Re: Database / XMP discussion

["John Stowers" <john stowers lists gmail com>]

All,

> I do agree on the spirit of FreeDesktop, but am not quite sure I agree
> with it fully. There are a number of standards for how to store metadata
> in XMP format, and they are not duplicated (in general anyway, main
> difference is the title/comment field). Why not use the standard
> NameSpaces consisting of
> * EXIF Data
> * Dublin Core
> * Iptc4XmpCore
> Which is what F-Spot is doing for the moment.
> If we start to use FreeDesktop namespace, then we would have to duplicate
> the information since currently not many applications uses this one.

Sorry I didnt explain that very well.

I was just suggesting that the 20 or so pieces of metatdata that
Freedesktop selected appear to cover a useful subset of all those
fields present in EXIF, Dublin Core, and iptc4XMPcore. I.e. The 20
most useful that should cover 95% of the use cases.

They would naturally be written back into the file in the most
appropriate natural format and namespace (XMP, EXIF, etc).

By selecting the 20 most common fields it allows the content of said
fields to be stored in the DB and wrote back to the file, in the most
appropriate natural format/namespace when appropriate.

> I do agree that F-Spot should store a lot of the existing tags that are
> embedded in the photo. Mainly EXIF (Camera, Model, Flash, Size, Portrait
> etc) related tags, but any others that F-Spot handles also of course.
>
> This would enable F-Spot to do very fancy searches :)
>
> One other option would be to use Beagles database (or interface) but
> personally, I would rather Beagle would use F-Spot (if it existed) than
> F-Spot beeing dependent on Beagle (which some people perhaps do not want
> installed)

Perhaps, my opinion is that I prefer Tracker, but a good database
abstraction will make both possible.

> Version Database should probably have more information.
> * Path to version (versions could be in different locations perhaps.)
> * Version Created?  (A cropped photo's modified timestamp is probably
> different from when it was imported, as well as different from when you
> did minor touchup on it)

I agree

> Having Rating as a tag should be workable (as we have stated in bugzilla),
> but it would mean making some very special codes for it in the tag code.
>
> On the other hand, we should enhance the tag code so we can ALWAYS (no
> matter what language) find Location, City, Country, State etc to store
> these tags in the proper XMP field (not tags). But then we also force the
> user to have these special tags (which is not in current F-Spot idea)
>
> This propably means that Tag database has to have three special fields.
> * Visible (rating tag should not be visible
> * Deletable (special tags should not be possible to delete)
> * ID (So F-Spot can find it, irrelevant of language)

Agree completely

John