1

I've been using IPv6 for many years now, but it wasn't until I switched ISP recently (and with that, from 6to4 to 6rd) that I found out that stateless autoconfiguration only works on /64 subnets. I had always assumed that it worked on subnets of arbitrary prefix length.

This took me quite a bit by surprise, and I must admit I don't quite understand the further implications of it. In particular, the following questions arise for me:

  • Is stateless autoconfiguration supposed to be the default method for interface initialization, or is DHCPv6 supposed to be used in most places?
  • If the latter, then what is stateless autoconfiguration supposed to be used for?
  • If the former, does that mean that "all" subnets are supposed to be /64? Why would one want to waste all that address space on "terminal" subnets? Is a network administrator not supposed to be able to subdivide his networks further?

In my particular case, I have a few different physical subnets beyond my router, so when I had my /48 6to4 prefix, I divided it in a couple of /64 subnets, which is why autoconfiguration always worked for me. Now, with 6rd, I just get a /64 prefix instead, which I then have to divide into something small than /64, clearly (I use /48 in practice). Am I not "supposed" to be able to further subdivide a 6rd prefix, or what is the matter here?

If this question looks confrontational, rest assured that is not the intention. I'm just a bit confused and I'd like to understand the intentions.

2
  • It's not a "waste", and you aren't supposed to have only a single /64. Contact your ISP. May 7, 2014 at 2:16
  • @MichaelHampton: RFC 5569, which details 6rd, specifically allows for 64-bit prefixes, so the ISP doesn't seem to be in error for using it. It would be hard to argue with them that they are in the wrong without some specification to cite. Is there any established praxis or specification that discourages or disallows smaller subnets than 64 bits? Is there any rationale somewhere that I can read for why having each physical link being a /64 subnet "isn't a waste"?
    – Dolda2000
    May 7, 2014 at 3:30

1 Answer 1

1

I encountered this too, and added a 6in4 tunnel to get around my ISP "only" offering a /64 via 6rd for my home networking experiments.

This is simply a very strong convention that the /64 is the smallest unit for automatically configuring subnets. It is easy to remember as half of the address. Some protocols and implementations have this assumption built in, and it would be silly to break that.

A /64 allows you to do neat things like fit an entire IPv4 or MAC address space in any arbitrary subnet. If embedding a MAC address in an IP seems like a bad idea, as an alternative generated addresses are unlikely to conflict.

RFC 6177 backs away from pushing a generous /48 to all sites and acknowledges other allocation schemes. As it explicitly highlights the use cases for multiple subnets, compromises like /56 are mentioned.

Even then, there is no shortage. Providers can offer /48 to most or all users without running out.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .