[Desktop12.04-Topic] Deeper Zeitgeist integration. Installation of datasources for default applications etc

[Jo-Erlend Schinstad]

Den 19. okt. 2011 08:24, skrev Didier Roche:
> [snip]
>
> Now some questions:
> - do you have automatic testsuite for them, running on different 
> versions of upstream projects?
> - how do oyu work with firefox in particular, where we update the 
> released version through release life? We generally avoid shipping 
> plugins for this reason.
> - can you elaborate on one of the major flaw of zeitgeist which seems 
> a bigger priority to me: when you plug an usb key, or have a 
> windows/ubuntu partition, as zeitgeist isn't a indexer, we can't see 
> them in the file lens in particularly. I know that Seif has a script 
> for that, but it doesn't seem to be suited for indexing and Mikkel has 
> some concerns about it. Can we put that on the table as one of the 
> priority for Precise?

I don't think this is simply a technical issue. It's first and foremost 
a design issue. When I have opened a file, then you can know that the 
file is of some interest to me. The fact that I haven't open a file, 
doesn't prove that it isn't interesting, but you just can't know. I 
regard Zeitgeist is a logger that enables applications to learn from my 
actions, not as a general indexer like Tracker. In order for the dash 
and lenses to be effective, I think it should primarily display files 
I've shown some interest in. Similarly, the web lens should only display 
sites I've actually visited, not intermingle results from Google, since 
I haven't shown any interested in all those other sites.

Searching for the unknown is completely different from searching your 
personal history. The thing I like most about the current way the lenses 
work, is that no results are ever entirely irrelevant, since at some 
point, I've chosen to use them all. I'm very concerned that mixing these 
types of searches will introduce many false positives, which will reduce 
the user experience. Searching for things you've never used is obviously 
quite useful, and an interesting field that should be treated as a 
separate topic. Because of its nature, you'll want the ability to define 
a lot of parameters for such a search, and I'm not convinced that lenses 
are ready for that. These are some of the parameters that the lens would 
have to have in order to provide a good search for unused things:

     *  Name (duh)
     *  Time created (from and to)
     *  Time modified (from and to)
     *  Specific folder(s)
     *  How deep to search
     *  Specific servers (nfs, samba, ftp, etc)
     *  Size (to and from)
     *  User or group the file belongs to
     *  File type
     *  Whether or not to search file files content
     *  Source (did you download it from the web, received it in email, 
bit torrent, etc)

These are only the parameters that immediately comes to mind. I'm sure 
there are many more. But already, this has become a fairly long list, 
and it's likely that you'd want the ability to store that search. From 
my perspective, it seems that forcing these types of searches into the 
dash will both reduce the quality of results from my log, and reduce the 
ability to search for things I've never used. For that reason, I would 
recommend that the dash be used only to search for things that are known 
to be interesting because it's been used, and that a more powerful 
desktop search engine be developed separately. Obviously, this 
application would be able to use the same data sources that are used in 
the dash, but would provide much greater level of detail. Then the dash 
could use stored searches from that app as a source, because then you 
have defined an interest, so it's no longer random data, and the results 
will still be relevant.

Does it make sense to you? :)

Jo-Erlend Schinstad