Fwd: Re: Naive question on DCAT versioning

Dear all,

I'm forwarding a question raised by Antoine Isaac which I feel deserves 
attention from the WG if possible as he's highlighting a possible 
inconsistency that might come back to bite us.

Edited e-mail exchange below...

[..]

PA:
>
> In my mind the main difference between DCAT and ADMS is that the latter does indeed include versions. It has a notion of a series of versions (we use sub properties of xhv:next/prev/last).

IA:
Yes, that's what I saw.


>
> Both ADMS and DCAT allow for multiple distributions of one dataset (Asset), and the date of creation/modification of those distributions is independent of the dataset. I might take a CSV from 15 years ago and create a new distribution of it in RDF for example. But it's still a distribution of the 15 year old data.


I have absolutely no problem with this. The patterns seem easy to get, here.


>
> At dataset level, I can imagine that a dataset might be modified without it being a new dataset in at least some people's minds. If I correct an error in a dataset is that a new version or not? I guess that depends on a point of view. ADMS also has the modified property on a dataset so it allows modification without it being a new version.


Ok, this one thing I had missed. From the ADMS document I had understood 
that new versions, even minor ones, would have to be published as new 
versions. I had not seen the dcterms:modified. That being said, I've 
search for "modif" in the ADMS document, and just found two occurrences, 
including the name of the property. I'd say that it is quite light in 
support to the "modifications not leading to new versions" compared to 
sentences like:
[
Assets can be versioned. Every time the intellectual content of an asset 
changes, the result is considered to be a new asset that can be linked 
to previous and next versions of the Asset.
]
But that's rather an editorial issue. I let you decide whether this 
can/should be handled, or not.

The issue I still have is how to handle ADMS versions in DCAT.
Imagine two major releases of an Asset. Then this results in two 
instances of adms:Asset, and thus two instances of dcat:Dataset. While 
DCAT seems to hint that all versions of a dataset be lumped in one 
instance: there's no occurrence of "version" in the entire document, and 
there are chunks like
http://www.w3.org/TR/2013/WD-vocab-dcat-20130312/#Property:dataset_update_date
http://www.w3.org/TR/2013/WD-vocab-dcat-20130312/#Property:dataset_frequency
that are quite strongly suggesting that every new version of publishing 
of a dataset shouldn't lead to a new instance of dcat:Dataset.

=== End quote===

So...

A: if I modify a dcat:Dataset is it a new dataset?

B: If I modify an adms:Asset is *that* a new asset?

If the answers to A and B are different:

1. is adms:Asset still a subclass of dcat:Dataset?
2. should ADMS show dcterms:modified as a property of adms:Asset?

If the answers to A and B are the same - should we make editorial 
changes to improve clarity?

Phil.

Received on Wednesday, 31 July 2013 11:51:31 UTC