[DDI-SRG] TIC Meeting: Identifier issues

[J Gager]

Pascal -

I am answering on behalf of Achim.
 
- what is the @action attribute in identifiers?
	This has values of Add/Delete/Replace. It is used for overriding in
the grouping meachanism as documented in the orginal Add/Replace/Delete
elements of the IdentifiableIDType et al.

- In your email, you mention that Name, VersionResponsibility and
VersionRationale remain elements. I assume that VersionDate is in there as
well.
	Version date will actually be a attribute of type r:BaseDateType
(which is a union of ISO year, year month, date, etc.)

- will the ReferenceType be revised as well and become attribute base and
reflect the same approah as for identifiers? This would make it more
compact.
	ReferenceType will not become attributes due to the fact that there
is substructure in this, such as the additional referencing necessary for a
schemes and exclusion of items from them. In addition, there are case where
ReferenceType is extended to add additional content (see
pi:PhysicalDataProductReferenceType). Since this can have multiple items, it
must be an element and the idea of a reference mixing elements and
attributes is very off putting from a consistent design perspective.
However, we are going to revisit all reference and strongly type them based
on what is being referenced. For instance, referencing a variable will now
require either a URN or a combination of scheme reference(ID, agency,
version) and item reference (ID, version) - all required.

- Have we defined a formatting regular expression for version. I do not
particularly support non-numeric characters in version as they may not sort
as well as numbers.
	Bug 00046 has not been resolved yet, but I too am of the opinion
that we should limit these to numbers (despite the original proposed
solution).

J

-----Original Message-----
From: ddi-srg-bounces at icpsr.umich.edu
[mailto:ddi-srg-bounces at icpsr.umich.edu] On Behalf Of Pascal Heus
Sent: Tuesday, October 30, 2007 1:07 PM
To: Chris Nelson
Cc: 'DDI Technical Implementation Committee'; 'Chris Nelson'
Subject: Re: [DDI-SRG] TIC Meeting: Identifier issues

Chris:
Thanks for your input. Regrading your note on mandatory ID in reference of
the enclosing maintainable object (for versionable and identifiable), this
is really something we want o avoid in DDI as it will lead to hundred of
repeat of the same element. To alleviate this problem, we have introduced
the "default" scheme mechanism. For example, in the PhysicalDataProduct
RecordLayout, a defaultVariableScheme is defined and use for all the
variable reference in the layout. Otherwise, I will repeat the same
infomration for every variable which can repeat hundred or thousands of
time. This element already exists in references as the optional 'scheme'. I
agree that it can be present but should not be mandatory (unless a default
is not provided, this should be validate by the DDI Parser).

Achim:
a few more questions:
- what is the @action attribute in identifiers?
- In your email, you mention that Name, VersionResponsibility and
VersionRationale remain elements. I assume that VersionDate is in there as
well.
- will the ReferenceType be revised as well and become attribute base and
reflect the same approah as for identifiers? This would make it more
compact.
- Have we defined a formatting regular expression for version. I do not
particularly support non-numeric characters in version as they may not sort
as well as numbers.

thanks
*P

Chris Nelson wrote:
> Joachim,
> I agree with option 4. This is, essentially, the way it is done in SDMX.
> Both the Id method and the URN method should be supported (as they are 
> in the option). For referencing in SDMX we allow either or both of the 
> full identifier by attributes (agency, id, version etc.) and URN. In 
> other words, if the URN is present then there is no need for the full 
> identifier by attributes. Is this the intent here?
>
> For specification of unique Ids, where the URN is composed from a 
> pre-determined set of components that include the Id, I would prefer 
> to see the specification of the full identifier by attributes to be 
> mandatory. The URN could be optional  i.e. for specification the 
> application is not allowed to specify a URN on its own, as it is not 
> allowed to specify "any" URN - the URN must be composed to the rules 
> of the DDI URN scheme which includes the unique Id.
>
> For both referencing and specification, if both are supplied and there 
> is a mismatch between the URN and the full id by attributes, then one 
> of them should take priority.
>
> Just a note, also about the Id as your example is not clear here. For 
> a maintainable the id is the id of the maintainable. For a versionable 
> or identifiable, it must also have the id of the maintainable within 
> which it is contained e.g. when referencing a variable it is mandatory 
> to identify also the variable scheme.
>
> Regards,
> Chris
>
> -----Original Message-----
> From: ddi-srg-bounces at icpsr.umich.edu
> [mailto:ddi-srg-bounces at icpsr.umich.edu] On Behalf Of Joachim Wackerow
> Sent: 29 October 2007 11:22
> To: Pascal Heus; DDI Technical Implementation Committee; Chris Nelson
> Subject: [DDI-SRG] TIC Meeting: Identifier issues
>
> Pascal, Chris, and others,
>
>
> We started today the meeting in Dagstuhl. We discussed again the 
> identifier issue. Please see the attached file for details. Currently 
> we are all opting for version number 4 (attributes) right now. We would
like to think on that.
> Do you have any opinions on that?
>
> Additional note:
> things like Name, VersionResponsibility and VersionRationale need to 
> be elements in base abstract Identifier ComplexType.
>
> Pascal:
> We will discuss data types. Do you have any additional opinion on the 
> issue beyond the recorded issues in Mantis?
>
> Achim
>
>
>   

_______________________________________________
DDI-SRG mailing list
DDI-SRG at icpsr.umich.edu
http://www.icpsr.umich.edu/mailman/listinfo/ddi-srg