Repository : ssh://git@open-mesh.org/doc
On branches: backup-redmine/2017-07-13,master
commit 4de7e64f182c35e57fb18b7b1062d0e13944c161 Author: Marek Lindner mareklindner@neomailbox.ch Date: Mon Jun 21 00:11:38 2010 +0000
doc: batmand/Why-starting-batman
4de7e64f182c35e57fb18b7b1062d0e13944c161 batmand/Why-starting-batman.textile | 56 ++++++++++--------------------------- 1 file changed, 14 insertions(+), 42 deletions(-)
diff --git a/batmand/Why-starting-batman.textile b/batmand/Why-starting-batman.textile index bc5830dc..1b71ec0e 100644 --- a/batmand/Why-starting-batman.textile +++ b/batmand/Why-starting-batman.textile @@ -1,44 +1,16 @@
-= Why starting B.A.T.M.A.N. ? = [[BR]] - -Often, we are asked why we dropped OLSR in favor of B.A.T.M.A.N. [[BR]] -To explain the initial motivation in those days to start developing a new [[BR]] -routing protocol optimized for lossy networks, we have summarized some [[BR]] -of those reasons in the following list in no particular order. [[BR]] - - * The OLSR protocol as specified in RFC 3626 was not completely [[BR]] - functional in practical scenarios. Tests in 2004 (namely at the [[BR]] - Wizards of OS conference) showed several issues like routing [[BR]] - tables taking a long time to build up and taking no time to break [[BR]] - down again, routing loops and flapping routes. We are not aware [[BR]] - of any mesh network installation in the field that runs the [[BR]] - protocol according to RFC3626 in a productive environment. If you [[BR]] - do, please let us know. [[BR]] - - * The OLSR protocol belongs to the family of link state routing [[BR]] - protocols which calculates a full routing path to all other [[BR]] - nodes in the network. All routing decisions rely on the fact [[BR]] - that every node has (mostly) the same information at any given [[BR]] - time. The more those information differ between the nodes in a [[BR]] - network, the probability of things like routing loops increases. [[BR]] - In our point of view, it is rather difficult to keep such [[BR]] - information synchronized in lossy environments like wireless [[BR]] - mesh networks, in that the effort needed for synchronisation [[BR]] - increases exponentially with each node. [[BR]] - - * The OLSR daemon as it is today (the one which can be [[BR]] - downloaded from www.olsr.org; we will further reference it just [[BR]] - as "OLSR daemon") comes with a lot of extensions / changes [[BR]] - (e.g. the default config disables most of the RFC3626 features as [[BR]] - MPR and enables ETX, !FishEye, etc) that alter the behaviour of [[BR]] - the routing protocol completely (compared to RFC 3626). [[BR]] - Those extensions have been developed by an active community [[BR]] - (Thomas Lopatic & Elektra and the OLSR-NG project later on). [[BR]] - Therefore the OLSR daemon today is not comparable with the [[BR]] - initial OLSR protocol at all, as many things which had been [[BR]] - specified in the RFC 3626 have been modified. Again, [[BR]] - because of our personal experiences, those extensions help [[BR]] - to decrease the problems inherited by the link state algorithm but are [[BR]] - unable to solve it which led to our decision, that it might be better [[BR]] - to try a different approach. [[BR]] + +h1. Why starting B.A.T.M.A.N. ? + + +<pre> +<code class="div"> + +Often, we are asked why we dropped OLSR in favor of B.A.T.M.A.N. To explain the initial motivation in those days to start developing a new routing protocol optimized for lossy networks, we have summarized some of those reasons in the following list in no particular order. + +* The OLSR protocol as specified in RFC 3626 was not completely functional in practical scenarios. Tests in 2004 (namely at the Wizards of OS conference) showed several issues like routing tables taking a long time to build up and taking no time to break down again, routing loops and flapping routes. We are not aware of any mesh network installation in the field that runs the protocol according to RFC3626 in a productive environment. If you do, please let us know. + +* The OLSR protocol belongs to the family of link state routing protocols which calculates a full routing path to all other nodes in the network. All routing decisions rely on the fact that every node has (mostly) the same information at any given time. The more those information differ between the nodes in a network, the probability of things like routing loops increases. In our point of view, it is rather difficult to keep such information synchronized in lossy environments like wireless mesh networks, in that the effort needed for synchronisation increases exponentially with each node. + +* The OLSR daemon as it is today (the one which can be downloaded from www.olsr.org; we will further reference it just as "OLSR daemon") comes with a lot of extensions / changes (e.g. the default config disables most of the RFC3626 features as MPR and enables ETX, FishEye, etc) that alter the behaviour of the routing protocol completely (compared to RFC 3626). Those extensions have been developed by an active community (Thomas Lopatic & Elektra and the OLSR-NG project later on). Therefore the OLSR daemon today is not comparable with the initial OLSR protocol at all, as many things which had been specified in the RFC 3626 have been modified. Again, because of our personal experiences, those extensions help to decrease the problems inherited by the link state algorithm but are unable to solve it which led to our decision, that it might be better to try a different approach.