RE: An updated draft of the schema.org/Action proposal

On Wednesday, November 13, 2013 11:36 PM, Sam Goto wrote:
> On Wed, Nov 13, 2013 at 12:06 PM, Markus Lanthaler wrote:
> > I'm still worried about the
> > whole ActionHandler stuff as I've already explained in the past. This
> > RPC-based model is quite anti-Web and thus I would like to see this
stuff
> > more aligned with how the Web works, i.e., the manipulation of resources
by
> > the exchange of state representations.
> >
> > I find this draft is a steps backwards in that regard as it couples the
data
> > expected by an action to the action itself:
> >
> >  "Each action has corresponding arguments/slots/parameters that
> >   are well defined. Actions define a standard programmatic
> >   predefined interface between parties (e.g. which arguments
> >   "Watching a Movie" takes), and ActionHandlers helps with the
> >   Mechanisms (e.g. invoking an action via an android intent vs
> >   a HTTP GET)."
>
> Just to be very clear, this specific regard (the fact that the schema
defines
> the arguments/parameters) is consistent with every single draft we put out
> in the past (you can find all the earlier drafts here). There are no
changes
> in this draft on this subject.

You are right(ish). What I meant is that in previous proposals the action
handler had requiredProperties/optionalProperties (just as GMail Actions
have) where as the latest draft doesn't have that anymore. Instead it seems
to suggest that the Action itself specifies which data is to be send to the
server. In the previous drafts the properties on the action looked like they
were describing the action in more details and not defining the interface to
the server.
 

> > Do you really expect to, e.g., have different actions to rent a
> > skiing shoes from renting a house?
>
> I expect there will be high coverage of common nouns and verbs in
> http://schema.org long term. We have a http://schema.org/RentAction,
> which can be used in combination with nouns like
http://schema.org/SkiingShoes
> or http://schema.org/House as these become problems that really need
> to be modeled.

Yet, RentAction "defines" only the two properties landlord and
realEstateAgent itself. This is hardly useful when renting skiing shoes and
will probably confuse developers.


> For the long tail of problems, the extension mechanism can help us
> understand verbs/nouns that are not yet in http://schema.org.

Agreed, even though we need to move away from the current extension
mechanism advocating non-resolvable IRIs
(http://schema.org/docs/extension.html) and instead move towards a more
linked data driven approach with resolvable IRIs.

 
> > Currently RentAction's "parameters" according to your
> > draft [2] are "landlord" and "realEstateAgent". What's the rationale
behind
> > this decision? I think the sole purpose of the action itself should be
to
> > convey the semantics of what happens or, in other words, what the
> > consequences I can expect when I invoke an action.
>
> I think that should be one of the goals, yes. In addition to that goal, I
> believe it is important for actions to define which arguments/parameters
> they take.

That may turn out to be very difficult IMO.


Cheers,
Markus 


--
Markus Lanthaler
@markuslanthaler

Received on Monday, 18 November 2013 10:19:42 UTC