On Thu, Jan 23, 2014 at 10:08 AM, Jon Phipps <[log in to unmask]> wrote:
> Hi Ben,
> On Thu, Jan 23, 2014 at 4:48 AM, Ben Companjen
> <[log in to unmask]>wrote:
>> Returning an HTML document (or XML document as I get) in
>> response to a request for an RDA property or class is wrong in the Linked
>> Data sense [note 1]. This is explained in the W3C WG Note that you
>> referred to in recipe 2 [2].
> I'm the co-author of that note, so I'm all too familiar with it. :-)
> At the moment, it shouldn't be possible to request html from
> without getting redirected to (hosted
> on github using github pages). Although I'm doing a minimal job of checking
> the HTTP Accept header.
>> Are you planning on introducing 303-redirects?
> I'm deeply embarrassed (really) by the fact that the redirect is not a 303
> and that it may not be consistent. As well as by the fact that it doesn't
> return the requested fragment (which I still believe is best practice). So,
> yeah, as soon as I get back from the ALA Midwinter conference (sooner if I
> can get some meeting-free time). I'll at least get a 303 redirect header in
> there (still learning nginx).

Oh. I'm going to take a guess that this announcement was pushed out to
meet an ALA Midwinter deadline, and therefore was a tad premature.

If that's the case (or even if not), why not market it as a beta,
collect up the known bugs in a visible place, and (perhaps most
importantly) invite the denizens of the W3C Public Linked Open Data
mailing list to weigh in on the opaque identifiers vs. meaningful
identifiers vs. both opaque + meaningful direction? You want this
vocabulary to be adopted and used; it would be really good to have
their buy-in to the vision.

In my opinion, I think it would be a mistake to continue with the
opaque identifiers as the primary identifiers; the vocabulary is
almost unreadable as it stands. And I believe they will make
communication between people trying to implement it harder, as they
continually struggle to translate to (because you *know* that
someone will insist on communicating in opaque identifiers: insert
flashbacks to "You know, 100 $a $4 act!"). If you *really* want to
keep the opaque identifiers as an option, you could invert everything
so that the "owl:sameAs" assertions identify the opaque identifier
instead, and make the rest of the assertions target the meaningful

Oh, and on that note there's another technical bug to add to the list:
the owl:sameAs assertions appear in the RDF/XML document, but they do
not currently appear in the Turtle document.