dedicate storage and bandwidth by joining a collaborative cluster, which will automatically stay in sync with the master node(s)
Thanks for asking this, one of the devs in the AMA was talking about Bittorrent for sharing the serving of instance-hosted files in a way that made me think “sounds like you want IPFS”. Now I’m not so sure.
Trash. I tried to download a few TB of images on it and it has horrible performance.
- It trashes spinners
- it’s garbage collection is delete everything. No joke when you have more than x GB they just delete everything to reclaim disk space. I ended up using ZFS volumes and just nuking the disk as it was faster.
- Networking code is garbage with no limits.
- CPU hog
A no longer developed, but much more promising and private alternative to IPFS: Theseus
imo its a nice way to make p2p hosted content available as a simple http link using an ipfs gateway (like ipfs.io or cloudflare-ipfs.com)
