Yunker/Nagahashi 1/26/2004: Question on Construction of GUIDREF /Versioning

["Monica J. Martin" <Monica.Martin@Sun.COM>]

> mm1: Kenji,
> With this proposal and today's TC call and vote, can you agree on 
> final verbiage with John Yunker to ensure we capture his comment on 
> name (item 3 below):
>
>    * Leave specification as is.
>    * Do not specify what UID is.
>    * ***Clarify: Can use name and perhaps give example of a condition 
> when used.***
>
> I have also attached John's response. I think we pretty much have 
> consensus except for the versioning question, Work Item 7.

mm2: For Work Item 43, have you had any luck so we can discuss in 
today's call? In addition to the response from Dave Webber [1], I've 
also included the initial response from Reg/Rep. We should think 
carefully before embedding information as suggested. John, can we have 
your response so we can finish this (see ***clarify.....***)?

Related and for Work Item 7, in addition, the ebXML JC is considering a 
larger-scale compatibility matrix that may be affected by/affect our 
discussion on UID and/or versioning. We may also look to RIM on how it 
sees versioning as I think it will come to a common solution between all 
the ebMXL TCs at some point.

Thank you.

[1] attached

>------------------------------------------------------------------------
>
>Subject:
>RE: [ebxml-bp] Nagahashi 1/7/2004: Question on Construction of GUIDREF
>From:
>"Yunker, John" <yunker@amazon.com>
>Date:
>Mon, 19 Jan 2004 11:12:47 -0800
>To:
>"Monica J. Martin" <Monica.Martin@Sun.COM>, nagahashi@fla.fujitsu.com
>CC:
>ebxml-bp@lists.oasis-open.org
>
>BPSS components have a name attribute, and the reference elements reference both a name and an ID.  Not sure what you mean by "not used for referencing BPSS components".
>
>Since IDs change when versions change (there is no version element in naming) and are maintained by software, the name reference is the only element that allows binding over version, domain (e.g. industry group), and usecase boundaries.
>
>Monica, I'm still in the midst of 2004 planning with all day meeting, and appologize for not being on todays call.
>
>John
>
>-----Original Message-----
>From: Monica J. Martin [mailto:Monica.Martin@Sun.COM] 
>Sent: Monday, January 19, 2004 11:58 AM
>To: nagahashi@fla.fujitsu.com
>Cc: ebxml-bp@lists.oasis-open.org
>Subject: Re: [ebxml-bp] Nagahashi 1/7/2004: Question on Construction of GUIDREF
>
>
>nagahashi@fla.fujitsu.com wrote:
>
>
>  
>
>>>Nagahashi: Hi,
>>>
>>>- name attribute carries descriptive text only. It is never used for
>>>referencing BPSS components. It is not required to be unique at all, 
>>>even within a package.
>>>
>>>      
>>>
>
>mm1: Need input from John Yunker on this one.
>
>
>  
>
>>>Any thoughts?
>>> 
>>>
>>>      
>>>
>
>mm1: We can discuss in today's call to come to closure. Thank you.
>
>  
>
>
> ------------------------------------------------------------------------
>
> Subject:
> Re: [ebxml-bp] Nagahashi 1/7/2004: Question on Construction of GUIDREF
> From:
> nagahashi@fla.fujitsu.com
> Date:
> Mon, 19 Jan 2004 10:50:05 -0800
> To:
> ebxml-bp@lists.oasis-open.org
>
>
>Hi,
>
>Sorry to be late. I was reading discussions on work item 43. 
>Here's my proposal for resolution of the work item.
>
>There are two points to address:
>- double reference with name and nameID
>- uniqueness rule for nameID (or name), i.e. scope of nameID.
>
>Terms:
>- BPSS components: such as BinaryCollaboration, TransactionActivity.
>- GUID: general term for globally unique identifier. UUID is a specific 
>example of GUID. Identifiers only unique within a document or package 
>are not GUIDs.
>- UUID: 128bit globally unique identifier defined by DCE. Used in ebRR. 
>
>Proposal:
>
>- Use only nameID for referencing BPSS components.
>
>- BPSS specification doesn't enforce any specific format for nameID. 
>Author may use arbitrary text (as far as it satisfies other rules).
>
>- name attribute carries descriptive text only. It is never used for 
>referencing BPSS components. It is not required to be unique at all, 
>even within a package.
>
>- Scope of nameID is a package. So reference to BPSS components (within 
>the same document) MUST be qualified with package names. As a special 
>case, qualification is not necessary for reference to BPSS components in 
>the same package. This means nameID is not a GUID (as defined above) --- 
>so type name for nameID should not contain "GUID" to avoid confusion.
>
>- external descriptions (eg CPP/A) reference BPSS components by 
>combination of 
>  1. globally unique identifier for BPSS document (eg. URL, or UUID 
>assigned by ebRR)
>  2. fragment = <package name> '.' <nameID>
>
>  (example: http://host.domain.com/brabra/process.xml#package.PO)
>
>Any thoughts?
>
>Regards,
>Kenji Nagahashi
>  
>

