Re: Unique handle generation

[Patrik Fältström <paf@cisco.com>]

At 12.49 -0800 01-03-13, Bill Manning wrote:
>%
>% Back to requirements...
>%
>% Are there any objections to these, with number 4 amended as suggested by
>% Paul George:
>%
>% [1] Every object MUST have an associated object identifier.
>%
>% [2] Object identifiers MUST be globally unique.
>%
>% [3] An object's identifier MUST NOT change during the lifetime of the
>% object, even if administrative control of the object changes over time.
>%
>% [4] An object identifier MUST contain information that unambiguously
>% identifies both the object and the object's administrative repository at
>% time of creation.
>%
>% [5] Object identifier format SHOULD be easily parsed and understood by
>% humans.
>%
>% <Scott/>
>
>I am concerned w/ two things:
>
>	) what constitutes "lifetime" in #3

I would define lifetime as the actual lifetime of the object. This 
doesn't mean that at very extreme situations a record can be replaced 
by some other record, and that way a new record is created.

For example, I see "Arin go away (not only taken over by someone 
else)" is an extreme situation. I see "record for a domain handled by 
the VGRS registry moving from one registrar to another" being a 
likely situation. The identifyer only have to survive the latter (or 
Arin taken over by someone else).

>	) I am opposed to the construction of #4.
>	  George's construct of an OI is as a composite, the object and
>	  the administrative repository.  The other items refer to
>	  atomic objects that have meaning independent of any associated
>	  repository.

My take is that the identifier should be created in such a way that 
given this identifier, the record should be able to be found -- 
without having to have one database with all record identifiers in 
the world in one place.

   paf