Re: Closure, registries and ADMS

Dave,

Just a quick response to this in public.

It's a *very* interesting set of points you raise here from all sorts of 
angles. I'm pushing colleagues on the team and elsewhere to look at 
this. Hmmm... checks process for workshops...

Phil.

On 19/10/2012 12:35, Dave Reynolds wrote:
> This is to explain and expand on my question on "closure" in ADMS.
>
> First some context ...
>
> I've been working [1] with a few groups [2] who feel the need for some
> (preferably standardized) notion of Linked Data Registry as a way to
> improve governance and interoperability in their use of Linked Data.
>
> Such registries would need to record things like ontologies, code lists,
> reference URI sets and datasets.
>
> There are a lot of important differences between a *registry* and a
> *repository* including:
>
> (a) A registry records that some asset exists but the asset itself may
> be served from elsewhere.
>
> (b) A registry can state the disposition of the register owner to that
> asset. So things like "status" may differ for the same asset in
> different registers. One register owner might think an asset is ready
> for widespread use, another might not.
>
> (c) A registry can definitively enumerate the set of things in a
> particular collection (which we'll call a "register"). It can in effect
> state "these and only these ontologies are approved for use in this
> context" or "these and only these codes are in this code list, using
> anything else would violate the schema".
>
> Operationally a registry has some controlled approval process to make
> (b) and (c) work. So while a "register" is, at some level, just a list
> it is a list that makes promises about the status of its content. That
> as much about governance and promise as it is about technical notions of
> closure.
>
> Most of the people I work with in this area have a geospatial
> background, and thus are familiar with OGC and ISO specs of which there
> are a lot that are relevant here.  In particular, a default expectation
> would be a linked data vocabulary for registries which broadly
> corresponds with  ISO 19135 (Geographic information -- Procedures for
> item registration).
>
> BTW This is also the context behind my questions on "Representation
> Technique" since all of these groups want to "represent" a single
> concept via a range of representation languages including non-RDF ones
> like UML and XSD.
>
> ... end context.
>
> So a key question for me is whether ADMS, which doesn't claim to support
> semantic asset registries only semantic asset *repositories* is a
> possible match to this need and something to build on. Or not.
>
> If the requirements I'm seeing from these groups are representative of
> other public sector organizations then this will be a question for GLD
> in how it decides to proceed with ADMS.
>
> Looking at my list above then (a) seems fine, (b) may just require some
> improved description of adms:status (or more likely an extension), (c)
> is currently not supported at all.
>
> So with this context my question translates to:
>
> """
> Do we see Semantic Registries as a use case for ADMS and so wish to
> extend/modify it to fit (possibly based on lessons from the existing
> standards) or is that out of scope?
> """
>
> Dave
>
> [1] Full disclosure some of this work is being commercially funded.
>
> [2] In the interests of transparency these include the UK Location
> Programme (who lead UK's response to the EU INSPIRE directive), UKGovLD
> (who have a coordination role in UK government use of linked data) and
> some groups in CSIRO (who in turn are working with UN agencies). I do
> not speak for any of these groups, this discussion is simply informed by
> my interactions with them and any errors here are my own.
>
>

-- 


Phil Archer
W3C eGovernment
http://www.w3.org/egov/

http://philarcher.org
+44 (0)7887 767755
@philarcher1

Received on Friday, 19 October 2012 14:55:45 UTC