Asemantics: A new generation of feed aggregators for Web 2.0 applications is being jointly developed by Asemantics and the British Broadcasting Corporation. As a first step in the process, Asemantics has completed the aggregator engine for the Memory Share service of the BBC.
From the puff piece:
All information is stored internally using the RDF model.
...because the most important thing about using RDF is that you keep it to yourself.
Asemantics has employed a novel approach based on Atom and optimised RDF/SQL data storages. Atom-based and REST interfaces allow for easy recombination of building blocks into a Web2.0 application
Sorry, I was put off by the “HOLY SHIT HERE COMES A PDF: Presentation” until I realized it was a Stylish script I’d written a long time ago:
@namespace url(http://www.w3.org/1999/xhtml); a[href$=".pdf"]:before { content:"HOLY SHIT HERE COMES A PDF: " }
Lessee:
We have a winner!
My uneducated guess: they’re polling Atom feeds, lossily converting them to RDF, serializing them as RDF/XML, and storing the serializations as text blobs in a MySQL database.
At least they didn’t spell it “ATOM”.
"My uneducated guess: they’re polling Atom feeds, lossily converting them to RDF, serializing them as RDF/XML, and storing the serializations as text blobs in a MySQL database."
According to page 14 of the presentation, you’re basically correct until that last step. There are two SQL databases, one that stores and indexes the fields they’ve designated as ‘searchable’, and another one that uses Redland to store the actual triples and from which the metadata can be retrieved for search results (using the URL as the joining key between the two DBs). This design suggests they started with everything in Redland but found query performance was poor.