4 ms·

Adding layers of indirection isn't going to give us a permanent addressing system. This service just adds new points of failure: the service itself, or the crea

by btrask 11y ago

Adding layers of indirection isn't going to give us a permanent addressing system. This service just adds new points of failure: the service itself, or the creator of an identifier who fails to update it or changes it maliciously. The real solution is some form of content addressing. Whichever one you want: URNs, magnet links, the "ni:" RFC[1], IPFS paths[2], or my own hash link system[3]. [1] http://tools.ietf.org/html/rfc6920 http://tools.ietf.org/html/rfc6920 [2] https://ipfs.io/ https://ipfs.io/ [3] https://github.com/btrask/stronglink/ https://github.com/btrask/stronglink/

javajosh 11y ago
Content-addressing suffers the exact same problems, they are just a little more abstract. Content addressing also contains indirection because TCP/IP is most certainly location-based. So, to map from a content hash to one or more IP addresses requires (drumroll) a service that is a new point of failure. Edit: So, there is certainly an algorithm that lets you go from hash to IP directly in a way that evenly distributes things, blah blah - but then you'll probably end up bugging a lot of hosts that don't participate in that algorithm! I pity the poor fools who get "lucky" IP addresses for certain content. :)
- rictic 11y ago
  IPFS solves all of these problems: https://ipfs.io/ https://ipfs.io/ Data only vanishes from IPFS when no one cares about it enough to host it, rather than vanishing when the original poster stops caring.
- btrask 11y ago
  I wouldn't say it's the exact same problem, because at least the address gives you something to go on (you can verify whatever data against the hash to make sure it's correct and unaltered). But the problem of resolving hashes is something every content addressing system must deal with. The most common answer is distributed hash tables. Personally, I'm fond of local resolution. One of the coolest things about content addressing is that you (or anyone) can choose how to resolve addresses, so an ecosystem is possible (sort of like the Wayback Machine today, or the coexistence of various search engines).
- danieldk 11y ago
  Content-addressing suffers the exact same problems, they are just a little more abstract. Actually, content-addressing already works pretty well. If you paste the SHA hash of a widely distributed file (e.g. an OS ISO or commonly used source package), you typically find many places to download the content. It's often more reliable then URIs for such kind of data. It is surprising that it already works well for the web, which was not built for CAS. Peer to peer CAS systems, such as IPFS make this much more usable. but then you'll probably end up bugging a lot of hosts that don't participate in that algorithm! ??? Trackers/DHTs were invented exactly to make this efficient and doable.
  - gioele 11y ago
    DBJ puts random IDs in his papers and uses that that as a persistent (work) identifier. An example from http://cr.yp.to/bib/documentid.html http://cr.yp.to/bib/documentid.html > Date: 2004.04.02. Permanent ID of this document: 46d904d0613f360904e88a85dcdaa52b. > When I cite a paper with a document ID, I list the paper's document ID in the bibliography entry: > [7] Daniel J. Bernstein. Sharper ABC-based bounds for congruent polynomials. Journal de Theorie des Nombres de Bordeaux, to appear. ISSN 1246-7405. Document ID: 1d9e079cee20138de8e119a99044baa3. URL: http://cr.yp.to/papers.html#abccong http://cr.yp.to/papers.html#abccong. Date: 2004.02.10. > The idea is that, even if the URL changes for some (probably bad) reason, you can still search the Internet for the words > Permanent ID of this document: 46d904d0613f360904e88a85dcdaa52b
    - danieldk 11y ago
      Thanks for pointing this out, that's a nice idea!
    - javajosh 11y ago
      Interestingly there's some ambiguity about which document - is it the one that includes the hash, or is the hash inserted after it is computed?
  - javajosh 11y ago
    > If you paste the SHA hash Okay, where are you pasting it? Wherever that is, it's a point of failure.
- endergen 11y ago
  To your point about 'poor fools who get "lucky" IP addresses for certain content', isn't this what Bittorrent solved by design where popular content is actually spread more quickly as more people spread it out all over and then you just need smart routing to find the nearest cache of that file?

by btrask 11y ago

javajosh 11y ago

Content-addressing suffers the exact same problems, they are just a little more abstract. Content addressing also contains indirection because TCP/IP is most certainly location-based. So, to map from a content hash to one or more IP addresses requires (drumroll) a service that is a new point of failure. Edit: So, there is certainly an algorithm that lets you go from hash to IP directly in a way that evenly distributes things, blah blah - but then you'll probably end up bugging a lot of hosts that don't participate in that algorithm! I pity the poor fools who get "lucky" IP addresses for certain content. :)

rictic 11y ago
IPFS solves all of these problems: https://ipfs.io/ https://ipfs.io/ Data only vanishes from IPFS when no one cares about it enough to host it, rather than vanishing when the original poster stops caring.
btrask 11y ago
I wouldn't say it's the exact same problem, because at least the address gives you something to go on (you can verify whatever data against the hash to make sure it's correct and unaltered). But the problem of resolving hashes is something every content addressing system must deal with. The most common answer is distributed hash tables. Personally, I'm fond of local resolution. One of the coolest things about content addressing is that you (or anyone) can choose how to resolve addresses, so an ecosystem is possible (sort of like the Wayback Machine today, or the coexistence of various search engines).
danieldk 11y ago
Content-addressing suffers the exact same problems, they are just a little more abstract. Actually, content-addressing already works pretty well. If you paste the SHA hash of a widely distributed file (e.g. an OS ISO or commonly used source package), you typically find many places to download the content. It's often more reliable then URIs for such kind of data. It is surprising that it already works well for the web, which was not built for CAS. Peer to peer CAS systems, such as IPFS make this much more usable. but then you'll probably end up bugging a lot of hosts that don't participate in that algorithm! ??? Trackers/DHTs were invented exactly to make this efficient and doable.
- gioele 11y ago
  DBJ puts random IDs in his papers and uses that that as a persistent (work) identifier. An example from http://cr.yp.to/bib/documentid.html http://cr.yp.to/bib/documentid.html > Date: 2004.04.02. Permanent ID of this document: 46d904d0613f360904e88a85dcdaa52b. > When I cite a paper with a document ID, I list the paper's document ID in the bibliography entry: > [7] Daniel J. Bernstein. Sharper ABC-based bounds for congruent polynomials. Journal de Theorie des Nombres de Bordeaux, to appear. ISSN 1246-7405. Document ID: 1d9e079cee20138de8e119a99044baa3. URL: http://cr.yp.to/papers.html#abccong http://cr.yp.to/papers.html#abccong. Date: 2004.02.10. > The idea is that, even if the URL changes for some (probably bad) reason, you can still search the Internet for the words > Permanent ID of this document: 46d904d0613f360904e88a85dcdaa52b
  - danieldk 11y ago
    Thanks for pointing this out, that's a nice idea!
  - javajosh 11y ago
    Interestingly there's some ambiguity about which document - is it the one that includes the hash, or is the hash inserted after it is computed?
- javajosh 11y ago
  > If you paste the SHA hash Okay, where are you pasting it? Wherever that is, it's a point of failure.
endergen 11y ago
To your point about 'poor fools who get "lucky" IP addresses for certain content', isn't this what Bittorrent solved by design where popular content is actually spread more quickly as more people spread it out all over and then you just need smart routing to find the nearest cache of that file?