In comments, Adewale Oshineye pointed me towards Aggrevator, which is a MySQL-backed desktop aggregator that can handle lots of feeds/entries, at least 5000/500,000. Now we're talking. The tool has a simple thumbs up/down rating system, which helps focus on the feeds of interest.
Ade also asks how RDF (and Attention.xml) could help with a tool like this. Heh, won't be the first time this has cropped up on this blog. The answer is long, but for now I'll say there are two aspects that can be of use: data modelling and interop. The model can be made richer much more easily, rather than adding a table for every different aspect of the posts you want to model/store (e.g. tags), with a triplestore you already have support. You may have to pre-process the data (say the tags come in a <category> elements), but you'd have to do that anyway.
For interop, you can immediately share your data with other people without having to make up a new application-specific format. Instead, you can describe your resources in a language designed for the purpose. A good example there would be the info in the ratings part of the system. If I've not seen a particular feed before, I'd like to apply the ratings Ade has already compiled, because I trust his opinion (it says so in my RDF profile). The systems with which you share don't have to be the same software, or even the same kind of applications - how about including data from Film Trust?
I've not yet installed Aggrevator (I want it on my kiloFeed list), but had a nosey at the source - the SQL schema is very simple. Once I've got my current batch of stuff out of the way I might just have a look at moving that to a triplestore. Then maybe incorporate tags and the Attention stuff ;-)
Ok, I really must get some work done today, but Aggrevator does
look very much like something I've been looking for. It's GPL btw.
Couple of questions, someone may be able to help save me trawling:
What's the current status of Kowari? - Is there a nice smallish,
everything-included (HSQL maybe?) binary? Does it have
SPARQL yet? Or
Jena be a better
approach? What about named graphs/contexts in this (errm) context?
Or what about RDFX? (Aargh, 404′ing) That's hot on info-integration, and I believe is SWT-based like Aggrevator. Might it be better to use the reification approach to attribution like Rich talks about the slides? (Somewhere near that 404).