Pekka Nikander wrote:
I think we know, at least for some value of "know". We can use DHTs.
What we don't apparently know are how to secure them against
freeloading in the case of multi-operator DHTs and how fast/slow they
will be in widespread use. But there are experiments and lots of
research going on in those areas.
Yes, but IMHO the hardest problem is scale.
If we want the architecture to move towards a flat identifier space
(whether fixed length or crammed into 128 bits) we need a way to do
lookups that scale to when every host on the planet has such an
identifier. RFC 1726 talks about at least 10^12 hosts.
I don't know of any DHT research that is pushing towards such scale, but
I'd would be interested in finding out if there is.
More seriously, I would rather see the Internet to develop towards
fewer contention points (like the current DNS and IP address
assignments) rather than towards more of them.
But if we don't have any indication that we can scale flat spaces, then
we need hierarchy. Hierarchy doesn't necessarily imply administratively
controlled root of the name/number space; one can in theory construct a
hierarchy based on multiple hashes that give statistical uniqueness at
the top level.
I think it would be wise to expand the KHI draft to allow for
hierarchical identifiers, so that we have flexibility as we learn more
about a 128-bit identifier space.
For instance, allocating binary 0001/4 for IP identifiers, and then
000100/6 for flat Keyed Hash Identifiers might make sense, leaving room
for other work in 0001/4.
Erik
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------