10 Sep
2018
10 Sep
'18
12:44 p.m.
I tried libremesh but I didn't like. It was very strange the way new
options are added. When I discussed this problem with some of the
libremesh developers they did not see complexity so it was clearly a
different point of view about the problem. I discussed with some users
and they also find this problematic. I think we suffered too much the
problems of the configuration mess with qMp. In the end, the most
valuable thing of qMp was the good integration of bmx6 with openwrt.
Yanosz's work on https://github.com/yanosz/mesh_testbed_generator/
showed what was missing to build your firmware in an easy manner.
Building firmwares and images is complicated in general, but the build
system for openwrt it's the exception: it's great. I enjoyed so much
having the opportunity to select specific version of software I want
to build with the firmware.
I was thinking that another approach should work and I found the opportunity.
Our users are not hackers, 99% buy the same device and 99% have the
same configuration pattern. The idea is that there are people
responsible of the community firmware decide the design of the network
and the supported devices. What needs to be templated, what not and
what are the parameters for user. Temba regular node just need 3
parameters: node name, wifi channel and IPv4 (because our
compatibility with classic guifi.net webpage / but IP planning is very
useful in my opinion!).
It just "works for us", and we are sharing our work to others if that
is useful. Not pretending anything.
With this "kind of hack" I feel a lot of control of what I'm doing:
clear way fof changing configuration options, of changing parts of the
firmware (its software versions), changing to new release... It is so
simple (once you get all the pieces and relations) that you don't want
to simplicity (and inherently destroy required complexity).
The project is new, young and growing. At some point perhaps I
discover that this template idea is a mess. We will see.
I also feel I am more close to openwrt community.
I'm not worried about maintaining temba. It is too simple, just a
templating thing with some things to consider.
From the temba point of view: your project (libremesh)
is still very
important because generates important configuration options we use/we
copy.
Libremesh, temba, and other community network firmware, are different
approaches to the same problem.