[Olsr-users] Survey about fib_metric usage
Mon Sep 7 13:55:23 CEST 2009
Does anyone really needs fib_metric, i mean requires another value than
if kernel routing metrics are applied (e.g. fib_metric="approx" -> kernel
routing metric = hopcount) to olsr routes,
this makes managing kernel routes more complex,
and bloats the code of olsrd.
Furthermore many of my improvements to the routing code (to ensure that all
olsr routes are correctly applied to the kernel (and stay there)), only work
partly, or not effcient with fib_metrics != "flat",..
for example if olsr crashes and starts again, with fib_metric flat, the old
routes will not be able to cause problems, they will get overwritten as soon
as olsr has new routes again.
if not they may stay and may even interfere badly with routing if their
metric (hopcount) is lower, than current best routes hopcount,..
also routes with an "flat" metric can get replaced directly in kernel, if
they get different nexthop, and olsr does not need to make an delete && add
like now in this case
So if u need this feature, please let me now, including for what u use it,
beause there may even better/less_problmatic ways
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Olsr-users