The hash-based OEMID is removed because I think many may get it wrong and it is hard to explain how to get it right.
The input to the hash needs to identify the OEM globally. The text about 96 bits of entropy is confusing. It's not 96 bits as in the output of an RNG. It's 96 bits in some space of global OEM identifiers.
The one use I can think of for it is when the the device has a domain name of the OEM handy. They can run that through the hash and get a good globally unique OEM ID. The uniqueness is coming from the fact that domain names are managed to be globally unique by a central authority.
But maybe we should just have an OEM ID type that is a DNS name or the hash of a DNS name?
The hash-based OEMID is removed because I think many may get it wrong and it is hard to explain how to get it right.
The input to the hash needs to identify the OEM globally. The text about 96 bits of entropy is confusing. It's not 96 bits as in the output of an RNG. It's 96 bits in some space of global OEM identifiers.
The one use I can think of for it is when the the device has a domain name of the OEM handy. They can run that through the hash and get a good globally unique OEM ID. The uniqueness is coming from the fact that domain names are managed to be globally unique by a central authority.
But maybe we should just have an OEM ID type that is a DNS name or the hash of a DNS name?