Re: Blog Post: 5 Simple Provenance Statements

On 25/10/2011 16:19, Simon Miles wrote:
> (2) Saying less in the specifications, introducing fewer concepts
> and/or with less complex language.
[...]
> (2) may conflict with the coverage of the specification's concepts,
> which we might argue is needed to allow interoperability (e.g. we
> 'need' entity attributes, we 'need' accounts, we 'need' agents, etc.)
> But our specifications can't provide complete interoperability by
> themselves, however expressive they are, as there will always be
> domain-specific information expressed by one system/person that
> another needs to interpret for the provenance to make sense.

+1

> ... So, being
> expressive enough to allow interoperability is a matter of degree, and
> if the specifications (DM and OM) are currently too complex, we may be
> trying to be too expressive. In some cases, we might reduce what we
> say by 'allowing' for the expressivity required for interoperability,
> rather than saying how it would be expressed ourselves, e.g. we don't
> say how you must assert that something is an agent, but allow for that
> to be standardised as a supplement to Prov-DM.

FWIW, in the case of "entity attributes", the point that Jim made that I found 
compelling was his report that in the provenance challenge, use of entity 
attributes was a key feature that allowed for mapping existing and proposed 
provenance usage.  It may be more expressiveness than is strictly needed, but 
having a graceful migration from or interopability with current practice is a 
key element for achieving success in standards deployment.  (That, and avoiding 
unnecessary complexity :) )

(I do very much support the general thrust of what you're saying here.)

#g
--

Received on Wednesday, 26 October 2011 11:08:58 UTC