On Aug 4, 2012, at 12:41 , Eugen Leitl <eugen@leitl.org> wrote:
On Sat, Aug 04, 2012 at 10:31:02AM -0700, Owen DeLong wrote:
IPv6 missed a great chance of doing away with all the central waterfall trickle-down space distribution.
There was no need to fix what wasn't broken.
Let's say I want to plunk down a zero-administration node somewhere, as an end user. The most natural approach is where addresses are derived from constraints, and address collisions are identical to physical space collisions. No two nodes can occupy the same space. By the time you're beyond these 2^24 lat/long resolution IPv6 is probably on its last legs anyway, and there's way to do renumbering with more bits, at the very least.
This ignores the many many studies of the idea of geo-based addressing which have proven its unfeasibility as well as the fact that not everyone wants their address to reflect the exact coordinates of where the box is located. Also, any such scheme would depend on defining an arbitrary "minimum" sized box and ignores the possibility of needing many addresses for the same physical box containing multiple virtual nodes. It sounds great in theory until you actually compare it to the real world. IP addresses are not physical addresses and trying to correlate them only leads to artificial limitations and other problems.
Luckily, /64 looks like large enough to bypass that by offering address space sufficiently large while co-existable with legacy addressing and routing.
Why on earth would you be messing around within /64? It should be easy enough to get a /48 (it certainly is now).
It's a lowest common denominator, at least as long the ISPs are playing by the rules. If end users conspire to use a new addressing scheme bypassing the ISP infrastructure as the crow flies, a freely modyfiable address field within your ISP-assigned address space is the best label you can hope for.
Uh, good luck with that.
I hope eventually somebody will start tinkering with mesh radios which also have GPS onboard (as most smartphones and tablets do). 24 + 24 + 16 bits are just enough to represent a decent-resolution WGS84 position fix. Plus, GPS gives you a pretty accurate clock.
That could be an interesting project. Limiting it to a /64 still doesn't make a lot of sense to me.
I'm actually glad it's a /64. MAC space is a lot more cramped, and that information doesn't travel at all far.
You misunderstand... I'm suggesting a /48 (65,536 /64s), not a /80 (48 bits to mess around with). Owen