[olsr-dev] Make ETX metric hop-count insensitive

[Acinonyx]

O/H Daniel Poelzleithner έγραψε:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Acinonyx wrote:
>
>   
>> So i decided to try to experiment with olsrd in order to achieve a
>> hop-count free metric. The idea I have is to create a new metric, based
>> on link quality statistics, where the calculation of a path will produce
>> a path metric ranging from 0.0 to infinity. This I believe could be
>> implemented qutie easily on olsrd by simply subtracting 1.0 from the
>> original ETX.
>>     
>
> I already suggested this month ago after discussion on the freifunk
> mailinglists, but got no answer from any dev. One problem could be, that
> all olsr clients must be use this patch or it could cause routing loops.
> This was more or less my question :)
> But i stopped working on olsr, its a dead protocol for me.
>
> i suggested to have a per interface setting between 0.0 and 1.0 which
> will be substraced on the nodes neighbor connections. this way you can
> tune a little bit like setting 1.0 for lan connections, 0.8 for
> backbones and 0.0 for normal connections etc.
>
> kindly regards
>  daniel
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
> Comment: GnuPT 2.7.2
> Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
>
> iD8DBQFGEo5uy/mkIQp7AD0RApseAJ493QX/I9nVrxSgDvgjPCY1az5anQCePuIl
> pp/HM7ynFC5gxOjP9GhfHTA=
> =wjaZ
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> olsr-dev mailing list
> (spam-protected)
> https://www.olsr.org/mailman/listinfo/olsr-dev
>
>
>   
Allowing metric-base mixing would be dangerous. ETX is calculated on 
each node so each one has its own "metric view" of the network. Also a 
transistion from ETX to another metric on large networks is very 
difficult. But on relatively small networks, like the BGP confederations 
we have in Athens, it is feasible and worths a try.

Vassilis
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.olsr.org/pipermail/olsr-dev/attachments/20070403/10de2168/attachment.html>