Re: whine about dctypes:Software

On Wed, Jan 16, 2013 at 5:03 PM, Paolo Ciccarese
<paolo.ciccarese@gmail.com> wrote:

> FOAF has always been our first pick, we are on the same page for that.
> The issue came up when talking about Software Agents, what do we do there?
>
> We can certainly say:
> :paolo a foaf:Person, prov:Agent;

That is the style intended by PROV (perhaps with prov:Person as well)
- that more custom ontologies like FOAF would be used to describe
people. prov:Person is just a 'marker' class and there are no
implications for properties and states, it is therefore not directly
related to other *:Person classes, although foaf:Person is an obvious
candidate for real use.


In PROV we felt we did not want to mandate use of FOAF because we
wanted PROV to be mainly self-contained (but extensible), particularly
due to the different serialization formats. An agent being a person or
a piece of software felt like an important piece of information for
the sake of provenance. I think this is also the case for provenance
of annotations.

In practice PROV WG would recommend people to use FOAF to describe
people. For organizations we know there are also other competing
vocabularies. For software there are many various degrees of
granularities ("MS Word", "MS Word 6 for Windows", "MS Word on my
particular machine", "MS Word as started today") and we did not want
to delve into that, and so only made prov:SoftwareAgent.  We actually
had a discussion if we should also have a prov:HardwareAgent or a more
general term for them both - robots and machines after all also can do
considerable causation in the real world - but we dropped it as it
felt a bit more exotic.



> But for a software agent we would have:
> :tool a prov:SoftwareAgent
> with no link to FOAF as PROV does not link them.
> Are we happy with this?

foaf:Agent can still be inferred from the range, right? You can also
still say foaf:Agent if you want.


> The other option would be:
> :tool a oa:SoftwareAgent (subclass of foaf:Agent).

I don't see a big gain in making yet another SoftwareAgent. We've got
enough *:Agent already, and as we see that is giving some source of
confusion.  This would just raise questions about oa:Person etc.



-- 
Stian Soiland-Reyes, myGrid team
School of Computer Science
The University of Manchester

Received on Thursday, 17 January 2013 09:15:40 UTC