W3C home > Mailing lists > Public > www-rdf-interest@w3.org > July 2005

RE: [VM] Configuration management for RDFS/OWL ontologies

From: Miles, AJ \(Alistair\) <A.J.Miles@rl.ac.uk>
Date: Thu, 7 Jul 2005 13:36:13 +0100
Message-ID: <F5839D944C66C049BDB45F4C1E3DF89DEE9E4B@exchange31.fed.cclrc.ac.uk>
To: "Thomas Baker" <tbaker@tbaker.de>
Cc: <public-swbp-wg@w3.org>, <www-rdf-interest@w3.org>

Perhaps a slightly confusing name, but in a project management (e.g. prince2) context 'configuration management' means a system for controlling change to ensure quality, and that's what we need for RDF vocabs.  I.e. we need to know how to support *commercial strength* RDF vocab and ontology development.

If you've got 'The Little Prince2' look at section 6.1.1 'Planning Quality' ... very useful, tho I better not reproduce it here for fear of copyright infringement.  It highlights 5 processes:  

Planning: this is what we did when we discussed the policy statements section of the SKOS Core spec - we decided what level of configuration management is required, and we wrote a process for achieving it.

Identification: this means identifying all the components of a product.  In the case of SKOS Core this is all the properties and classes, in the case of a generic RDF vocab it could be modules as well.

Control: this means 'freezing' products and making changes only within a formal (or at least clearly defined) procedure, involving e.g. access rights, version tracking.  For SKOS Core this is editorial responsibility, historical snapshots, and the review process.

Status accounting: this means keeping a record of current and historical data for a product, especially relating to the status of the product.  For SKOS Core this is per-term stability levels. 

Verification: verifying that actual status matches recorded/authorised status.  We could do that e.g. by checking if changes have occurred to a class or prop between versions that are not allowed by the term's stability level.

An analogy is e.g. car or aerospace engineering.  With good configuration management you can track a problem back to the specific batch of faulty nuts or bolts.  With poor configuration management you have no idea what went wrong or how to fix it. 

Cheers,

Al.

[1] http://en.wikipedia.org/wiki/Configuration_management

> -----Original Message-----
> From: Thomas Baker [mailto:tbaker@tbaker.de]
> Sent: 07 July 2005 13:03
> To: Miles, AJ (Alistair)
> Cc: public-swbp-wg@w3.org; www-rdf-interest@w3.org
> Subject: Re: [VM] Configuration management for RDFS/OWL ontologies
> 
> 
> On Wed, Jul 06, 2005 at 02:51:12PM +0100, Alistair Miles wrote:
> > After the discussion on the SWBP-WG VM telecon yesterday I put 
> > down some thoughts on how configuration management for RDFS/OWL 
> > ontologies ought to be done, see:
> > 
> > http://esw.w3.org/topic/ConfigurationManagement
> > 
> > Has anyone written anything like this down already?
> 
> DCMI practice for "versioning" terms is described
> -- unofficially, from a DCMI perspective -- in
> ftp://ftp.cenorm.be/public/ws-mmi-dc/mmidc148.pdf.
> 
> I'm curious about the choice of words "configuration
> management".  A Google search on '"configuration management" 
> site:w3.org'
> does not show any obvious sources for a definition in the W3C context.
> 
> Tom
> 
> -- 
> Dr. Thomas Baker                      baker@sub.uni-goettingen.de
> SUB - Goettingen State                            +49-551-39-3883
> and University Library                           +49-30-8109-9027
> Papendiek 14, 37073 Göttingen
> 
> 
Received on Thursday, 7 July 2005 12:36:19 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:14 GMT