Unleash the Distributed Web

David Dias (@daviddias)

dependency on the web causes problems
web moved from distributed to centralized due to politic decisions and convenience
resource discovery problem
- web is fragile due to location addressing (domain => resolve to ip address => contact ip address)
- multiple unicasts are expensive (100 clients accessing the same 1 resource cause 100 sends)
- problems with the web:
  - disconnected
  - bandwidth
  - offline
  - performance
  - security
  - IoT
  - control
IPFS (the interplanetary filesystem)
- protocol to upgrade the web, basically a big merkle tree
- features:
  - links do not break
  - verify what you recieve
  - works offline
- reference element by cryptographic hash (=> MERKLE LINK, see Merkle Tree)
- IPLD: InterPlanetary Linked Data => links are hashes
  - http://example.com/hugo resolved to http://10.11.12.13/hugo
  - /ipns/example.com/hugo resolved to /ipfs/QmW98oJrc5FT6/hugo
- distributed resource discovery
  - p2p routing overlay over IP network => not bound to ip addressing scheme
  - announce file possession to close nodes (e.g. hash(f) == 91, announce to 91&plusmin;5)
  - find files by looking in nodes close to hash
  - libp2p is the network stack of IPFS
distributed web applications
- use cases
  - content distribution
  - sensitive data (b/c data is verified)
  - huge data sizes
  - webapp w/ components from different endpoints

Provide feedback