Hi Stephane,
On Tue, 2006-11-28 at 11:41 +0100, Stephane Bortzmeyer wrote:
let's assume two registries R1 and R2
manage the namespace ".example". A customer C1 wants to create
foobar.example and asks to registry R1. A customer C2 wants to create
foobar.example and asks to registry R2. There is a conflict.
You can solve the problems in various ways (see Emin Gun Sirer's
message) but most of them create a "super-registry" on the top of R1
and R2 and you are back to the unique registry model.
This is a false statement. A basic course on distributed systems will
cover lots of design alternatives where R1 and R2 are symmetric,
mutually distrusting and there exists no super-registry, yet there is a
way to establish whether R1 or R2 acquired the name first.
Other ways to solve the problem, without a super-registry, have other
issues:
* You can share the namespace, for instance, R1 has the right to names
whose length is > N characters (this was proposed in the WSIS: ITU
could manage names <= 2 and ICANN the rest). In essence, it makes
several namespaces in ".example" so, again, you're back to the rule
"One namespace, one registry".
This is a non-solution. The namespace is not shared.
* You can give in the "non conflicts" rule. This is what happens to
most "alternative inclusive roots". Ask your users how they feel about
it.
This is underspecified, but because words like "secure timestamp" or
"blinded signature" or "commitment" or "quorum" or "consensus" or
"agreement" do not appear in it, it is likely a non-solution. Btw, not
_all_ of those keywords need to appear in a good solution, but probably
at least one should appear somewhere.
If anyone would like some design alternatives for a way to establish
whether R1 or R2 acquired the name first, without a super-registry, I'll
be happy to outline them. There are lots of possibilities, depending on
what one is willing to assume of the environment. But I'll do this only
if asked. Judging by the tone of the msgs, the folk on this list believe
themselves to be an expert at all things distributed and resembling a
name-service, so no point covering elementary topics.
Gun.
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/ietf