[Olsr-dev] MID optimization

[Hannes Gredler]

On Sun, Dec 14, 2008 at 10:25:23PM +0100, Henning Rogge wrote:
| On Sunday 14 December 2008 22:18:47 Hannes Gredler wrote:
| > markus,
| >
| > i have some concerns about your patch:
| >
| > http://gredler.at/hg/olsrd/rev/d411b618bff4
| >
| > we should not fix things on the receiver side
| > by faking up protocol state. - questions that come to mind:
| >
| > - what happens if the neighbor has different sequence spaces for
| >   MID vs. TCs (then you may prevent the MIDs getting learned)
| >
| > - what happens if the TC interval is lower than the MID interval ?
| >   (then your faked up MID will time out anyway )
| >
| > IMO the more correct thing to do is to fix the send side of things.
| > i.e. if an (non main IP address) of a node changes then immediatly
| > generate and flood a MID entry. please back out htis fix and
| > fix it on the sender side in the interface code.
| This does not solve the problem... because the MID might be lost.

well even if it's lost it does not get worse than it is today ;-).
 
| Our routing daemon should be able to set routes for this case too... okay, I'm 
| not sure if "setting a pseudo MID" is a really good thing to do it... but we 
| should not depend on MIDs if we have the necessary informations available 
| without them.

agreed - so what we could do is to insert 1hop routes as part of the link-set code.
-> whenver a link transitions to 'up' then we should make sure that the 1hop route
pointing to the far-end is there prior to installing 1+ hop routes.

/hannes