DALI examples: Vocabulary and a spec bug

[Markus Demleitner]

Dear DAL, Dear Semantics,

DALI examples are a way to give machine-readable examples for service
usage; currently, it's mostly used with TAP.  If you know the
"service-provided" entries in the Examples popup in TOPCAT's TAP
window: That's filled from DALI examples.

I have just built support for its continuation property (which lets
you build, in effect, hierarchical examples lists) into DaCHS, and
while doing this, I first noticed that DALI 1.1, sect. 2.3 says:

  In future, this URL [http://www.ivoa.net/rdf/examples] may resolve
  to an RDF vocabulary document in human and machine-readable
  format(s) and if so will contain the normative list of values for
  the property attribute and will supersede the ones specified below.

Well, with Vocabularies 2 we have the infrastructure, and though
there's little immediate benefit, I'd say we ought actually provide
this vocabulary.  And so I have added DALI Examples as a draft
vocabulary to the RDF repo, <http://www.ivoa.net/rdf/examples>.
This covers the terms introduced by DALI as well as the ones from
TAP.

In terms of procedures, this is a bit tricky; normally, a vocabulary
comes into existence when a standard needs it, and it is reviewed
together with the standard.  In this case, the two standards
contributing are already passed and reviewed, and we certainly don't
want to wait for the next versions of the documents before we review
this vocabulary.  I'd hence suggest to the TCG that they should
review it as a belated part of the reviews of DALI 1.1 and TAP 1.1.
Would anyone be really strongly opposed to this slight bending of
rules?

Meanwhile, I invented the labels and descriptions for the existing
terms (and the vocabulary), as they have not been written into DALI
or TAP.  Feedback and fixes on these are very much welcome (as usual,
the vocabuary URI gives you a rendered version when retrieved in a
web browser: <http://www.ivoa.net/rdf/examples>).  I am particularly
grateful if these come as PRs against
<https://github.com/ivoa-std/Vocabularies>.

And then I have a brown-paper-bag embarrassment.  DALI 1.1 has to say
this on continuation in sect. 2.3.4:

  ...the link elements must contain the following additional attributes:
  a property attribute with the value continuation, *a resource
  attribute with an empty value* (referring to the current document),
  and the href attribute with the URL of another document formatted
  as above (i.e. another collection of examples that clients should
  read to collect the full set of examples).

(emphasis mine).  The examples below that language are missing the
property attribute.  Which is fine, because, as it immediately turns
out when one feeds such a document to an RDF processor, setting
resource produces bad RDF triples.

What it should be is, in Turtle:

  <> example:continuation <http://dc.g-vo.org/rr/q/ex/examples>

That is: http://... is a continuation of the current document (that's
what the "<>" means).

With resource="" the result is (the equivalent of):

  <> example:continuation <>

That is: The current document is the continuation of the current
document.  This is clearly not what should be conveyed.

Sure enough, RDFa says on the meaning of the resource attribute:

  @resource: a URIorSafeCURIE for expressing the partner resource of a
  relationship that is not intended to be 'clickable' (also an
  'object'); 

So, it's clear that the resource attribute must not be used here.  I
*think* this is my text, and I can't really say what I had thought
when I wrote it (dumb excuse: it was before I really did anything
serious with RDF).

As repentance for having written a nonsensical spec, I'll keep
wearing a brown paper bag until the next Interop, and I offer writing
an erratum which would simply say: "Drop 'a resource attribute with
an empty value (referring to the current document),'".

Are there opinions on that plan?

      -- Markus (from below the brown paper bag)