Re: Regrep data integrity (referential integrity)

[Len Gallagher <LGallagher@nist.gov>]

All,

Below Terry asks what "referential integrity" is.  It's a formal term from
relational database theory, but it also has a lot of informal usage.  In
relational theory, when one defines a referential integrity constraint from
table A to table B, it means that every row in table A, with non-null
foreign key values, must point to exactly one row in table B, i.e. if the
item has a parent of a given type, it has a unique one.

Informal usage in the OASIS spec could mean that every URIreference in the
DTD points to something else that exists in the DTD.  Or every
"classification" in a submission points to a "classification-scheme" that
exists in some publicly accessible repository. The spec is also using
referential integrity when it says that a "registration-status" must come
from a "registration-status-list", etc.

-- Len G.



At 06:48 PM 3/28/00 , Terry Allen wrote:
>Thanks for the quick response, Lisa.
>| The statement that Terry writes below that specifies "that a 'data
>| integrity' policy should exist and be published" is correct, I think,  for
>| this specification.  To require  more than that (i.e., to require a
>| 'defined' security policy and mechanisms (e.g., digital signatures)) is
>| beyond the scope of our specification and should be implementation-defined.
>|   The IMS meta-data specs do not address data integrity; however there may
>| be a different IMS spec that does.  
>
>Good.  Then we're set.
>
>| NEW ISSUE: On the same but actually different note,  the issue that I
>| wanted to raise in the conference call was not 'data integrity' as Terry
>| defines it below, but rather data 'preservation'. (I'll stick with using
>| 'data integrity' in the security context.)   Our specification should state
>| that a registry will return, without loss or error, the metadata that was
>| submitted/agreed to by the SO.  IMS has a few good statements in their
>| spec:  (LOM is learning object meta-data)
>
>Ah, good.  We missed that; "Persistence" is about saying how long you
>expect to stay in business.  It's perhaps implied by the statement
>about "INTEGRITY" in repreqs.htm,
>
>"The system must maintain data integrity (consistency, referential 
>integrity)."
>
>but it probably should be specified better and this whole nonnormative
>document should be scoured for what should go into the normative spec.
>And I'm not sure what referential integrity is.
**************************************************************
Len Gallagher                             LGallagher@nist.gov
NIST                                      Work: 301-975-3251
Bldg 820  Room 562                        Home: 301-424-1928
Gaithersburg, MD 20899-8970 USA           Fax: 301-948-6213
**************************************************************