Re: [Tracker] Tracker and RSS

[Ivan Frade <ivan frade nokia com>]

Hi Thomas,

First of all, with our idea of tracker-0.7 (and as is prototyped in the
rss-reader example) you should write two processes:

* A daemon monitoring the online activity and translating it into our
ontology, inserting it using sparql

* The "viewer", the window the user sees and uses, which talks with
tracker to get/set data.

El lun, 15-06-2009 a las 01:29 +0200, ext Thomas Meire escribiÃ:

> * The first one would be to implement a plugin for tracker-indexer
> (which is already there, i think) and tracker-extract (which i didn't
> find) and let tracker parse the rss file.

 No, that was the way to do it in tracker-0.6. The "tracker-indexer" in
the 0.7 is going to be just a daemon monitoring and indexing stuff in
the file system... what you need to do is a "tracker-rss" process doing
the same with online sources :)


> * The second implementation would be to fetch the rss file for the
> FeedChannel in my app, parse it there and then push the data to
> tracker.

 Uhmmm yes. You need to have a process that retrieve the online feed,
parse it, translate the content to nepomuk ontologies (our flavour now,
lets hope soon it is upstream) and INSERT it into tracker using Sparql
via DBus. 

 Some implementation details:
* Your daemon can start asking tracker all active FeedChannels (better
than keep that configuration in parallel).
* Your daemon should store the "last update time" in tracker
* Your daemon should delete the entries older than X days (user
configurable)
* It is not forbidden to have your local caches to avoid unnecessary
queries
* Try to parse and save correctly the author of the post! (specially in
planets). That would allow us to link the blog posts with the contacts
in the address book.

 Note that if you write a proper daemon, the data will be in tracker
available for everybody. So different applications will be able to use
those data. (Corner stone of our world-domination plans)
 
 About the viewer, you can write your own, but what about to port one
that already exists (e.g. liferea) to use tracker as backend? that could
save you a lot of UI work. 

> What would be the preferred the option to do this? Any common pitfalls
> i need to be aware of?

 Sounds like a great project and really useful for tracker. Well, if you
could also integrate google reader sources into tracker... (the parsing
part would be trivial: they convert everything into ATOM, but you have
the "synchronization" problem...).

 Feel free to ask in this mailing list, and the IRC channel (#tracker in
FreeNode).


> ps. I know there are a lot of rss readers already and the need for
> another one is very small. I'm just using rss as a case to play with
> tracker to figure out the basics. Something more useful might follow
> when i get a hang of it :)

 Here the interesting part is the daemon, pushing data into tracker so
it is reusable for every application!
 
 Regards, and good luck ;)

Ivan