Re: "shape" as a relationship, not a class

Our customers often want to know when a fact was added, deleted, modified, etc. and by whom. In other words, they need to track metadata at a level more granular (and more precisely defined) than a resource. To address this requirement, we use separate graphs that contain only this information. In these graphs, each triple in the 'data graph' has its own identifier (is reified) in order to track its provenance. 

Creating another ID for a document would not address such requirement. Not to mention, that a notion of a document is pretty ephemeral. 

They also have requirements for tracking provenance information for a group of triples. For example, a spreadsheet may have been imported on a certain date by a certain person. It contained various information about hundreds of resources. These triples get their own named graph (in addition to being part of other named graphs) and the provenance/versioning information is associated with the named graph. Again, it is typically kept in a graph(s) that is different from the data graphs.

Sent from my iPhone

> On Feb 21, 2015, at 9:31 PM, Karen Coyle <kcoyle@kcoyle.net> wrote:
> 
> 
> 
>> On 2/21/15 4:51 PM, Irene Polikoff wrote:
>> It is very common for database records to have fields for, let's say, person's name, birthdate, the date they received the driver license on and so on and the 'created by', 'created on', 'modified by', 'modified on' fields. The latter fields are understood as 'data management' information. It is about the data, not about Alice. This has been supported by systems forever, yet I don't believe there has been a need to create another ID to make this distinction apparent.
> 
> No, we didn't create different identifiers for those bits of information, but that doesn't mean that we shouldn't do so now. In the past, the data I worked with was record-based, and identifiers weren't heavily used. Even if we had understood the concept, we did not have a way to identify the metadata and the metadata subject differently because we hadn't considered it an issue at the time.
> 
> With linked data, it is indeed an issue. We have situations today where we cannot tell the difference between the identifier for the metadata and the identifier for the thing the metadata describes. This didn't matter when we were do very little with our metadata other than creating displays for humans, but as soon as we contemplated linking between various data stores, this came back to bite us. It's something we are definitely struggling with, actively, at the moment. In practice, the difference matters.
> 
> kc
> -- 
> Karen Coyle
> kcoyle@kcoyle.net http://kcoyle.net
> m: 1-510-435-8234
> skype: kcoylenet/+1-510-984-3600

Received on Sunday, 22 February 2015 17:00:28 UTC