How's the race between IPFS and Dat Protocol these days? I keep waiting for one of them to pull ahead and solidify themselves.
Not that popularity matters exactly, but I used to be super hyped for IPFS but my fervor stalled when it *felt* like the team got super distracted by FileCoin and IPO. Combine that with technical hurdles like the time it takes to spread IPNS updates and i started to pull back.
These days i'm wondering if a less grandiose toolset like Hypercore might solve a lot of similar problems with better usability and short term traction.
Interesting nonetheless.
edit1: Been googling this question to answer it myself, and i now realize Dat _is_ Hypercore .. i think. Ah name churn. Lol.
Yep, there's been some name-churn. Dat the name is no longer used, and all the tech that was powering it is now under https://hypercore-protocol.org/
From an architectural level, IPFS is more focused on the "one unified namespace for everything" application. IPFS is designed to transparently share blocks of files no matter what namespace/key that block lives under. For example, many projects have the MIT License in their source directory, so when retrieving that file, any node that has pinned a project containing that file can help seed you. This is handy from a sharing point of view, but also potentially difficult from a privacy issue.
The hypercore ecosystem is more focused on private / invite-only workspaces. Peers only exchange data with other peers in the same hyperswarm address.
Another difference in focus is mutability - Hypercores are designed to be updateable by exactly one person (an append-only log), and IPFS is designed to be updated by zero people (an immutable tree).
I haven't worked much with IPFS, but Hyper* makes it easy to mix and match only the modules you need for your application. There is some cool software and demos out there that use only hypercore and hyperswarm modules, and bypass the Hyperdrive file-tree representation entirely.
Not sure when you last paged into IPFS - the team has made some pretty awesome performance improvements in the past year and seen strong adoption to match (ex, nearly all NFTs are stored on IPFS today).
Definitely check out the IPFS 0.9 release from last week if you haven't seen it! Has some MASSIVE speed ups to IPNS propagation - we're talking less than a second resolution now! There's also a new experimental DHT client that can provide 30M records in about 2.5 hours (which the folks at nft.storage are using reliably at scale). -- https://github.com/ipfs/go-ipfs/releases/tag/v0.9.0
I’m still consistently impressed with the work Maf and Andrew are doing on Hypercore. The next major release will include multiwriter and has some nice networking-layer improvements, and I really like how they’ve implemented the updates. Don’t know the timeline for release but it’ll be this year.
Both are good tech, both are still in dev, so whichever floats your boat
Please, someone help me understand. Why would I want this instead of ENS or a traditional domain?
Let's assume that we are interested in decentralization for the sake of censorship-resistance and lets imagine I want to run a website to allow people to share "extreme" political opinions that might not otherwise be normally tolerated on the clearnet web. I set it up with IPFS, and then I register:
1. a traditional domain name
2. ENS (Ethereum Name Service) domain
3. a Tor Onion Hidden v3 Service
4. IPDS "IPFS Subdomain"
After attracting attention of the FBI, the traditional domain gets redirected to an FBI seizure page, presumably.
The ENS domain is governed by a contract that would require a manual M-of-N override of the ENS contract. (I'd love to see a threat analysis of ENS from this angle). I've never heard of this happening and it seems like it would be a very notable event.
The hidden service is, short of attacks on Tor, quite resistant (though usability is quite poor).
And then finally, this submission, IPDS -- I'm not sure I believe that they will be any better than a traditional domain (given the warning at the bottom, and on the interest form).
Can someone help me understand? EDIT: The more I look at this, the more I'm struck by just another sign of how immature the IPFS ecosystem is. Seemingly destined to NIH every last thing. How can you put this landing page up without any other details or acknowledgement of alternatives and expect buy-in?
EDIT2: of course they're collaborating in a cleartext Telegram room. People, get your stuff together, open source collaboration belongs on Matrix. Not Discord. Not Telegram. Period.
Hmm, I wouldn't assume the goal here is censorship resistance (esp given IPFS+ENS already fits that niche better). Looks to me like someone in the ecosystem is seeing the recent massive speed up to IPNS publishing (launched last week) and testing the waters on whether an "IPNS+DNS as a service" site would be useful to the folks that host dweb sites on IPFS.
I think this makes sense. Another sibling comment implied that this was a product-ified rollout of existing tech, which is of course a useful thing if you're into dnslink. :)
It probably just is DNSLink, only instead of you going out and buying your own domain name and setting the record, they create a subdomain and set the record for you.
