if you ever had the feeling that rc1 is somehow unstable and many obvious routes were simply ignored, relax, you were totally right. Our optimism drove us to introduce just *minor* changes to the code just five minutes to 12 and which eventually turned out to be buggy. So happened with rc1. We are trying hard to improve.
So what has changed since then?
- Fixed relevant bugs in the debug-system and an incredible bug in the neighbor-detection algorithm causing one of two neighbors to simply ignore the other (what a mess).
- Now batman is sending originator-messages with TTL of 50 on every interface. (results in: no more dead back-routes to nodes with multiple interfaces but also larger routing-tables, more processing and traffic overhead)
- Reduced bidirectional-link-check-range to 2 and increased sequence-range to 128 (as some experiments in the berlin freifunk mesh indicated that these parameters worked best for our purpose)
- AND, in order to resolve existing bugs instead of introducing new ones, we took some time to test all changes in our local setup which includes freifunk-based WRTs, buffalos, and foneras as well as 4G-MeshCubes and i386 based machines ... hoping that you can confirm the good results
Please update your nodes to batmand-0.2.0-rc2 by visiting our new SUPA-DOWNLOAD-STORE at http://open-mesh.net/batman/downloads offering precompiled binaries and packages for quite a number of linux-based systems. Also we are willing to include other targets in the auto-compilation toolchain which currently allows compilations for openWrt ( http://wiki.openwrt.org/TableOfHardware ) and openembedded ( http://www.openembedded.org/filebrowser/org.openembedded.dev/conf/machine ) based systems. So if you find your HW supported by one of the mentioned projects (and are willing to test next autogenerated results) please give us a hint.
very best wishes from the B.A.T.M.A.N. development team