How do decentralized/P2P applications connect to each other?

[Deozaan]

Over the past few months my interest in P2P/decentralized systems has really been piqued.

One thing that has continued to puzzle me, however, is this: How do these applications know how to find each other without some sort of centralized server to connect to and register themselves with? And how would one prevent that single point of failure from keeping peers/nodes from connecting to each other?

[MilesAhead]

This Wiki page explains it better than I could. 

I had a chance back in my 486 days of building and running a small free CORBA implementation.  Amazing some of the stuff you can get for free for Linux.  The fascinating part of it was that it created C++ Objects inside your program.  The server that hosted the object may be on the same or another machine.  Kind of like DCOM only more sophisticated.  In the CORBA scheme I believe there was the ability to use the concept of a cluster to route the resource requests, object creation etc..  But it was a long time ago that I messed with it.  Amazingly my 486 built the entire library and all the test programs from the make file, overnight, without an error.  Not amazing that my machine did not crash.  It was amazing that they totally tested their code in this giveaway CORBA implementation.  Really something.