TeleHash: Distributed JSON Peering Protocol

Re:TeleHash: Distributed JSON Peering Protocol by tlack on Thursday May 20, 2010 @12:09AM (#1363229)

It's a great idea but I feel like this protocol is a bit more heavy than it needs to be.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by jeremie on Thursday May 20, 2010 @06:32AM (#1363679)

I've done my best so far to reduce it to it's most minimal required components to just get the job done, and in the last six months it's slimmed down significantly already, not sure much more can get cut out.

My plan is to get some standard libs and utils written next to hide the frustrating complexity required to build this mesh with peers and proxy the chaos-management for apps.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by grayrest on Thursday May 20, 2010 @12:31AM (#1363255)

Care to expound on that?

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by tlack on Thursday May 20, 2010 @01:22AM (#1363311)

The idea of using JSON in a very informal, global messaging system is a really strong one. Messaging is huge these days but pub/sub systems still aren't as commoditized as they should be.

However, Telehash's reliance on UDP means we'll still be worrying about whether or not our message made it. Its' confusing lexicon (telex? dial? listen?) is unnecessary for simple cases.

Telehash should be simpler and closer to HTTP.

tldr: we spend too much time writing pub/sub systems that talk JSON, but this is too much to learn.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by kqr2 on Thursday May 20, 2010 @04:00AM (#1363491)

I think the choice of UDP is based on the use of Kademlia:

http://en.wikipedia.org/wiki/Kademlia

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by catch23 on Thursday May 20, 2010 @01:51AM (#1363340)

maybe they used udp for speed and for udp hole punching over firewalls.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by adulau on Thursday May 20, 2010 @01:59AM (#1363354)

If you want to use the protocol within a browser, there is no standard (included by default in the browser) UDP socket interface accessible from js. Nice to use JSON but I'm still wondering how they could use UDP from the Browser? using "faked" DNS queries to generate UDP traffic?

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by tlack on Thursday May 20, 2010 @02:18AM (#1363372)

There's a reason most of the world uses TCP.. even for things that sound ideal for UDP.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by agentultra on Thursday May 20, 2010 @08:32AM (#1363845)

When you've got a hammer...

TCP isn't ideal for everything. Ad-hoc networking and broadcast streaming being just two.

There's a reason why UDP was invented.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by jrockway on Thursday May 20, 2010 @10:50AM (#1364152)

There's a reason why UDP was invented.

And that reason was not sending time-insensitive text messages.

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by pohl on Thursday May 20, 2010 @10:05AM (#1364024)

JXTA handles this by specifying neither. Instead they have a "pipe" abstraction. Once peers discover each other they can open a pipe that may, depending upon the network conditions, be a fast TCP implementation (when a direct connection is possible) or a slower, but usable pipe built on HTTP with polling (for being able to traverse firewalls).

[ Parent | on HN ]

Re:TeleHash: Distributed JSON Peering Protocol by jeremie on Thursday May 20, 2010 @06:28AM (#1363672)

Exactly, it's the only option left in order to let apps talk directly across the unwieldily myriad of network configurations (NATs).

[ Parent | on HN ]