Re: Last Call for Comments ... Re: proposal for updates to http://schema.org/Event

On 1 March 2012 02:06, Jason Douglas <jasondouglas@google.com> wrote:
> On Wed, Feb 29, 2012 at 11:50 AM, Dan Brickley <danbri@danbri.org> wrote:
>>
>> On 24 February 2012 16:39, Dan Brickley <danbri@danbri.org> wrote:
>> > I've just posted another draft proposal in the W3C wiki,
>> > http://www.w3.org/wiki/EventSchemaUpdate
>> >
>> > From the wrapper text there,
>> >
>> > "The proposal comes from the Google teams working with the existing
>> > Event markup, and has been checked by the other schema.org partners
>> > prior to publication. See PDF for full details of the proposal."
>> >
>> > * "Proposes 3 new properties of Event: eventStatus, previousStartDate,
>> > previousEndDate to support canceled or rescheduled events.
>> > * Adds eventCategory to support categorised events.
>> > * Supports recurring events by making startDate and endDate repeated.
>> > * Encourages use of existing 'url' property (of Thing) to link to
>> > associated Web pages."
>
>
> I think this last one is worth highlighting for broader discussion too.  As
> I understand it, the question is about whether a) Thing/url is the identity
> of the item (equivalent to itemid and conceptually akin to rel=canonical for
> that specific item) or whether b) it's ok for Thing/url to point to any URL
> that represents the same real-world entity, even if it's a different
> manifestation (e.g., someone else's database record for that entity).

Ah yes, I'd forgotten that nuance. I tend to read Thing/url as approx
what we had in FOAF isPrimaryTopicOf (but more accessibly named :)

But yes, it seems to have more of the flavour of (a), ie. id for the
description/record not the thing it describes.

> My understanding has been a) (equivalent to itemid), and that Thing/url was
> provided mostly as a convenience for being able to markup existing anchor
> tags without having to repeat the URL in the page markup (which itemid
> requires).

A fine motivation.

> However, accepting the proposed change would effectively eliminate option a)
> and mean that Thing/url was instead meant to express equivalence rather than
> identity.  This would also mean that itemids (rather than Thing/url) would
> have to be declared in order to link schema.org objects (meaning specifying
> objects as values by reference rather than nesting them).

Absorbing this, but yes I think I agree.

> The alternative, as stated in the doc, is to create Thing/sameAs for these
> equivalence use cases.  I personally prefer that option.

In terms of the broader issue, not just Event, it feels like we'd
benefit from a little 3 or 4 file set of test cases. Maybe describing
a movie and actor, to show how we want things to connect up, and the
role sameAs might play. I'm used to the role owl:sameAs plays in the
RDF and linked data world, but not sure that's quite what you're
looking for here. Is it a relationship between an entity and another
description of that same entity?

cheers,

Dan

Received on Thursday, 1 March 2012 16:23:18 UTC