3 ms·

> HTML has some serious problems with that regard I think the problem is not with HTML but with HTTP as location-addressed protocol. For future-proofing, and D

> HTML has some serious problems with that regard I think the problem is not with HTML but with HTTP as location-addressed protocol. For future-proofing, and DDOS/censorship mitigation, content-addressed storage (DAT/IPFS/Torrent) makes sense. I would love to seed my favorite blogs, if i was given the possibility to do so: in this sense, a web browser based on IPNS would be rad: too bad the only one i know of is bundled with JS (instead of operating a paradigm shift) and produced by an adware company. > How will SPAs fare when the backend server is down in 30 years? How much value will be lost? You don't even need to wait 30 years for SPAs to be broken. By the next API update they may stop working, and subtle changes in browser sandboxing could kill them just as quick.

iggldiggl 5y ago
> For future-proofing, and DDOS/censorship mitigation, content-addressed storage (DAT/IPFS/Torrent) makes sense. Well great, now you can't even change the content one tiny bit (not even definitively benign changes like fixing a typo or a bug in your CSS or whatever) without invalidating all links pointing to it. Okay, so as long as anybody still has that version cached and is seeding it at least the link still works (though over time it's not excluded that that number might drop to zero and the link still break after all), though now of course everybody who comes in gets the old version and might not even new that some update exists. While I acknowledge that the silent updates possible with HTTP for an URL's contents can be a curse as much as they are a blessing, I'm not sure if "absolutely no updates ever" are the right answer, either.
- southerntofu 5y ago
  Very good point, though you're missing a layer of indirection in your thinking. I should have made my point more explicit! Usually you would subscribe to a cryptographic pointer (the publisher's public key) whose value is stored in a DHT. Then the pointer can be updated to point to new revisions of the content/website. IPNS is a famous implementation of that for the IPFS protocol. So as a client you can seed a specific revision of the content, or all of them, or just seed the latest version.

iggldiggl 5y ago

> For future-proofing, and DDOS/censorship mitigation, content-addressed storage (DAT/IPFS/Torrent) makes sense. Well great, now you can't even change the content one tiny bit (not even definitively benign changes like fixing a typo or a bug in your CSS or whatever) without invalidating all links pointing to it. Okay, so as long as anybody still has that version cached and is seeding it at least the link still works (though over time it's not excluded that that number might drop to zero and the link still break after all), though now of course everybody who comes in gets the old version and might not even new that some update exists. While I acknowledge that the silent updates possible with HTTP for an URL's contents can be a curse as much as they are a blessing, I'm not sure if "absolutely no updates ever" are the right answer, either.

southerntofu 5y ago
Very good point, though you're missing a layer of indirection in your thinking. I should have made my point more explicit! Usually you would subscribe to a cryptographic pointer (the publisher's public key) whose value is stored in a DHT. Then the pointer can be updated to point to new revisions of the content/website. IPNS is a famous implementation of that for the IPFS protocol. So as a client you can seed a specific revision of the content, or all of them, or just seed the latest version.