The last years have seen a significant rise of a enterprise organizations acting as LIRs (encouraged to do so by RIPE itself). These organizations' networks have different properties and attributes than most service provider networks, amongst those properties the quite limited number of peerings points and the widespread use of stateful firewalls. Unfortunately these properties lead to a desire to advertise "more specifics" without a covering aggregate which in turn is against the legitimate interests of several other RIPE members striving, amongst others, to preserve precious TCAM memory of core devices. While the RIPE-532 "Recommendations on IPv6 Route Aggregation" document correctly states that "the majority of ISPs accept prefixes up to a length of /48 within PA space", the speaker has 1st hand observed several cases where more specifics up to /34s have been rejected from the upstream providers of enterprise organizations, which might somewhat slow their enthusiasm as for deploying IPv6.
To get a better understanding of the real-world dynamics of ("strict") IPv6 prefix filtering in RIPE space we've recently started a small research project, in the course of which we've analyzed data from five Remote Route Collectors (RRCs) across Europe from 2010 to 2014. We have looked specifically at the distribution of IPv6 prefix lengths within the address space IPv6 Provider Aggregatable (PA) allocations, how this distribution changes over time, and what the reasons for the observed trends might be.
While it seems that eventually the second law of thermodynamics kicks in there's still some interesting observations as for the distribution of prefix lengths between /32 and /48 and we will present detailed results of the project during the talk.
Enno Rey @Enno_Insinuator is an old school network security guy who has been involved with IPv6 since 1999. In the last years he has contributed to many IPv6 projects in very large environments, both on a planning and on a technical implementation level.