Subject:
Re: [ebxml-bp] Name and GUID
From:
Farrukh Najmi <Farrukh.Najmi@Sun.COM>
Date:
Sun, 14 Dec 2003 16:13:33 -0500
To:
David RR Webber <david@drrw.info>
CC:
"Monica J. Martin" <Monica.Martin@Sun.COM>, Dale Moberg <dmoberg@cyclonecommerce.com>, Jean-Jacques Dubray <jeanjadu@Attachmate.com>, ebXML BP <ebxml-bp@lists.oasis-open.org>, Nikola Stojanovic <Nikola.Stojanovic@RosettaNet.org>

David RR Webber wrote:

> I'm VERY happy if GUID is not synonymous with UUID,
> but rather may or may not be a UUID.
>
> If we simply state that - and then people are responsible for implementing a GUID that is appropriate for their implementation environment needs - that's all my issues addressed.
>  
>
My 2c follow. Apologies in advance if I am missing some context.

The registry defines three requirements for metadata:

1. Unique Id

Id must be universally unique, need not be human friendly and used only to identify the object in a unique unambiguous manner. The Id must not be overloaded to contain any encoded information.

The registry uses a URN based upon DCE 128 bit UUID. This could be generalized to be any URN that matches one or more canonically supported URN schemes (including UUID URNs such as: "urn:uuid:d242d228-43e0-91ce-88aacbcc167c"

2. Human Friendly Name:

In the registry this can be any String and need not be unique at all.

3. External Identifier

A unique value within a unique namespace that serves as an identifier for the object.
An example is the DUNS # for a company within the DUNS namespace.

Which of the above requirements does GUID map to? Is it Unique Id? If so, I think we should not have a GUID as well as a UUID.

-- 
Regards,
Farrukh

Content Enable your enterprise with the freebXML Registry:

http://ebxmlrr.sourceforge.net/presentations/freebXMLRegistryBrochure.pdf
http://ebxmlrr.sourceforge.net
-- 


Subject:
[ebxml-bp] UID versioning note
From:
David RR Webber <david@drrw.info>
Date:
Sat, 24 Jan 2004 00:07:13 -0500
To:
BPSS ebXML <ebxml-bp@lists.oasis-open.org>

Team,
 
Here is the detail on this extracted from the BCM and CAM specs'.
 
This was an action item from last weeks call.
 
Thanks, DW.
================================================

UIDreference
	

Valid UID and optional associated registry and taxonomy that points to an entry in a Registry that provides contextual metadata content such as a [valuelist] or other information
 

Similarly the taxonomy preferred is that of the UID system, however where legacy schemes exist such as EDI element dictionary numbering, then the UIDReference can accommodate such values accordingly.    The UID values themselves are composed of an alpha prefix representing an acronym for the domain organization, followed by a simple 6-digit numeric.  Optionally a UID can also have a suffix of colon, major version, and colon, minor version, to provide version control.  When the version information is omitted then the UID reference points to the latest current information from the registry by default.