dret
association of registry entries with protocol or organization?
comment from john curran (comment 2/2, other one is #27):
- You don’t highlight the need to be exceptionally clear about what is being associated with the registry values… for example, a specific UDP port may be reserved by a company for protocol “foo”, but it is the protocol that gets the reservation, not the company. There are registries (e.g. the general portions of the AS and IP address spaces, private enterprise number, etc.) where the value has no protocol meaning aside from uniqueness, and is truly associated with an organization. Protocol designers who specify such new general purpose registries should do so with significant care and clarity, as the administration of these registries is often more complex that typical.
@jcurranz, i think i understand your concern. but this really is about the semantics of registry contents, right? in this case it's a special kind of semantics because it is about the fact whether the value is indeed self-describing (i.e., when i see it i know what it means), or just reserved (when i see it, i have to follow some process to find out what it means). is that the distinction you are referring to?
No, I'm thinking of something much more basic... can someone reserve an entry in IANA registry and sell it to another party? If the assignment of an entry is to the specification of that particular option value, then the assignment is to the purpose. If the assignment of an entry is to the requesting party, then the answer is far less clear.
The IETF has had occasions where other parties (e.g. another standard body) just said "just give us a entry in that registry, and we'll define what it means at some future time when we finish the standard"
The IETF has had occasions where other parties (e.g. another standard body) just said "just give us a entry in that registry, and we'll define what it means at some future time when we finish the standard"
now that you're mentioning it, https://twitter.com/dret/status/699335652572295173 was something i thought about just today: isn't DNS a registry of some kind? and if that's the case, then what you're saying here makes a lot of sense in terms of people buying domain names, and then selling them at a profit.
Right... The point is that some registries are simply uniqueness values, which implies that the registrant may potentially transfer their assignment, whereas others are assigned to particular code points as represented by extensions to the standard.
On 2016-02-15 23:31 , John Curran wrote:
No, I'm thinking of something much more basic... can someone reserve an entry in IANA registry and sell it to another party? If the assignment of an entry is to the specification of that particular option value, then the assignment is to the purpose. If the assignment of an entry is to the requesting party, then the answer is far less clear.
i completely agree. but in the end, that's a question of the entry semantics, right?
- if the registry value should identify "protocol meaning", then the entry itself must be associated with an entry establishing that meaning.
- if the registry value only is a unique value, and the protocol meaning has to be discovered independently of the registry, then all the registry can do is point to the owner of the entry, and discovering the value's meaning is out-of-band from the registry point of view.
The IETF has had occasions where other parties (e.g. another standard body) just said "just give us a entry in that registry, and we'll define what it means at some future time when we finish the standard"
i'd say is up for the registry to decide whether that's acceptable or not. but if that's allowed then it certainly violates some principles of RFC 7500, because then there's no stability in the value's meaning, which probably makes such a registry not quite as useful as it could be.
On 2016-02-15 23:45 , John Curran wrote:
Right... The point is that some registries are simply uniqueness values, which implies that the registrant may potentially transfer their assignment, whereas others are assigned to particular code points as represented by extensions to the standard.
i think we're in agreement here. but if a registry only guarantees uniqueness, then that's a very weak registry, because some of the principles of RFC 7500 are not adhered to.
but in some cases that may be enough. if you look at DNS domain names, then maybe it's good enough that all the registry (the DNS) guarantees is that a value can be resolved. that's all that matters for this particular registry, that i can ask at some point in time which host address is associated with that entry value. this may change over time, but in this case that's actually the goal, and DNS makes sure that i have an infrastructure that can handle this scenario well. but in this scenario it's a feature than assignments can change over time.
My principal concern is that there is clarity in any draft that proposes a registry regarding the entry semantics.