[ckan-discuss] Comments on DCat spec: naming of Distribution

Thu Apr 19 23:25:47 BST 2012

On 19 April 2012 13:27, Fadi Maali <fadi.maali at deri.org> wrote:
> Hi Rufus, all,
>
> my two cents inline…
>
> On 11 Apr 2012, at 09:49, Rufus Pollock wrote:
>
>> Hi Faadi, Richard, Oly and Phil,
>>
>> I'm writing to you guys directly because the email list mentioned on
>> the DCat spec page seems incorrect: "undefined at w3.org" :-)
>>
>> I'm sorry I missed the call a few weeks ago. In catching up I've been
>> working through the spec as currently is:
>>
>> http://dvcs.w3.org/hg/gld/raw-file/default/dcat/index.html
>>
>> I have one immediate concrete suggestion:
>>
>> Rename dcat:Distribution to dcat:Resource.
>>
>> Why? Distribution has strong sense that this is a distribution of the
>> dataset -- i.e. the data in the dataset in some concrete form.
>> However, in reality what is being termed distributions can include
>> APIs, saved queries or even visualizations of the dataset in addition
>> to what would be a classic distribution: e.g. zipped version of the
>> data in the dataset. (See also this email from Richard regarding
>> "types" of Resources in CKAN [1]).
>>
>
> I agree it is not perfect. However as long as this provides a way to access the data it is *not too bad* to call it distribution
>
>> Why Resource? We thought about the naming of this quite a bit in CKAN
>> and the best we came up with was Resource. I understand that Resource
>> is perhaps an overused term in RDF (and computing generally) but I
>> think it is probably the most appropriate in this circumstance -- that
>> said, we could perhaps think of a better one (better than both
>> Distribution and Resource).
>
> My argument against Resource (in addition to the fact that it is overloaded) is that it implies everything related to the dataset such as documentation, its void description, its homepage etc.

I hear you but we could be stricter on that with a dcat:Resource being
"data"-related (though including visualization / apps might be nice
...)

> IMHO, distribution is still better, not perfect though. A name better than both would be great but I can't think of any!

I hear you though I do prefer Resource to Distribution for reasons
outlined above. Distribution really has a strong overtone that is
inaccurate here.

>> (I wonder whether we may have had a role in the naming of this since I
>> remember back when we first discussing DCat the CKAN project, inspired
>> by software, had been using concept of a Distribution of a (Data)
>> Package where the Distribution equated to the serialization of a
>> Dataset into some concrete form (zipfile on disk)).
>
> Actually to a high degree yes :)

:-)

I guess that is a point here in that Distribution model is yet to take
hold while a Resource style seems more prevalent.

Anyway we can keep thinking on this (and I have a bunch more DCat
suggestions / comments).

Rufus