Re: SOSA/SSN namespaces - ontology

I agree, Option 8 is technically no more difficult to implement than any other option. As you mentioned, probably easier than Option 1 and Option 2.

However, personally I am concerned about the fact that there is an OWL import in SOSA, meaning if you use Protégé or some other editor, you will not only get all the additional axioms for a term, but these additional axioms may refer to terms that are defined only in SSN full and as such, the tool will even import SSN. This is probably something Kerry is concerned with and why she is thinking of a method to separate parts of the axioms out that refer terms that are not defined within SOSA. Which I think is actually impossible in this solution. The fact that at the end, in most tools you will end up with all axioms of SSN when you import SOSA is the reason that I, personally, prefer Option 1 and Option 2 (and even Option 4 and Option 5) over Option 8.

From: Rob Atkinson <rob@metalinkage.com.au>
Date: Thursday, 9 March 2017 at 8:15 am
To: SDW WG Public List <public-sdw-wg@w3.org>
Subject: SOSA/SSN namespaces - ontology
Resent-From: <public-sdw-wg@w3.org>
Resent-Date: Thursday, 9 March 2017 at 8:16 am

In SDW plenary Kerry expressed view that Option 8 looked elegant but we wont have time to implement.

I personally dont see any qualitative difference - or if anything Options 5 and 8 are probably easiest to implement in time available.

So I have offered to implement - as i think its simpy a matter of splitting SSN into tow files:
sosa-owl.ttl - anything that just adds sosa axioms without mentioning SSN
ssn.ttl - everything else

this will IMHO only take a few minutes if SOSA + SSN in two namespaces is available.

Looking at the ontology files however I don't see any artefact where SSN adds such axioms,

Please point me to such a thing if it exists...

If not, then I would propose:

Option 8 allows us to stabilise sosa and ssn - and add the axioms to sosa-owl.ttl in a separate process. (i think every other option would require us to version at least SSN files as this work is done, but option 8 would allow us to split out the task of axiomitisation)

If it does exist, and someone points me to an artefact we all agree is the current modelm then I think I can whip up the option 8 implementation quickly :-)

Rob