[Rift] Negative disagg question

["Pascal Thubert (pthubert)" <[e-mail]>]

Dear all:

Yesterday at RIFT there were 2 very good questions asked upon the great presentation by Bruno:

1) difference before withdrawing prefix P and advertising it in a negative way

Withdrawing P removes it from RIB on the receiving router R (if it's there, else nothing to do) and then forget.

Advertising P as negative installs a negative route in the RIB in R. This negative route carves a logical hole in the aggregation above. 
If there is no aggregation above in R, R needs to take no action apart from installing the negative route in RIB. 
If the aggregation comes later, the FIB will be installed with the hole carved in it at that time. 

2) if the parent aggregation goes away, is the hole carved of the upper level of aggregation?

Short: Yes. 
Long: the carving is done by adding the route to P in FIB via the other nodes that advertise the encompassing aggregation (competing for the same longest match). 
So if the longest aggregation changes to a shorter one, the set of successors has to be recalculated. 
This is an unusual situation in RIFT if all leaves are similar since in that case we do not find nested aggregations, but say you do.
You could find the sitiuation e.g., if the parent aggregation is the positive disaggregation of a larger prefix. When the problem that causes the positive disaggregation goes away, meaning that a parent could not reach that aggregation and now it can, the result is that this parent is added to the FIB for the negative prefix.

I hope this helps

Pascal