Dear Christoph,

Christoph Helma wrote:
> Dear Nina, All,
>
> As a followup to our discussion before I want to specify the obstacles,
> that I have experienced for dataset creation.
>
> Lets assume I am working on a feature generation service, that creates a
> dataset as output and sends it to a datasetset service. The same problem
> arises, if I want to create a dataset in RDF "by hand" to submit it to
> the dataset service. Another use case would be the exchange of datasets
> between two dataset services (e.g. if I want to import a dataset from
> Ambit into my service).
>
> First:
> I can create only incomplete/wrong metadata (it is e.g. impossible to
> know dc:identifier in advance). Shall I send an incomplete/incorrect RDF
> (i.e. a RDF that does not pass our specifications) to the dataset
> service and rely on the dataset service to complete/update the
> information?
>   
I would say yes, in fact I don't think all dc: properties are mandatory,
at least this is not formally required.  And if you create a dataset,
you (your service) knows most about it - in the sense
which features are included, how they are generated, who generated the
dataset, etc.  One could have information about the same dataset on
different places and refer to it via owl:sameAs and owl:seeAlso.

(I know it sounds weird, but seems an accepted practice in RDF world).

> Second:
> I am a little bit concerned about conflicts between anonymous nodes. I
> see that Redland creates unique IDs for these nodes, but they might
> conflict with IDs that exist in the dataset service. Do you think, that
> could lead to problems - I have no idea how these IDs are handled during
> merging.
>   

It seems a generic issue which should already have solution.
I'll try to find out how this work with merging, but most probably this
should not be an issue ( on import one could always rename the anonymous
nodes with unique identifiers within the system, I think I have seen
this happening in Sesame as well as in Jena).

Best regards,
Nina
> Best regards,
> Christoph
> _______________________________________________
> Development mailing list
> Development at opentox.org
> http://www.opentox.org/mailman/listinfo/development
>