> there should be no problem, default gateway is always the next hop, so<br>> if there are more default gateways the olsrd should take the one with<br>> the lowest metric.<br><br>does that mean to take all subnet routes of the kernel routing table into acount<br>
or should it just compare it with the default route olsr generates,.. (bad idea)<br><br>on a device running olsr you can manually configure a metric 0 default route<br> olsr will generate its own default route with metric >=1<br>
<br>or even worse: you can have some /1 or /2 or ... subnet routes or so,..<br>i'm sure olsr ignores them completely (while it deletes manually configured default routes)<br><br>in this scenario it makes a huge difference whether all hostroutes get exported from olsr or not,..<br>
<br>but on most routers this scenario will never occure, so a plugin that reduces the route count olsr exports makes still sense, but it should be a plugin or configureable option,..<br><br>and in combination with an "slaveroute-daemon" on a neighbouring "virtual lan-wlan bridge" * it makes even more sense,.. <br>
<br>Markus<br><br>* i refer to an wireless - lan bridge i currently develop <br> in fact its just an arp proxy an olsr repeater,..<br>and at the moment an full-olsrd to get quite useful routes<br>getting the routes from an master olsrd running on the central device of a node, could create a bit better and consistent routes.<br>
<br> this affort aims to have a net of multi-device-nodes which act (and look) like single device nodes with many wireless interfaces,.. removing the hated routing costs of 2 meters of ethernet cable,.. <br><br>why doing this, instead of investing in olsrd support of better metrics ??<br>
<br>because i think that even an ETT or any multiplicated metrik cannot reduce this burden very much,.. (because multiplicating link costs of 1 is jsut the same bad idea as an adding an ETX of 0, (it will generate temporary loops)) <br>
<br>ETT its just very useful to stop olsrd to route over slow links <br>furthermore in my opinion ETT costs should not get multiplicated, .. (multiplicated transmission times = ???)<br><br>but i`m getting off topic,..<br>
<br>markus<br>