It’s been a while since we defined the first prototype architecture for itsme OS, at the beginning of this year. Of course, implementing it whe found some ways to improve it, by using flexible solutions, better tools than the ones we first adopted, and by consolidating the elements providing us with ground for innovation. The following diagram shows what we have now in our latest prototype (which is actually quite different from the one we showed you in our latest video – as you can see from the post header… and maybe we’ll do a new video soon).
itsme OS architecture, november 2009
Just to mention some of the major changes in this version:
- we moved from Nepomuk stuff (Virtuoso, Soprano, Strigi) to Tracker 0.7: there are several reasons for this change, basically we are aiming at having a strong implementation and a strong reference community. The impact we had from Maemo Summit and by meeting some Tracker developers convinced us that we are on the right path.
- we adopted a model-state-view architecture for the itsme front-end: we are aiming at having an implementation that can dynamically and easily adapt to changes. We’ve made some of the engineering features that could allow multi-platform interaction slimmer (since they have lower priority at the moment), and gained a great amount of speed and scalability.
Tags: architecture, diagram, features, Guglielmo, itsme os, prototype, tracker
Hello, The ItsMe approach looks very promising. That’d be interesting to know more about the reasons why you decided to move from the Nepomuk libraries. In what sense do you consider that the implementation and the community are not strong enough? Also, do you plan to support data model customization by end users – i.e. addition or extension of ontologies? This is a built-in capability of Nepomuk, which we are promoting in the Mandriva desktop, while Tracker’s backend doesn’t seem to have this level of flexibility, or has it? Greetings.
Hi Stéphane,
.
Nepomuk libraries are too high level for our needs.
We tried to use soprano with sasame2 backend, and we found it quite slow, and it eat many resource.
Also we tried to use the soprano inference model, but it’s quite limitative for our use case.
Tracker seems to be faster, we can use sparql update to modify/delete things in metadata store, that’s nice, and we love the miner architecture for fetch metadata all around
We met some tracker devs at maemo summit, and talking with them we understood that their roadmap match our needs. Nokia is working hard to make tracker avaiable also on mobile system, that can be very useful for us.
This is some of the things behind our choice.
If you want more info, don’t hesitate to ask
Michele
Hi Michele, Thanks for your input. Have you given a try to the Virtuoso backend as well? It’s much faster and more lightweight than Sesame. As for inferencing, it is getting more deeply integrated into Soprano currently (see [1]), and Virutoso has good inferencing capabilities. Feel free to give your take to the Soprano devel mailing-list! The Soprano and Nepomuk communities would probably be glad to know about your goals and see how the API can evolve to match your needs in the future. Cheers.
[1] http://sourceforge.net/mailarchive/forum.php?thread_name=200910211012.59544.trueg%40kde.org&forum_name=soprano-devel
Stéphane
Hi,
how I can see the prototype?
Tks,
Antonio
@Antonio: At the moment we are just showing videos on this website, and presenting the prototype in interactive demos live (at conferences and workshops). The code cannot be downloaded (yet), but we could arrange for a demo if you happen to be in range
[...] itsme OS (Prototype) architecture [...]