I’ve attended the recent FOSDEM 2009 (great as always), where a number of presentations triggered a lot of my interest.
First @DebianRoom where Lucas presented UDD, the Universal Debian Database. This database groups facts about the Debian project, to ease the creation of queries on what’s happening in the Distribution. This is for instance very helpful for QA tasks, like counting bugs with certain characteristics, or comparing packages in various ways.
Note that a complementary presentation by Enrico was very interesting, on DDE : Debian Data Export (edit: see also his post on DDE), showing ways to offer services to query UDD.
Another presentation, @CrossDesktopRoom introduced the Flossmetrics database, which is collected out of many libre software projects, by extracting contents of the project data from the hosting forges. Very much interesting, in particular since the data becomes available, and a large number of projects allow researchers to compare them in many ways.
Maybe Flossmetrics could benefit from data coming from the Debian UDD… or vice versa ? I think contacts have been taken to think about potential future interchange between the 2.
A general criticism I could make on these two databases is that their schema (the tables & columns layout, as well as the eventual relations), and the code of the data “harvesters” is the only way to understand the real meaning of these data. There’s not so much semantics. Sometimes for known reasons, because, as explained by the UDD developers, there’s actually much incoherence in some of the Debian tools already, and it still it happens to deliver 
I’m thinking of a way to produce similar databases of facts (results of queries on these) with Semantic Web standards, to try and convey some bits of commonly agreed semantics, hence fostering interoperability of these databases, and maybe allow comparison of facts relating to different projects. Edit: actually, this idea has been in the air for a year, as readers of a provious post may remember. See in particular the paper by James Howison presented last year at WOPDASD 2008.
It happens that Mandriva, as a followup of the Nepomuk project is indeed trying to setup such a database (called SWIM at the moment. Edit: it has been renamed to MEPHISTO. See my other post on it.) with the use of RDF ontologies, to store facts and annotations about its distribution (more details here). In the HELIOS project, we’ll certainly try and investigate the use of such techniques to try and manipulate such data, like bugs for instance.
I’m thinking about providing an access to UDD with the use of a SWIM-like service, so maybe we can imagine things like more linking of facts about packages, people, bugs and such between Mandriva and Debian, for instance.
Note that at the FOSDEM there were also interesting presentations relating to these kinds of semantic techniques, both relating to outcomes of the Nepomuk project : one about the integration of KDE 4.2 in Debian, where tools like Soprano were mentioned, and another about Tracker in Gnome (which I haven’t attended) about the same kind of techno on the Gnome side.
The future seems semantic, somehow… and we have then a lot of work ahead of us. More to come.
