- From: Myers, Jim <MYERSJ4@rpi.edu>
- Date: Sat, 3 Sep 2011 19:06:41 +0000
- To: Simon Miles <simon.miles@kcl.ac.uk>, Provenance Working Group WG <public-prov-wg@w3.org>
I'm not sure what will make sense to someone who hasn't been through the standardization grinder on this, but I think what you propose might help. I wonder if that could be morphed a bit to talk about the 'OPM'-style case first and then expand. OPM didn't have to deal with entity/thing distinctions because it only considered things that were naturally defined as invariant with repect to the processing that happened. So - entities are PIL constructs representing things in the real world that, in the simplest case, are used and generated by process executions. To address cases in which process executions merely modify something, PIL provides two mechanisms: the ability to say an entity participates in a PE - this allows one to record that a PE occurred in the entity's lifecycle but does not allow one to describe the change in the entitiy itself To describe the change itself, PIL recognizes the possibility to characterize things in the world in different ways and uses this allow provenance records to include entities that characterize things such that changes are viewed as the use of one entity (file-in-location-A) to generate another (file-in-location-B). The definition of entity thus allows one to describe what characterization is intended by identifying fixed attributes that are tied to indentity in that characterization. PIL also includes the ivpOf/complementOf relationship to assert that two entities are characterization of the same thing... Perhaps better ways to say it, but hopefully the general idea - only introducing the issues of thing/entity and ivpOf/complementOf to talk about mutability after covering the simpler case - comes through. Cheers, Jim ________________________________________ From: public-prov-wg-request@w3.org [public-prov-wg-request@w3.org] on behalf of Simon Miles [simon.miles@kcl.ac.uk] Sent: Saturday, September 03, 2011 2:19 PM To: Provenance Working Group WG Subject: Re: Is _this_ what is meant by "Entity"? Jim, OK, thanks. I think it's what I thought you meant :-). If we are to get across clearly and succinctly what you describe, wouldn't it be preferable to define entity only after process execution and participation (both in model and primer documents)? Defining participation requires saying what is participating, which cannot be entities without definitions being circular, so I would assume something like: Defn 1. A thing is anything that can be identified. There are no assertions directly about things in PIL, but instead about entities (defined below). Defn 2. A process execution is an identifiable activity, which performs a piece of work. Defn 3. Participation is the involvment of a thing in an activity. Defn 4. An entity is a view on a thing defined by those characteristics that do not vary in the process executions it is asserted to participate in. ["executions it is asserted to participate in" may not be all processes we want to define an entity relative to, but I can't see what else is concrete enough to be usable.] Defn 5. processExecution(id,rl,st,et) is an assertion that a process execution, identified by id, occurred, and followed recipe rl (optional) from start time st (optional) to end time et (optional). Defn 6. entity(id, [ attr: val, ...]) is an assertion that a thing existed and the entity defined by characteristics [ attr: val, ...], identified by id, was a view on that thing consistent with the participations asserted about it. Defn 7. hasParticipant(pe, e) is an assertion that the thing on which entity e is a view participated in process execution pe. Are those definitions adequate to capture everything yet? They may still be more complicated than is helpful in getting people to adopt the standard... Thanks, Simon
Received on Saturday, 3 September 2011 19:07:10 UTC