[haiku] Re: Need Some GSoC Advice

  • From: Truls Becken <truls.becken@xxxxxxxxx>
  • To: haiku@xxxxxxxxxxxxx
  • Date: Tue, 24 Mar 2009 14:50:19 +0100

Ankur Sethi wrote:

> Although I didn't follow most of the discussion about the finer points
> of OpenBeFS, I gather that most people recommend attaching certain
> attributes to each file (and everyone seems to have different opinions
> on how that should be done). Wouldn't this involve going through each
> file on disk and reading its attributes every time someone runs a
> query? Wouldn't that slow down the process?

No, because BFS can index the attributes. Using a database for the
same values as are also stored in attributes will result in confusion
because the two can and will get out of sync.

Another benefit of extracting the data to attributes is that you at
the same time solve the problem of manually populating attributes for
things like media files.

François corrected me before when he said that the search interface
does not need to know what attributes should be searched because the
mime db already knows this. Add a rule about only searching fields
that are indexed and you are good to go.

I think the storage part is more or less solved. The missing parts are
an index feeder with examples of file type plugins (e.g. mp3 and plain
text) and a library and UI for performing a spotlight-like search.


Other related posts: