Re: Regrep data integrity (referential integrity)

[Len Gallagher <LGallagher@nist.gov>]

Terry and others,

Re: referential integrity

I agree too -- there's certainly no reason to reference some other
standard's formal definition for the term "referential integrity".  Regrep
can define the term for Regrep purposes if it wants to, and use it, or not.
I agree with Terry and Robin that it doesn't seem profitable for Regrep to
try to define or use the term at this time. 

Instead -- what is needed, I think, are some semantic rules to indicate
what is being referenced in certain situations, and what to do if the
referenced item is missing.  For example, it doesn't do any good to record
in the registry a reference to a <classification-scheme> if that
classification scheme can't be found in some publicly accessible place.
Likewise with referencing the URN of a <data-element> to say that it's been
"superseded" if it doesn't first exist in some repository.  I think the
Registration Authority for a given registry/repository should be free to
ignore certrain parts of a <data-elemnt> submitted for registration if they
don't make any sense.  We need semantic rules to know when things make
sense and when they don't.

NOTE: I am NOT suggesting that a Registration Authority can delete portions
of a <data-element> submitted for registration.  If someone asks for what
was submitted they should receive the entire item submitted.  But the RA
should be free to record in their implementation of the repository only
data that makes sense.

-- Len


At 11:58 AM 4/2/00 -0700, you wrote:
>Thanks for the help on "referential integrity".  I think I agree with Robin
>that we can get the same effect by careful wording and DTD construction,
>without specific reference to a relational database model.  
>
>regards, Terry
>
>