Hacker News Reader

Comment by rhodey | Hacker News Reader

This is really unfortunate. When cloudflare dropped IPFS you could say this next step was sort of already on the way. I may be biased but I think when IPFS decided to put so much time into "IPNS" in order to support non-static webapps years ago what they came up with did not fit the need. And without webapps on IPFS things were going nowhere.

A year or so ago I wrote IPFS-boot which allows serving webapps on IPFS while providing also an update path and without breaking content hashing:

https://github.com/rhodey/IPFS-boot

But now if you want to serve a secure webapp and not use IPFS IMO the only option you have is to tell users to install Tailscale and to host the webapp themselves and then to install Tailscale on all devices.

View on HN

oh man and when Brave removed native IPFS support. https://github.com/brave/brave-browser/issues/37735

Oh wow they did? I had no idea. Unfortunate

I agree that IPNS has always seemed a bit naff; but alternatives naming systems can be used too (if your system's name resolver can understand them); e.g. this uses pkarr addresses for IPFS content: http://www.chriswarbo.net/blog/2026-05-08-pkdnslink.html

Regarding an "update path", GNS has support for that built-in; though I've not been able to try it myself, since I can't get GNUNet to bootstrap :-(

I always thought that they should have used ENS instead of whatever coin they ended up making. GNS is cool but IIRC (years ago) it had no incentive mechanism which is I think where a lot of these solutions fall down. Even imaginary "ratio points" that p2p file sharing sites have being using for years work better than "donate your resources out of the goodness of your heart".

Half baked comment but I feel like many of these efforts are too purist. Sure they need to think more carefully about incentives from a systemic angle but also why not accept a hybrid design that doesn't depend on those in the first place? For example plenty of people operate their own nameservers for their own websites to use, or spin up a VPS to support [thing] they make use of. So at a minimum I'm sure webring-style infra where you cover for your friends and they cover for you would work quite well.

For a real world example of a hybrid approach that works see bittorrent where torrents can optionally contain urls that point to alternate download sources in order to speed up the initial seeding phase.

My personal opinion is that any notion of "incentive mechanisms" for IPFS is bullshit: whether Filecoin, or otherwise.

There is already a perfectly good incentive mechanism: if someone wants to host some data, they can; and if nobody wants to host it, it doesn't get hosted.

That's worked perfectly well for HTTP, without any shitcoin needed. The advantage of IPFS is that its URLs resolve to anybody who cares enough to host that data; so links will only 404 when nobody cares (unlike HTTP, which relies on a single point of failure).

Did it actually work for HTTP though? Look at the modern web and tell me it's perfectly well.

I see this very similar to IPv4 and IPv6, and adoption. Non content addressable URL addressing simply is "good enough" for most use cases, and bittorrent is "good enough" for serving content durably and somewhat in a content addressable manner (file hashes, magnet torrents, immutable torrents). Do we need IPFS URLs? It doesn't appear so, it seems like a solution seeking a problem. IPFS gateways will always be a target for abuse and copyright claims as well.

If you want to reference resources securely, https://developer.mozilla.org/en-US/docs/Web/Security/Defens... is available and widely supported.

I think something like IPFS is definitely needed. There is nothing stopping you from distributing a webapp on bittorrent but you'll also need to add a README for instructions on how to run the webapp on localhost so a browser can render it. So thats not something that any non-technical user wants to do.

Can you share potential webapp use cases that would rely on IPFS or using IPFS for referencing resources would be an improvement? Perhaps my mental has gaps on this topic, and my thoughts and assertions could be incorrect.

Content-addressable storage as a whole is a non-starter with non-technical users.

I don't understand. Billions of non-technical users are loading URLs like `https://cdn.foo/?guid=deadbeefdeadbeefdeadbeef` all the time. Why would they care if those URLs became `ipfs://deadbeefdeadbeefdeadbeef` instead?

The number of normies visiting urls directly from a CDN is a rounding error.

Non technical users like pronounceable domain names. It's why we aren't just using IP addresses.

CAS is intimidating to non-technical types.

Pronounceable names can point to a content address, the same way they can point to an IP address. For example, https://dnslink.dev points a DNS name to a content address; git refs point to git objects; etc.

This is irrelevant. The point I'm responding to is this:

>I think something like IPFS is definitely needed. There is nothing stopping you from distributing a webapp on bittorrent but you'll also need to add a README for instructions on how to run the webapp on localhost so a browser can render it. So thats not something that any non-technical user wants to do.

This quoted comment assumes there's a web app that is sufficiently complex as to require a readme to run it, but that's too complex for normal people.

But somehow, a link that looks like `ipfs://12340-9q5ui1j2;lk3rjq;ladksfj` won't scare those same people off?

Color me extremely skeptical.

> a link that looks like `ipfs://12340-9q5ui1j2;lk3rjq;ladksfj` won't scare those same people off

Why would non-technical users be looking at raw URLs? Surely links "look like" blue underlined text; or a button; or whatever?

>Why would non-technical users be looking at raw URLs?

Surely you jest. Have you heard of this thing called the address bar? Do you think non technical folks don't know what "Google.com" is?

How many normies working office jobs get it drilled into them "don't open links if you don't recognize the domain"?

How do normies share urls? They paste them into Whatsapp group chats.

Arguably that's a browser shortcoming that applies equally to IPFS (as well as all the other "alternative" internet protocols). (But yes I realize IPFS is a much better fit for the purpose when you dig into the details.)

My go to example: IPFS is Plan9 to bittorrent's UNIX. Some achingly beautiful design, willingness to completely disregard existing messy conventions(for example: The ipfs unified url format) and, like you said, still not enough better for people to want to change.

Why do you need Tailscale?

Edit: Okay, so a couple of people have downvoted but no-one has answered.

Why would you use tailscale to secure a web app?

I think parent is talking strictly about P2P, so you'll need (typically) something to help you tunnel between clients, as there are NATs and all sorts of things between two typical consumer computer at two different homes.

Ah, okay, but that sounds like an incredible pain in the arse to do no matter how you approach it.

libp2p (what IPFS uses) and Iroh does make it relatively painless :)

Reply on news.ycombinator.com