IPFS Supported In FFmpeg 5.1, IPFS Devs Envision Support In More Open-Source Projects

[markg85]

Do all these tools ship with a complete IPFS stack? Do they all have to connect to IPFS and pull the files themselves? Do they have a private IPFS storage?

Or is there a daemon running that connects to IPFS and the curl IPFS support is "merely" redirecting ipfs://QMxxx to localhost:12345/ipfs/QMxxx

Their current implementation is to rely on the gateway. You are heavily encouraged to run your own local gateway. As when you do, ffmpeg for example will detect it and you'd benefit from your local IPFS node (in 0.15.0, it's about to be released in a couple weeks). ffmpeg does warn you when it can't find a gateway and will use dweb.link (run by Protocol Labs, the same folks who develop IPFS) as a last effort fallback mechanism.

Moral of the story is that you can share just a CID with someone. Or `ipfs://<cid>` to be more exact. You should not ever need to provide the gateway URL as that is a single point of failure and might be different per user. That gateway should either be automatically detected on a per user basis or be set by the user (ffmpeg allows for both variants), it should never have to be send back/forth when sharing something.

[cl333r]

Is it correct that IPFS is about storing (media)files found on the internet but not about recreating the webpage/website?
Because I can't see how you could store i.e. youtube in a file since it's a collection of services that create the webpage on the fly - instead you can store a certain video with IPFS - and those who know its hash or whatever can then request it from peers like with bittorrent.

PS: I just heard of it for the first time and did some quick googling.

[bug77]

Is it correct that IPFS is about storing (media)files found on the internet but not about recreating the webpage/website?
Because I can't see how you could store i.e. youtube in a file since it's a collection of services that create the webpage on the fly - instead you can store a certain video with IPFS - and those who know its hash or whatever can then request it from peers like with bittorrent.

PS: I just heard of it for the first time and did some quick googling.

It's BitTorrent with a paid-for tier. As long as someone is still seeding you can get the file. So yeah, it's a file storage, it knows little about how a web page looks like.

[markg85]

Is it correct that IPFS is about storing (media)files found on the internet but not about recreating the webpage/website?
Because I can't see how you could store i.e. youtube in a file since it's a collection of services that create the webpage on the fly - instead you can store a certain video with IPFS - and those who know its hash or whatever can then request it from peers like with bittorrent.

PS: I just heard of it for the first time and did some quick googling.

Partly. For this to make sense, consider IPFS to be like a filesystem. You can add anything you want to it. If you create a site you can also add that. Say you added a folder to IPFS with a structure like this:
- MyAwesomeSite (folder)
- index.html
- etc...

You would then add your folder to ipfs (like `ipfs add -R MyAwesomeSite`)
That gives you back a hash per file and one for the parent folder.
Now you can point your site to that hash. Be aware that you need DNSLink for this (https://dnslink.dev/)

This is all for static sites. If you want a dynamic site (where you need a server component in web2 talk) then it becomes a lot more complicated.

[markg85]

It's BitTorrent with a paid-for tier. As long as someone is still seeding you can get the file. So yeah, it's a file storage, it knows little about how a web page looks like.

That's all wrong, sorry.
- IPFS has no paid-for tier. Filecoin has but as of now, files added to filecoin can't be immediately accessed on IPFS so it would be unfair to consider that the paid tier.
- It's not file storage! You add files to the network (making it know you have certain hashes). The files are hosted on your own pc where you run your IPFS node. If you don't do anything special besides the defaults then this is all there is. You have to manually go the pinning service route to make someone else host it. Or ask other IPFS users to specifically pin your files (which is also a manual task)
- IPFS knows nothing about how a webpage looks. Don't know where you get your information, but that's just plain wrong. IPFS has it's own webserver build in for the gateway. That logic + DNSLink allows you to host a website on IPFS. It has no more web page knowledge then any other webserver has.

[bug77]

I think you got me all wrong, let me try again.

That's all wrong, sorry.
- IPFS has no paid-for tier. Filecoin has but as of now, files added to filecoin can't be immediately accessed on IPFS so it would be unfair to consider that the paid tier.

I meant for advanced usage (save my file here, access it anywhere), it resembles BitTorrent. That you may be asked to pay for such convenience, I have no problem with, I did not mean that in a derogatory sense. Nobody will host your crap for ages, for free, IPFS/FileCoin has no reason do it any differently.

- It's not file storage! You add files to the network (making it know you have certain hashes). The files are hosted on your own pc where you run your IPFS node. If you don't do anything special besides the defaults then this is all there is. You have to manually go the pinning service route to make someone else host it. Or ask other IPFS users to specifically pin your files (which is also a manual task)

Ok, it's a file system, not a storage. But its advanced features do require storage outside of your machine (see above).

