The Unique Identifier rathole - was: Re: [ubl] Missing Schema Annotationsand other schema matters

MHonArc v2.5.0b2 -->

ubl message

[Date Prev]
 | [Thread Prev]
 | [Thread Next]
 | [Date Next]

--

[Date Index]
 | [Thread Index]
 | [List Home]

Subject: The Unique Identifier rathole - was: Re: [ubl] Missing Schema Annotationsand other schema matters

From: Tim McGrath <>

To: 

Date: Fri, 19 May 2006 14:46:36 +0800

i dont want to nit pick here but the CCTS only mandates a Unique 
Identifier for either BIEs or CCs when it is to be stored in a Registry.

We have been discussing this ad nauseam since 2002.  i dont want to go 
over it again (but i will ;-) ).  in terms of outcomes i agree with what 
Syvlia says.  we will generate some numbers from EDIFIX and use these in 
our schemas but I dont think we are kidding anyone that they will 
actually be useful or maintainable.  They are to stop us having this 
debate every time we do a review.
 
CCTS (2.01) defines a Unique Identifier as "The identifier that 
references a Registry Class instance in a universally unique and 
unambiguous way."  That is, the storage/registry identifier Stephen 
noted earlier.  Nothing to do with the defintion of a BIE or CC.  So it 
probably best not to call anything a Unique Identifier unless it is the 
storage key for a Registry Class.

As far as identifying definitions of BIE and CCs, the CCTS says when 
Applying the Naming Convention (Figure 5-4 ) - "Step 6 Assign Unique 
Identifier to the New Item" - this suggests its storage into a registry. 
However, in section 5.3.1 CCTS goes on to say  "Step 6. Assign a 
Temporary Identifier to the new item in the form of a 6 digit 
alphanumeric string, chosen at the discretion of the user."   These 
temporary identifiers are also used in the examples of core component 
catalogues.  But this identifier does not form part of the Core 
Components and Data Types Metamodel (Figure 6-1.) of CCTS.  So I am not 
sure when we started thinking of  them as being mandatory.

In fact i am not aware anyone (including the TBG groups) who actually 
uses these temporary 6 character strings.  I suspect the motivation for 
this was  to make harmonization easier. So it is worth noting that the 
new TBG17 format now requires a UUID or GUID type identifier (eg. 
cdf11383-243a-4b7e-a573-f73c8415743b) of any BIE or CC from its 
submitters.  Maybe this will make it into the next CCTS but it isn't in 
2.01.

I suspect the reason this is such a grey area is because we dont 
actually need another identifier for BIE or CCs.  That is what the 
Dictionary Entry Name does.  Commonsense tells us that having 2 "unique" 
identifiers is unnecessary and potentially confusing when we try and 
maintain these things.  And it doesn't matter if this identifier is 
numeric, english terms or hebrew - we only need one of them. And as we 
must have a Dictionary Entry Name, lets make it that one.

Sylvia Webb wrote:

>Stephen,
>
>Unique Identifiers are mandatory per the NDR. See the DOC "X" rules. They
>are also mandatory in CCTS. 
>
>They were not included in the first public review schema because we had
>decided not to include them in UBL 1.0, and, GEFEG was not given enough time
>to include them after we decided to do so for UBL 2.0.
>
>Unique Identifiers will be in the data models and auto generated in the
>schema for the next public review. 
>
>With respect to the extension point, I would suggest that you read the CTD
>and GXS rules in the last NDR draft to determine if your questions are
>answered.
>
>Regards,
>Sylvia
>
>