datalink-terms

[Kristin Riebe]

Hi Pat, Arnold, Norman and all the others,

just a few comments from my side on the terms, though I must admit that
I haven't looked closely at DataLink yet:

* "mirror of" and "derived from":
The W3C provenance model (see http://www.w3.org/TR/prov-dm/) provides
the terms "wasDerivedFrom" for entities (data sets) that are derived
from an activity.
Additionally there are  "AlternateOf" as well as "SpecializationOf" to
specify if a data set is an alternative or a specialisation of another
data set.

I'm not so sure if we really need to distinguish between original and
mirrored instances (they should be exactly the same, in fact, and
treating them separately may cause more work than necessary); but if we
do this, I suggest to use the W3C terms "wasDerivedFrom" and
"AlternateOf". "AlternateOf" could also be used for e.g. data sets in
different formats, representing potentially the same thing.

And I like Norman's suggestion and would define them as subproperties of
#from.


* #derivation for forward links:
W3C uses the term "Derivation" in the passive form ("wasDerivedFrom",
http://www.w3.org/TR/prov-dm/#term-Derivation), i.e. it is used there
for pointing backwards to where something is coming from.

So I would like to avoid confusion when just using #derivation and
specify more explicitly that this is a forward link, e.g. using
#derivedEntities or similar.

Cheers,

Kristin







On 10/21/2014 10:53 PM, Patrick Dowler wrote:
> 
> Yes, I had missed out on the provenance related terms. I like the idea
> of having the generic top-level term(s) now and can add specifics later.
> 
> Normally, we think about having a link from a product back to the input,
> but it is equally valid and potentially more interesting to provide a
> link to a (better) derived product.
> 
> In that light,
> 
> #derivation describes a forward link to a product aka "is a #derivation
> from #this" +1
> 
> It's kind of a "you might be interested in this thing over here..." :-)
> 
> For the backward link to "inputs" I find #predecessor pretty vague as
> there are many sorts of sequences, not just provenance... predecessor
> also implies that the "original" has been replaced or made obsolete...
> something less specific than #input would probably be good. #precursor
> and #progenitor are ok, but a bit fancy sounding and I'm tired of going
> around thesaurus circle... anyone have a better word for this?
> 
> 
> 
> 
> On 21/10/14 03:50 AM, Markus Demleitner wrote:
>> Let me propose #derivation as a top-level label for the merged-order
>> file in the split-order datalink.  I can't see myself liking
>> #predecessor or #ancestor for the files some reduced data set is
>> derived from, so better ideas  are solicited.
>>
>> Having #predecessor and #derivation (or whatever) would, I'd argue,
>> do for now.  Further details would be up to a proper provenance data
>> model or actual software using those wonders.
> 

-- 
-----------------------------------------------------------------
| Dr. Kristin Riebe
| eScience & GAVO
|
| Email: kriebe at aip.de
| Phone: +49 331 7499-377
| Room:  B6/25
-----------------------------------------------------------------
| Leibniz-Institut für Astrophysik Potsdam (AIP)
| An der Sternwarte 16, D-14482 Potsdam
| Vorstand: Prof. Dr. Matthias Steinmetz,  Dr. Ulrich Müller    	
|
| Stiftungsverzeichnis Brandenburg: 26 742-00/7026
-----------------------------------------------------------------