simple search api (was Re: mimetype standardisation by testsets)

[Joe Shaw]

Hi,

On Mon, 2006-11-27 at 23:02 +0100, Thiago Macieira wrote:
> Joe Shaw wrote:
> >Shouldn't this be the job of the MIME detection code instead?  To
> >inspect the container and say "this is audio/x-vorbis+ogg" or "this is
> >video/x-theora+ogg" ?  That way it works correctly desktop-wide.
> 
> So, video/x-theora+vorbis+vob-subtitles+ogg ?

No, of course not.  There's are degrees here.  If I am a file manager,
and the user clicks on an ogg file, what application should open it?  It
is very possible that a different app handles theora video from the one
that handles vorbis audio.  Whose responsibility is it to make that
determination, and how much information is needed to make that
determination?

A better example would be XML.  You want inkscape opening SVG files, but
not XSLT files.  The indexer's code will be very different for SVG files
than it would be for XSLT files, even though both are parsing XML.

> Hell, no. If a format allows for compound data (like message/*), the 
> indexer should be able to read and understand what that compound data is.

There's no question about this.  It's the indexer's responsibility to
extract the relevant information from the file, but fundamentally what
kind of file that is (again, think SVG vs. XSLT) is something the MIME
code should handle.

Joe