This looks basically good but I think it is worth distinguishing between behavior towards the author & towards the reader. There should be two modes, selected by an attributes in the exhibit html, that says whether various problems should lead to warnings/reports or should be silently handled. Eg right now an extra comma throws you to the exhibit validator which is good for authors but bad for readers---for readers, you just want that comma dropped.
David Huynh wrote: > Johan Sundström wrote: > >> On 3/18/07, David Huynh <[EMAIL PROTECTED]> wrote: >> >> >>> Keith Alexander wrote: >>> >>> >>>> You can give properties a "uri" property. (The JSON->SPARQL->JSON php >>>> script I mentioned before stupidly made use of an "iri" property; I'll >>>> change this.) This lets you use proper vocabularies for your properties. >>>> >>>> You can give each item a "uri" property, but all this does is give you >>>> the 'real' uri in the html display (for URI:). You still get the >>>> uri-as-fragment-id in the rdf:about. >>>> >>>> >>> It's a bug in the exporter... What you're doing--giving explicit >>> URIs--is correct. I need to fix that exporter. >>> >>> >>> >>>> Also, even if this did work (one >>>> could presumably modify the exporter script to do this), Exhibit doesn't >>>> use the uri as the id (it defaults to the label), so you've got to have >>>> a lot of bloat and duplication in your json if you're using the uri as >>>> the Exhibit id. Is there a way to tell Exhibit to use another property >>>> as id? >>>> >>>> >>> We could potentially make Exhibit detect if an ID is already a URI, in >>> which case no namespace is prepended. >>> >>> >> Would this proposition make the id generation work like this, the way >> I would presume/hope? >> >> * items with a given id property get that id. >> Otherwise: >> * if there is a given uri property, copy that to id. >> Otherwise: >> * if there is a given label property, copy that to id. >> Otherwise: >> * fail miserably. >> >> Or what do you have in mind? >> >> > I'm thinking of something like this: > > * an item with a given id property gets that id > otherwise > * if there is a given label property > * if that label hasn't already been used as an id, make that the id > otherwise, still make that the id but give some warning? > otherwise > * ignore that item and issue some warning > > * an item with a given uri property gets that uri > otherwise > * if the item's id looks like a uri, make that the uri > otherwise > * append the item's id to the page's url and make that the uri > > Whatever scheme we decide to go with should maximize convenience and > minimize error. > > David > > _______________________________________________ > General mailing list > [email protected] > http://simile.mit.edu/mailman/listinfo/general > _______________________________________________ General mailing list [email protected] http://simile.mit.edu/mailman/listinfo/general
