W3C home > Mailing lists > Public > semantic-web@w3.org > March 2016

Re: Handling multiple rdfs:ranges

From: Ross Horne <ross.horne@gmail.com>
Date: Thu, 3 Mar 2016 16:52:25 +0800
Message-ID: <CAHBrK_gDQzOQGYnj2DJuMxzD5EgtZHGAPvFe8dWRk9PgrKTAig@mail.gmail.com>
To: Hugh Glaser <hugh@glasers.org>
Cc: SW-forum Web <semantic-web@w3.org>
Hi Hugh,

I'm getting your hair-shirt feeling! At least, with another modality, we
can express the necessity of never copying the LoD-list from this thread;)

There is basically no perfect solution to multiple ranges just a least bad
option. Correctly, the RDF WG chose that least bad option.

It is unavoidable that people can have different intuitive models when
publishing data -- a problem aggravated by when data published with
different intuitive models is ... linked. Perhaps the LoD community is well
placed to offer solutions to this problem:

 - Option 1: stick mainly to safer things such as triples, SPARQL,
dereferencing and coreference (the status quo).

 - Option 2: offer tools for Linked Data consumers that comply with the
recommendation on multiple ranges whenever possible. However, when a
publisher strays from the unenforceable recommendation, the tool makes an
effort to accommodate the weaker model implied (usually subconsciously) by
the publisher rather than rejecting the data or reverting to the status
quo. Of course, such a feature should be a mostly transparent part of tools
used by consumers, making the consumer's life easier.

Option 2 is more an accommodating work-around for a corner case than an
ideological shift; and doesn't need any new standard, since the standard
already makes the best recommendation possible.

Best,

Ross


On 1 March 2016 at 20:25, Hugh Glaser <hugh@glasers.org> wrote:

> Thanks all, I find myself really enjoying this discussion.
> (Well, in a sort of hair-shirt way!)
Received on Thursday, 3 March 2016 08:52:57 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 3 March 2016 08:52:59 UTC