[okfn-discuss] implications of licensing restrictions on ontologies

[Rufus Pollock]

Jo Walsh wrote:
> dear all,
> 
> i am back in write-only mode again but wanted to braindump a bit 
> about the question of what happens when ontologies are under licensing
> constraints specifically noncommercial use, but also attribution.
> 
> http://www.ordnancesurvey.co.uk/ontology/ OS geodata ontologies (e.g.
> taxonomical/logical models intended for partial reuse in structured 
> data description). They are being released under a CC-NC-BY-SA license. 

This is going to get interesting. The crucial thing to distinguish here 
is the ontology as idea and the concrete expression of that ontology in, 
say an RDF/OWL file. What is difficult here is that the two seem 
practically identical (how can I copy the given ontology without copying 
the actual expression). One answer would be if you copied the ontology 
but changed the naming (but then one would be wary of various Software 
related cases that have found copyright infringement where the infringer 
used substantially the same structure in their code).

> What gives if an ontology is licensed for noncommercial use only (and
> has an SA clause?)
> - If i subclass something in it for my own use does that become a
>   'derived work'?

My guess would be no. You after are just subclassing the idea (though 
referencing a specific implementation as released by the OS). Of course 
if you were to take their actual file and add your ontology at the 
bottom of it then yes you would have created a derivative 
(copyrightable) work.

> - If i make a data set using its terms does that become subject to the 
>   same license constraints e.g. that data cannot be then be subject to
>   commercial use (because it depends on the OS ontology?)

I doubt it. Again it is the question of dependency. A book of criticism 
of say Arthur Miller (picking a playwright who is still in copyright) 
may certainly  require one to reference his work but is not a derivative 
work. However this is a good question and just shows the negative effect 
of using non-open licenses in creating gray areas of uncertainty (cf. 
Harlan Onsrud's lenghty comment on your original blog post).

> - What happens if instead of subclassing i make a direct, exact copy  
>   at a different URL and maintain it in parallel. Am i infringing some
>   "copyright" on the information stated in a model - an abstract model
>   of how towns relate to counties relate to regions doesn't really 
>   seem copyrightable. So exactly what is being copyrighted in the
>   ontology, if not "only the model available at this URL" (which is
>   what the OWL definition implies... 

In that case you are making a derivative work in the copyright sense and 
would be infringing I think. As you say the underlying abstract model 
clearly should not be copyrightable but the expression might well be. 
this is definitely a difficult question and to where the idea/expression 
divide kicks in in a case such as this.

> But when i had a look at it, the RDF itself still states that the model
> is Crown Copyright. I pinged a contact at the OS's semantic web group
> to ask to fix, and enquire what possible justification there is for
> imposing a noncommercial use clause on an OWL ontology...
> http://www.ordnancesurvey.co.uk/ontology/AdministrativeGeography0.1.owl

I think that is the crucial point. Why on earth aren't they just 
releasing this stuff PD or similar ...

> The geodata database rights question is covered in the Guardian and
> goes round the geoblogosphere, that is just great to see.
> http://www.allpointsblog.com/archives/2708-Report-OS-Data-Not-as-Copyrighted-as-We-Thought.html

Yes we got the same information in a comment on the blog. Someone 
should go off and read the report that claim is based on.

I also think it might be worth starting to draft a data rights FAQ on 
the wiki.

Regards,

Rufus