Friday, November 11, 2011

Odd IPv6 ULA use cases

I have to be honest, I am not a huge fan of the idea of IPv6 ULA (unique local addressing) at all. I have seen several use cases presented and even some knowledge based articles written saying to use it such as this one by Apple. There are ULA address prefix generators like this one at SixXS which are useful if you want to do ULA, my question is WHY?

At the core of the question is what do you gain by doing ULA that you don't get with doing Global Unicast Addressing? I would argue you get no benefit of having to global register a /48 ULA then simply applying for a /48 or larger from ARIN or one of the other regional registries that provide public IPv6 address space with the exception of price (which could be a big deal for some small businesses but just get your IPv6 address space from your provider for free then.) Furthermore, ULA by definition in rfc4193 cannot be routed globally and must be filtered at the edge which very much limits your IPv6 deployment and ensures you have to either deploy Global Unicast Addressing at a later date or do prefix translation as described in rfc6296 which is a viable solution but seems to introduce yet another network translation component on the network when one is not needed if you simply used Global Unicast Addressing the first time around.

The other concern I have is some OS platforms not behaving as expected when getting ULA addresses. Ideally all OS behavior with ULA would know that you don't have global IPv6 access with a ULA at all but if you are using prefix translation is that still true? Also, since IPv6 is preferred do we run into a case where the network team is putting ULA in play and breaking some of the default OS behavior that is desired for transitioning to IPv6?

Given the fact that the effort is almost identical for deploying ULA and it is Global Unicast I am not convinced that ULA is something that is needed or should be recommended. I would love to hear feedback on this one. The few corner use cases I have heard still do not seem to overcome the argument of just using Global Unicast.
- Ed

No comments: