- From: Harshvardhan J. Pandit <me@harshp.com>
- Date: Mon, 29 Jul 2019 11:12:47 +0100
- To: "pieroandrea.bonatti@unina.it" <pieroandrea.bonatti@unina.it>, Javier D. Fernández <jfergar83@gmail.com>
- Cc: Axel Polleres <axel.polleres@wu.ac.at>, "simon.steyskal" <simon.steyskal@wu.ac.at>, apollere <apollere@wu.ac.at>, public-dpvcg <public-dpvcg@w3.org>
- Message-ID: <d5304361-320b-9590-ef3e-ab9d7e457ca2@harshp.com>
Thanks for the clarity of explanations on these.
Should we go ahead and change the example in the spec to use
sub-classing instead, and also add this rationale to why we prefer
sub-classes instead of instances?
- Harsh
On 29/07/2019 05:45, pieroandrea.bonatti@unina.it wrote:
> Another advantage of purpose *classes* is that it supports data re-use
> for "compatible purposes", which is allowed by the GDPR.
> In general, compatibility can be assessed by lawyers only; with
> purpose classes, controllers and subjects agree a priori on a set of
> "similar" purposes, so later data can be re-used for all such
> purposes without asking subjects for more consent.
>
>
> -------- Messaggio originale --------
> Oggetto: Re: some more comments on the paper drtaft and spec
> Da: "Javier D. Fernández"
> A: pieroandrea.bonatti@unina.it
> CC: Axel Polleres ,"simon.steyskal" ,apollere ,public-dpvcg
>
> I agree with Axel and Piero,
>
> Of course instancing could be another way to do it, but I
> think we early decided to follow the SPECIAL approach and use
> subclassing. This allows for flexibility and reasoning
> strategies as mentioned by Piero. At the end of the day, I can
> imagine that it would be relatively infrequent to have a named
> instance of a purpose per se (e.g. it would be the concrete
> "item", text. etc., reflecting the purpose class), and more
> common to have a full policy using those classes to compose
> its components.
>
> Best,
> Javier
>
> On Sun, Jul 28, 2019 at 12:13 PM Piero Bonatti
> <pieroandrea.bonatti@unina.it
> <mailto:pieroandrea.bonatti@unina.it>> wrote:
>
> Subclassing IMHO is preferrable because it is more general.
>
> It encompasses instances throug singleton classes, when
> needed (for
> example in SPECIAL we are using this approach for dynamic
> consent).
>
> Subclassing gives full flexibility and an amazing range of
> granularity
> choices, including co-existence of orthogonal formalizations.
>
> Best,
> Piero
>
> On 25/07/19 17:27, Axel Polleres wrote:
> > FWIW, I think subclassing was so far the mechanism we
> areed upon (and which is compatible with SPECIAL's
> compliance checking algorithm as well),so I'd prefer to
> keep that...
> > Would appreciate Piero's and/or Javier's comments here!
> >
> > Axel
> >
> > Prof. Dr. Axel Polleres
> > Institute for Information Business, WU Vienna
> > url: http://www.polleres.net/ twitter: @AxelPolleres
> >
> >> On 25.07.2019, at 17:24, simon.steyskal
> <simon.steyskal@wu.ac.at <mailto:simon.steyskal@wu.ac.at>>
> wrote:
> >>
> >> isn't it just personal preference though?
> >>
> >> while it certainly makes sense to use sub classes for
> more generic purposes, I wouldn't create a sub class for
> each and every purpose..
> >>
> >> just my 2 cents,
> >> simon
> >>
> >> -------- Original message --------
> >> From: apollere <apollere@wu.ac.at
> <mailto:apollere@wu.ac.at>>
> >> Date: 25/07/2019 07:00 (GMT+01:00)
> >> To: public-dpvcg@w3.org <mailto:public-dpvcg@w3.org>
> >> Subject: some more comments on the paper drtaft and spec
> >>
> >> While Harsh and myself are working on the paper draft
> for ODBASE (again,
> >> feel free to also comment/help),
> >> I was reading over the spec text for personal data
> categories again,
> >> where it says:
> >>
> >> "We therefore suggest to declare the specific context
> as an instance of
> >> one or several dpv:Purpose categories and to always
> declare the specific
> >> purpose with a human readable description (e.g., by
> using rdfs:label and
> >> rdfs:comment)."
> >>
> >> I think this is wrong, because it is not an instance,
> but a subclass. I
> >> reformulated that whole paragraph in the paper draft
> (but not yet in the
> >> spec):
> >>
> >> "DPV provides a list of suggested purposes which may be
> extended
> >> as shown in Listing ~\ref{lst:purpose-example} by
> subclassing existing
> >> purposes to create more specific ones: as regulations
> such as the GDPR
> >> generally require a specific purpose to be declared in
> an understandable
> >> manner, we suggest to such declare specific purposes as
> subclasses of
> >> one or several \texttt{dpv:Purpose} categories and to
> always declare the
> >> specific purpose with a human readable description
> (e.g., by using
> >> \texttt{rdfs:label} and \texttt{rdfs:comment})."
> >>
> >> This should also be changed in the spec.
> >>
> >> Likewise, the example in Listing 2 (Example 2 in the
> spec) uses
> >> instantiation instead of subclassing...
> >>
> >> :SomePurpose a dpv:Purpose ;
> >> rdfs:label “Some Purpose” ;
> >> dpv:hasSector dpv-nace:M72 .
> >>
> >> Isn't that also an error and should be subclassing?
> >>
> >>
> >>
> >>
> >> Axel
> >>
> >>
> >
> >
>
>
>
> --
> Javier D. Fernández García
> jfergar83(at)gmail.com <http://gmail.com/>
>
--
---
Harshvardhan Pandit
PhD Researcher
ADAPT Centre
Trinity College Dublin
Received on Monday, 29 July 2019 10:17:54 UTC