Domain Name System for IPFS. You can get a subdomain for your Dapp
IPFS is one of the coolest things out there, but as much as we like it we still need a human-readable name for the URLs as a Source of Trust, this is why we have built IPDS.io
You can have a domain with us but only if you dox yourself to Google to give us the name to hash mapping for it. (The signup is via Google Forms and asks for email and requested subdomain and hash.)
No about, no technical info on how it works, the only contact is through telegram. It seems something related or even made by the IPFS team, but it's not.
"We are hiring a solidity developer": how is this related to IPDS?
TL;DR: The website does everything possible to look like scam.
That is correct. If you have the IPFS Companion[0] browser extension installed then it can automatically recognize URLs of this form and rewrite them to use your local IPFS node instead of the public ipfs.io gateway. The advantage of using https://ipfs.io/ipfs/... in your links instead of ipfs://... is that the former still works even if the visitor doesn't have the extension installed.
I saw a pirate site trying to use it as a DMCA resistant CDN for movies. It was a disaster.
Their idea, I think, was that if their site got taken down, they could easily pop back up at another address so long as the big heavy movies were safe and they just needed to make a copy of the HTML site.
So long as the movies were pinned by more than one IPFS server, they would remain available at their IPFS hash, even if one server was subject to a takedown.
The site could then link to the hashes via a public gateway and use an embedded player to play the movies all in simple static html. Very light to copy around.
1. It turns out that the public gateways were not actually willing to handle that amount of traffic for free. Over a few months they got blocked by all the good gateways. They could have set up their own gateway at this point but the following happened so the project died for other reasons.
2. I think the admin was hoping people would pin (seed) the popular files on IPFS and share the server load. This never ended up happening, so in the end there was just one server 'hiding behind' IPFS, burning through bandwidth. It ran out of bandwidth and the (college kid?) trying to run it ran out of money.
3. IPFS isn't designed for anonymity. When they got enough traction, they got a letter and shut down.
This is why ipfs probably isn't the right platform for libgen really.
But if you were, say, The Pirate Bay and all you are pinning is the torrent file then it seems like that might have worked as torrents are only a few kb in size rather than a Gb movie file? What was the name of the site?
If I'm being totally honest, I'll believe IPFS is worth something when I see a website open seamlessly and feature-equivalent to the real one. A distributed archive.org, but not a static totally nerfed version of it for example.
Yes. Millions of dollars worth of ETH are transacted through NFT platforms every day. Whether you personally consider that useful is up to you, but people are using it.
An incredibly rich ecosystem. I would recommend checking projects such as Uniswap, Aave, Sushiswap, Compound finance, Maker DAO, Yearn finance, Balancer, Synthetix, Alchemix, etc. There are over 52 billion dollars worth of capital being used in these projects today https://defipulse.com. Uniswap v3 is an extremely sophisticated dapp being used today https://uniswap.org/blog/uniswap-v3/. Their code is a masterclass in solidity optimizations. dapps on Ethereum are thriving
I'm always happy to see projects trying to make a decentralized future.
I'm always sad to see the critical , pedantical side of HN's userbase point out that few ecosystems are completely decentralized and then claim that ruins the whole enterprise.
Similarly, when someone wants to run free opensource OS people will criticize them for relying on closed source hardware or closed source communication networks.
Well, dredging up useless facts from the old cobweb attic, back in the 80's if you wanted to have a domain name you had to give Jon Postel a ring[0] and ask politely for him to manually edit the zone.
I suppose you gotta start somewhere, "worse is better" MVP and so on.
I’m not 100% sure what you mean, but hopefully my answer is still helpful. Content routing is performed by a DHT, so nodes do indeed maintain a lookup table [0] from hash to “provider” (i.e. an address) but a given node does not store all such mappings, only a subset.
[0] It is possible to run a node in DHT “client mode”, in which case it stores no mappings and relies on peers to resolve content ids.
Does this answer the question at the level you were hoping for?
Not that popularity matters exactly, but I used to be super hyped for IPFS but my fervor stalled when it *felt* like the team got super distracted by FileCoin and IPO. Combine that with technical hurdles like the time it takes to spread IPNS updates and i started to pull back.
These days i'm wondering if a less grandiose toolset like Hypercore might solve a lot of similar problems with better usability and short term traction.
Interesting nonetheless.
edit1: Been googling this question to answer it myself, and i now realize Dat _is_ Hypercore .. i think. Ah name churn. Lol.