W3C home > Mailing lists > Public > public-rdf-dawg-comments@w3.org > January 2011

Dataset, Graph Store, etc.

From: Gregg Reynolds <dev@mobileink.com>
Date: Wed, 26 Jan 2011 07:32:13 -0600
Message-ID: <AANLkTinyCMyZ=QHnKtD2hZpoDJL+neGsDVnC4ag_9QEN@mail.gmail.com>
To: public-rdf-dawg-comments@w3.org
Hi WG,

Just a quick note regarding some concepts and terminology.  I'll leave
detailed pros and cons for later; for now I'd just like to suggest the
following to put it on your radar:


   - RDF data, data content, etc.  Suggestion:  replace all uses of "data"
   with "resource".  The obvious reason is to maintain consistency with other
   W3C docs; the less obvious reason is that it really is about resources.
   - RDF data store and dataset.
      - I see no essential distinction between dataset and data store.
       Somewhere a text points out that data stores but not datasets can be
      updated .  That gets it backwards.  It's the language that allows update
      expressions or not; the capabilities of graph database
implementations are
      out of scope.  Or to put it another way, a query/update language
definition
      is supposed to define a language, not a database.  If your implementation
      understands update operations, then it can update its datastore,
which may
      be the same "dataset" used in query operations.  Using two terms
is likely
      to lead to confusion and uncertainty.
      - Suggestion:  replace both with "RDF graph region".
      - "Graph region" captures the essential semantics necessary for the
      language definition, and has a nice clean real-world analog: the piece of
      paper upon which one sketches one or more graphs.  The paper
itself, as well
      as any chunk of space on the paper, can be considered a region.
      - The graph region addressed by an expression in a query/update
      language is independent of any notion of server; but at the same time it
      admits of a simple analog:  the graph region controlled by a
server is the
      piece of paper it uses to inscribe graphs.  Multiple servers may
address the
      same region, etc.; the nice thing about "region" is that it allows us to
      disregard such implementation issues for the language definition, yet it
      also works for talking about implementation issues.
      - Alternative:  graph space.  I prefer "region" for some reason, maybe
      because "space" is too big; I think the intended meaning is
something less
      than the entire RDF graph space.

Thanks,

Gregg Reynolds
Received on Wednesday, 26 January 2011 13:32:46 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 26 January 2011 13:32:46 GMT