- IPFS knows nothing about how a webpage looks. Don't know where you get your information, but that's just plain wrong. IPFS has it's own webserver build in for the gateway. That logic + DNSLink allows you to host a website on IPFS. It has no more web page knowledge then any other webserver has.

I meant there is no "web page" concept in IPFS. In the sense that you cannot tell IPFS "give me amazon.com". You explained pretty well above that it can work as a time machine for the Internet, but it's still working with files, as far as it is concerned.

And while you correctly point out that a lot of things work locally, you can do a lot (all?) of them with any other file system. Or a file system + a web server. I believe the true power of IPFS is sharing.

[archkde]

Not really.
You're pulling it out of context and basing your whole idea on that. Never a good idea.

IPFS works without any coins. If you run a node you're part of the network and can access anything on the network. If you don't run a node you can still access anything on the network but via a so called gateway.

If you add a file to IPFS it's available to anyone on the IPFS network as long as there is someone hosting that file. If that is you and if your node is up: sweet! But there is a real need for files being "pinned" (made persistent) by other people. This improves how fast your file can be found and makes it more resilient. You have these so called "pinning services" for that. Most are free for limited amounts of data and paid for larger amounts. Still none of this includes Filecoin.

Filecoin comes in the mix when you want to guarantee data is stored. It's most comparable to tape backup storage, a cold storage. Anything you store on filecoin is contractual stored on there for as long as you paid. Thus far however, data stored on filecoin isn't "immediately" accessible on IPFS (though this is about to change). This is because the node storing your data needs to do "stuff" to make your data available. It therefore also costs filecoin to get your data (yes you pay for storing your data and for retrieving it). The golden mix is to use services where you can "pin" data on IPFS (thus get it immediately available throughout the network) where those pinning services back your data up on filecoin. There are quite a few services that do this but one such service is https://estuary.tech/

I hope this is more educational then the biased first impression of filecoin being a shitcoin. It might still be a shitcoin, just not to my knowledge.

That "it" was referring to the company's business model, not to IPFS itself. As said, I can see the value in IPFS. I can even see the value in the "pinning" services you described, and in fact my first idea for the business model would have been a paid pinning service, but apparently that isn't the case and filecoin is the business model.

I would not mind a pinning service that you can pay in euro or US dollar. But shitcoins in any form have many times shown to be most useful for speculation, fraud, money laundering, payment of criminal activities or circumvention of trade sanctions, no matter the original idea behind them. I will not support such schemes.

[Mathias]

I would not mind a pinning service that you can pay in euro or US dollar. But shitcoins in any form have many times shown to be most useful for speculation, fraud, money laundering, payment of criminal activities or circumvention of trade sanctions, no matter the original idea behind them. I will not support such schemes.

Filecoin basically is a paid pinning protocol. My hope is that the value of filecoin is not based on speculation, but on the value of storing data. If storing data is comparable to (or cheaper) then backblaze & others, then filecoin is not a shitcoin IMO. If the value skyrockets and it is not used for sane storage, then it should die like all the other shitcoins.

I'm not convinced it even will be a sane way of storage (because of overhead reasons). But if they solve that problem, it might be a good coin. But I also see the misuse of all Cryptocoins for laundering and criminal activities.

[PlanetVaster]

Some comments really overcomplicate this.

Easy analogy:

IPFS is like HTTP (a protocol for transfer of data, in this case peer-to-peer so actually a bit more like bittorrent than http)

Filecoin is like AWS (i.e. completely separate from the former protocol but will be used by many to avoid the hassle of setting up their own hardware and software for an ̶h̶t̶t̶p̶ ̶s̶e̶r̶v̶e̶r̶ IPFS node).

Yes, Filecoin is basically a shitcoin, but you don't have to use it or interact with it at all to use IPFS, just like I don't use AWS because it's also shit IMO, I host my HTTP servers elsewhere.

[mb_q]

Filecoin basically is a paid pinning protocol. My hope is that the value of filecoin is not based on speculation, but on the value of storing data. If storing data is comparable to (or cheaper) then backblaze & others, then filecoin is not a shitcoin IMO. If the value skyrockets and it is not used for sane storage, then it should die like all the other shitcoins.

The problem is that this system has a catastrophic failure mode due to a fact that the only value of filecoin is hosting pinned data, and the only gain from hosting pinned data for filecoin is filecoin. Once filecoin price drops, due to some random event, under the costs of storage, people are going to start dropping pinned data, ripping filecoin out of its only value and trashing it even more. A classic vicious cycle that won't stop until all cold data is deleted and filecoin reputation is lost for good.

There is already an airdrop service offering pinning 1TB "forever for free", https://web3.storage/ seemingly functioning because filecoin is overpriced just for being a cryptocurrency. Not really reassuring for stability.

IPFS is undoubtedly great for distribution (not counting the fact that you effectively publish your browser history by using it), but we sort-of already had the same thing with BitTorrent & magnet links.