W3C home > Mailing lists > Public > semantic-web@w3.org > January 2010

Re: Alternatives to containers/collections (was Re: Requirements for a possible "RDF 2.0")

From: Sandro Hawke <sandro@w3.org>
Date: Fri, 15 Jan 2010 09:31:07 -0500
To: "Michael Schneider" <schneid@fzi.de>
cc: "Jeremy Carroll" <jeremy@topquadrant.com>, "Semantic Web" <semantic-web@w3.org>
Message-ID: <21200.1263565867@waldron>

> >I think the attractiveness of RDF is minimalism and quietly deprecating
> >containers without any explicit replacement is quite doable.
>
> What would "quietly deprecating" mean?

Perhaps we could run a survey of (self-identified) experts, asking which
practices they recommend.  If everyone, individually, says "don't use
feature X", and we make that fact public, that's a pretty good
un-official deprecation.  It could also, of course, guide a WG in making
such a change official.

If we're going to do this, I'd like it to cover more than containers.

Here are some practices that data providers might follow, which I've
heard people discourage.  For each of these, we could ask experts (you
folks) whether these are ever a good idea or should be universally
discouraged.  I expect people providing RDF data could benefit a lot
from know which of these practices are considered bad by 100%, 50%, or
0% of the experts.  (Some of them would need a more full explanation
than this, I think.)

     - using alt/bag/seq in your data
     - using parsetype collection or rdf:first/rest/nil in your data
     - using OWL cardinality to close off your instance data (bldg has 17 rms)
     - using rdf's reification vocabulary
     - using blank nodes
     - using plain literals (no datatype, no language tag)
     - using type xs:string
     - using language-tagged strings
     - having the order of triples affect the meaning of data
     - having the order of triples affect performance
     - having your data imply something new if triples are omitted
     - using IRIs for which there is no dereference mechanism (eg tag URIs)
     - using http IRIs which are 404 or otherwise are not linking data
     - leaving out rdfs:label information for some terms 
     -   ...

Any more ideas?  Should we make it about both bad and best practices, or
just bad/not-bad like I've done?  In some cases, I put in the negation
of a best practice, as a bad practice, but maybe that's too confusing.

       -- Sandro
Received on Friday, 15 January 2010 14:31:11 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 19:48:05 UTC