A couple of days visit to Quintana, that ended up being two weeks: Gui joined Jesi and Nico on a “micro-sprint”, to help expand the local network, and at the same time do some “pair development” on Libre-Mesh.org project, to solve several longstanding firmware tasks and work on new ideas, intending to increase the momentum and get the ball rolling.
Nico and Jesi drove to Buenos Aires to do some collective shopping: local funds collects were done in QuintanaLibre, AnisacateLibre, (yet-to-be) SerranitaLibre and DeltaLibre in order to (wholesale) buy routers, antennas and everything needed for putting up 40 nodes in total, distributed along the different networks. The “node kit” is based on TP-Link TL-WDR3500, weatherproofed in a plastic box (we just found these StationBox, which are perfect for the job, yet cheap!), with the factory omni antennas put up on the focus of a recycled parabolic grid, connected back to the router with male-to-female RP-SMA pigtails.
The parabolic grids come from a local dealer, who had a huge stock of (otherwise unsellable) grids with no feed, which we bargained for ~8 USD each. Then, hacked a plastic pipe to hold the omni antenna, and ended up with an incredibly cheap dual-band parabolic dish with an estimated gain of 18dbi / 25dbi (on 2ghz / 5ghz respectively). The first test was back in January, and given the impressive results, we went ahead and bought a pack of 60 grids.
A community workshop was held in the town in order to build the nodes, and Jesi spent the following days going to participants’ houses for actually mounting things up, getting a total of 6 new (up-and-running) nodes in less than a week.
At the same time, we sat down with Nico for a few long days, and coded some steps forward of firmware features:
- made and packaged a simple webinterface to visualize an ath9k spectral scan in realtime, currently running independent of luci framework, but looking forward to integrate it.
- made and packaged a first version of a web antenna alignment tool, that plots in realtime the received signal of wifi neighbours (associated stations in adhoc, for example). The first real-world test was that very afternoon, during one of Jesi installs, where the signal was being blocked by a tree, which was then cut. We jumped in excitement as we saw a clear step-up in the graph as the tree fell down!
- resuscitated, polished and packaged a luci module for batman-adv originally made by Jow for a presentation.
- restructured AlterMap database and code, according to what Nico and Andre discussed back in May, looking forward to merge the mapping projects.
- rewrote altermap-agent in Lua, dramatically decreasing the install size since it doesn’t depend on curl anymore. The new agent is easier to maintain and extend, allowing more information to be reported.
- we migrated several packages from AlterMesh repo to Libre-Mesh, the metaproject that aggregates development efforts with qMp.cat and eigenNet, and that will serve as a base for all our firmwares in the near future.
- we tested the latest compat-wireless version, and verified that it’s still broken (wifi behaviour is erratic at best) so we’ll still keep using the older version.
- we also stress-tested current batman-adv_2013.3.0, to see if some outstanding bugs had been addressed or not. Again, the result was not satisfactory, so we kept Chef shipping the older, more stable batman-adv_2012.4.0
- made some experiments with alfred, a recently released user-space helper for batman-adv networks. We plan on using it to auto-discover / generate bat-hosts file and other similar tasks.
- completed netifd integration of batman-adv configuration scheme in OpenWrt, and sent the pull request to upstream developers for consideration.