[Olsr-dev] LinkQualityMult problems in 0.5.6-r5

[Mitar]

Hi!

On Sat, Sep 5, 2009 at 2:15 PM, Henning Rogge<(spam-protected)> wrote:
> Multiple interfaces with the same IP is a problem. Not only for lqmult but
> sometimes even for the ETX calculation. That's because OLSR cannot tell the
> different LQ/NLQ pairs appart if you have a lot of interfaces with the same
> IP.

What kind of problems are there? Because we are using it quite
successfully and have not seen any problems with it. I think this
should be supported. To have multiple interfaces with same IP address.
It is like having one interface and seeing multiple nodes - for each
link you have LQ. So why not have another link which just happens to
go over some other medium/interface, like Internet (VPN) or another
wifi card.

And yes. We would not need to set LinkQualityMult on VPN interface as
we would just mark VPN node as (slightly) "bad" and nodes would prefer
other links. But currently we have to set LinkQualityMult both on
nodes and on a VPN server to have equally higher LQ and thus ETX. (And
changing LQ calculation algorithm with OLSR versions does really not
help getting all those LinkQualityMult values set properly network
wide.)

Still, there should be also an option on nodes to ignore "badness" of
another node. For example in our case we would set "badness" on a VPN
node but for some (accelerator) nodes (with high speed optical
uplinks) we would still prefer to have VPN connection be the same as
wifi links. So it would be necessary to be able to override "badness".

(Currently we disable LinkQualityMult both on "accelerator" node and
on a VPN server for that specific node IP.)

> I hope we can make everyone stopping this when Markus "source-ip" Patch will
> be introduced with the next version of OLSR.

Can you explain or point me to more information about this?


Mitar