Dont quite remember how, but I’ve stumbled upon this post suggesting
to use Plone as a semantic aggregator, which referenced a thread
where I participated just about the time I finished Marshall.

What’s nicer is that from there I got to here, which brought me to
mapping semantic web data with rdbms, which is not exactly what I
was looking for (well, I wasn’t actually looking for it actively), but
may give some interesting hints.

One of the projects I’m working on these days is about deploying
content to remote locations, using standards-based formats. The
testbed uses Plone as the frontend and EnSimpleStaging, which is
connected by another tiny product to the deployment system.

So what’s the relation with RDBMS and the Semantic Web you ask.

Well, I would like to use Zemantic somehow. However, the ZODB backend
is sub-optimal because it can only be accessed with Python, and the
deployment system goal is to be able to access content in many
different languages.

The solution? Building a RDBMS backend for Zemantic. At first sight it
sounds very simple. I guess the simplest representation of a triple
store would be a 3 (or 4) column table. However, it doesn’t sound like
that will scale. We probably want specialized tables for each kind of
node, that maps to integer primary keys, and then to a global
identifier. I’m not quite sure yet of the approach. What’s funny is
that most of the articles I’ve found so far talk about the reverse:
exposing RDBMS content as RDF.

Need to send an email to Michel Pelletier. If I recall he had an
article about building a lightning fast RDBMS backend for a triple
store.

Update: Found some information about scalability and about
Database schemas for storing RDF. Seems also that 4Suite has
support for storing triples in a RDBMS. Will have to dig more into
that one.