Notes on a topless echomail distribution system. By Michiel van der Vlist, 2:280/5555 Fidonet echomail distribution is traditionally based on top down structures. While Fidonet itself was designed as a peer to peer network, echomail terminology is full of terms like uplink, downlink, hubs and backbones. For a specific echomail area, a node has one uplink and zero or more downlinks. Peering links are only found at the top of the so called backbone if it consists of a backbone ring. AKA a fully connected polygon. Such a rigid structure is needed to avoid so called dupes. Messages arriving more than once. Dupes can happen when there is more then one path between nodes. One way to prevent dupes is to ensure that there is only one path between the downlinks and the uplinks. Backbone / \ / \ / \ Hub Hub / \ / \ / \ / \ node node node node Another one is dupe prevention by seen-by. A----------B | * * | | * * | | * * | | * * | C----------D These four nodes form a so called fully connected polygon. They are all connected to each other. The seen-bys prevent that a message from A arriving at B will be sent on to C and D, because they are already in the seen-by. There is no theoretical limit on the size of a fully connected polygon but there are practical limits. Five is about the practical maximum. A standard echomail topology is a backbone of one to five nodes forming a fully connected polygon. Each backbone member feeds a number of hubs who in turn feed a number of downlinks. Such a configuration works fine for preventing dupes. But is has drawbacks. It is a top down configuration and one of the drawbacks of that is that it has a top from which things can be controlled. The most recent example was in the summer of 2011 when the North American Backbone tried to grab control of Fidonews, causing a schism in te process, but there have been many other examples in the history of Fidonet. The R28 CSO wars come to mind. The cost sharing wars ended when the main distrubition went from POTS to Fido over IP, reducing the cost to almost zero, so there no longer was any cost to fight over. We thought that freed Fidonet from the power of the distributors to control the distribution of echomal, but unfortunately it didn't as the 2011 Fidonews coup has demonstrated. So in the summer of 2011 the NAB set all of Fidonet outside of Z1 to read only for the Fidonews echo. And than something unexpected happened. To circumvent the censoring, sysops all over the world starting laying new links. Without any coordinations whatsoever. Just find a link that is not censored. They knew this was going to cause dupes, but the idea was to first restore the links and worry about the dupes later. What evolved was what I have named the NADS, the New Anarchy Distribution System. Some have called it GONADS which is quite appropriate as it takes balls to knowingly and willingly break a 25 year old taboo: that of causing dupes. Up until then that had been an almost mortal sin. Yes, there were some dupes. But not as many as was feared. Not by FAR as many. So then the idea arose to lay some more links. And some more. Against inituition this did not cause an increase of dupes. In retrospect this is not so strange. When everyone links up with everyone we are back to a fully connected polygon. So when approaching such a configuration one can expect the dupes to go down again. And there we have it. The NADS. Others have called it the Fidoweb. Maybe a better term as that is what it looks like. The result is a topless distribution system. No "top" to control it from and as a side benefit, it is far less sensitive to SPOF. If there is a path that dupes can follow, there is an alternate path in case a node drops out. The price is a manageable increase in traffic. In the POTS age, that would have increased the cost. In the Fido over IP age: who cares? Originally published in Fidonews 29:40. 2nd edition nov 2013 ¸ Michiel van der Vlist, all rights reserved.