Re: Mono bindings a blessed dependency? [Was: Tomboy in 2.16]

[Jamie McCracken <jamiemcc blueyonder co uk>]

Ross Burton wrote:
> On Fri, 2006-04-21 at 17:58 +0200, Emmanuele Bassi wrote:
> > [1] Dublin Core anyone? Timestamps not using ISO8601 but something that
> > is similar-yet-not-quite-enough?
>
> Now look what's happened.  My main gripes are:
>
> 1) no generic way to say "title".  

Thats true - Im not sure why thats needed though.

You can search specifc metadata or a full text search on all fields



There is File.Description,
> Audio.Title, Audio.Comment, Doc.Title, Doc.Subject, Doc.Comments,
> Image.Title, Image.Comments and Image.Description.  What if my generic
> interface just wants to show the title of a result?  Ditto for many
> other fields: Doc.Keywords, Image.Keywords but no Audio.Keywords.

The interface is class based so all files have File.* attributes.

File.Keywords would be user definable emblems, leaftag tags etc
Doc.Keywords are an official embedded OOO/Ms word metadata field
Likewise with Image.Keywords.

Audio.Keywords is not a metadata (they inherit the File.Keywords)

The class namespace is supposed to differentiate them but I take it you 
are saying thats not sufficient?

> 2) as I mentioned in my other mail, until yesterday the File.* date
> fields were missing their timezone.  As Emmanuele points out are now
> almost but not quite ISO 8601, apart from Audio.Date, Audio.LastPlay,
> Doc.Created and Image.Date which still don't have a timezone, and really
> should.

okay will fix
> 3) "For images, most support the EXIF standard and so a subset of this
> makes up part of this specification.".  Why a subset?  What about the
> rest of the EXIF metadata?  My photos contain the exposure time and
> focal length, so why can't I use tracker to search for all long exposure
> photos?

Point taken! (I know Im proably wrong thinking the rest were just junk!)


> Considering Tracker has a RDF Query API, I'm shocked that it doesn't use
> a RDF triple store internally.  You can create a MySQL-based triple
> store using librdf in a few lines of Python, which reduces the effort in
> writing a beagle/tracker clone to the metadata extraction libraries, and
> nice front-end APIs.  Arbitrary metadata can be added without any extra
> work, and it would build on the years of work by semantic web people.

I know that but the reason tracker is so fast is it has a highly tuned 
DB architecture. Technically Beagle's lucene should be faster than 
mysql's full text indexing but I use a few optimisations in my design to 
ensure that full text searches are super fast (I wont bore you with the 
details!)

 We use a custom RDF parser because its dead easy to map RDF query to 
SQL and I can add extensions to my RDF Query parser to support full text 
searches and other stuff.

Using an official rdf lib would negate the above and force a fixed 
structure on the DB. So I went for speed and flexibility.

> Rant over.  The first person to say "stop energy" gets their username in
> a SJ blacklist. :)

Nothing can stop me now!


--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/