Releases: will anyone cut releases at all, or should operators pin current versions and plan around that? Should we expect a community fork?
Infrastructure: the post says Protocol Labs retains ipfs.io, dweb.link and delegated-ipfs.dev. Who operates them day to day after September, and what experience do they have running public infrastructure at this scale?
Direction: when NFT.Storage, Web3.Storage and then Storacha wound down its IPFS service this spring, users were pointed at Fil One (https://www.fil.one/), a paid AWS S3-compatible product. Is something similar planned for the public gateways, or will they stay a neutral public service?
None of this is a knock on the Shipyard folks, who kept these projects alive for years and are offering transition help through the end of September. The community got better notice than when Storacha or Fleek shut down abruptly. I am curious what happened to the independence announced in 2023? If a single funder’s decision can end maintenance, the Foundation looks like an accounting arrangement, not independence.
I just want to understand what’s next. Be honest with us: if the plan is to pretend these projects are maintained with LLMs, say so now, so we can plan migration away from IPFS.
I don’t work on Shipyard or anything related to IPFS, so not privy to any information. I saw someone published IPFS is moving beyond the sponsored gateways | IPFS Blog & News though. Which doesn’t really answer your question but that’s all we have so far as far as official infos go.
Hi, just joined your group. An investigation team I work with were looking at using IPFS as an alternative to YouTube’s algorithms being abused by malicious actors.
We have good funding and can crowd-source.
Can anyone put me in direct contact with someone who can discuss what you would need?
Thank you.
Depressing addendum: not only did core software lose continuity, we can’t even talk about it. Whoever took over broke its Matrix delegation, so people can no longer join #ipfs-space:ipfs.io
This is a bummer. I’ve had a feeling that IPFS was being propped up by a reducing number of maintainers for a while.
The underlying design of IPFS is awesome, however deploying it at any kind of scale has bitten me in a big way multiple times.
File / Data CID’s are complex / impossible to calculate outside of the IPFS ecosystem. (Namely, the only way historically to calculate them was “while adding data to an IPFS daemon”)
IPNS records depend on a single publisher who has the private key. If the publishing node goes down, all is lost
DNS-based IPNS is great, but requires constantly updating CID’s in DNS to update content.
Lots of crypto / web 3 excitement which faded quickly
Pinning large IPFS data sets results in a long and painful pinning process with poor client tracking and a big potential to “fail” due to garbage collection
multiformats are complex at times and require a strong learning curve for basic fundamental things like setting IP addresses
choosing the “wrong” backend repository format will eat your lunch. big server-specific slow+stable formats vs small desktop fast+corruption prone means you better choose correctly or risk data loss.
I’m trying to be constructive with these points. Some may have been addressed, the list was a lot longer a few years ago, however the IPFS team has done awesome work reducing the pointy bits. (Fuse mounts of MFS come to mind)
IPFS is great, and an awesome technical design, but the core use-case really needs realigned around the realities of users.
Although I am using crypto IPFS was never about crypto for me. Just a very convenient way to store data. To store it for free, even. So if I want to upload a file or music publicly I will do it through IPFS 10 times out of 10.
Same. I never really got the brief crypto hype for NFT’s on it. IPFS isn’t magic, and the data has to live “somewhere”
Nice! Yeah, i figured some of my pain points would have 3rd party workarounds by now (though, the fact that a 3rd party workaround is even needed for something so core is an issue )
My list wasn’t a “punch IPFS while it is down” thing, I want to see it successful, just want to keep the “reality of users” visible.
Is libp2p.direct from https://blog.libp2p.io/autotls/ dead dead?
I get ERROR autotls.start client/health.go:114 registration broker at https://registration.libp2p.direct did not confirm it is healthy and no cert.
I just wasted entire day recreating repos just to unblock data export, and new nodes can’t get TLS cert because of the above, so I had to do extra work with reverse proxy just so it works with https://inbrowser.link
I can’t confirm whether the AutoTLS registration broker is permanently gone, but an unhealthy health-check response blocks cert issuance regardless of whether the underlying libp2p software is fine. The distribution host outage is a separate operational failure; it breaks the usual Kubo update path, so pinning and installing from official release artifacts with checksum verification is the safer short-term route. Your reverse-proxy/manual TLS workaround is reasonable. What’s missing is a documented fallback and status page for these shared services, otherwise every operator rediscovers the same breakage.
One thing nobody here has mentioned yet is NFTs. Lots of contracts don’t store ipfs://CID, they store https ://ipfs.io/ipfs/CID as a plain string. Since the gateway retired, those tokens show blank in wallets and marketplaces even when the content is still pinned somewhere. The CID is fine, the address baked into the contract isn’t. If you control the content, pin it yourself and keep CAR exports. If you hold the token, check what it actually points at. I build Xtrata and we made a free checker for Stacks collections that shows if art is on chain, on one host or unreachable: xtrata.xyz/nftcheck. In our census IPFS art was ~2% broken vs 56% for self-hosted servers, so the CID model holds up. Hardcoded gateways are what bite.