On Monday, March 12, 2012 16:23:30 Daniele Furlan wrote:
2012/3/12 Marek Lindner lindner_marek@yahoo.de:
On Monday, March 12, 2012 15:49:26 Daniele Furlan wrote:
I don't quite understand what you are proposing. Are you saying removing the averaging mechanism makes this patch obsolete and allows replacing the TTL/TQ values as the code does now (without the patch) ?
No no this patch is good, the removing of TQ average can be a following step as now average is not forwarded anymore, but only used as a comparative value between path in another point of the source code.
This follow-up step needs a volunteer. It is a bigger operation to remove the averaging and ensure everything still works. :-)
I can try to send some patch next week to start this operation.. :)
Wow - cool!
Do you mind explaining in simple words how the averaging could cause routing havoc together with the strict OGM forwarding patch ? I did not understand. Or are you in favor of removing the averaging for style / performance reasons?
Actually the average as it is calculated in the current version (without "considering" lost OGMs) is not monotonically decreasing. This cause routing loops when it is forwarded so with this patch the problem does not happen.
That is what I thought. This patch also enforces monotonicity, thereby eliminating the loop "potential" created by averaging the TQ.
Also maintaing only the latest TQ guarantee to have a more updated information about the path. Furthermore an average is already done at every link TQ calculation so it is superfluous to have another window. This reduce also memory and computational requirements.
Full agreement on this part. At the moment, I was more worried about temporary loops which seems to be addressed with this patch.
Thanks for all your feedback!
Cheers, Marek