40 likes | 117 Views
Thoughts on dividing the P2P functionality. Henning Schulzrinne Columbia University hgs@cs.columbia.edu. Diversity in DHT. Generally, only super nodes really need to know DHT thus, can imagine “provider-P2P” that runs proprietary DHT Three basic interfaces:
E N D
Thoughts on dividing the P2P functionality Henning Schulzrinne Columbia University hgs@cs.columbia.edu
Diversity in DHT • Generally, only super nodes really need to know DHT • thus, can imagine “provider-P2P” that runs proprietary DHT • Three basic interfaces: • UAs treat SN as DNS server (get SRV) • doesn’t work: SN depends on user, not just domain • UAs have common lookup protocol: user SN • but this is essentially SIP functionality • UAs treat SN as outbound proxy • can redirect or proxy
Two-layer architecture DHT SN Supernodes act as proxies
How to run multiple DHTs • Idea: no perfect DHT provide multiple DHTs • need at least one mandatory-to-implement one • Unlikely to have common information across all plausible DHTs • intersection of information = TCP • no major advantage of common protocol: all the hard work is in the DHT, not the protocol • We have no plausible global resource discovery mechanism to find such nodes • but might be able to have DNS-based “seed” lists: PTR _supernodes._tcp.p2psip.org