[Digikam-users] How to remove "left behind" tag tree

[Jean-François Rabasse]

On Tue, 20 Nov 2012, Marie-Noëlle Augendre wrote:

> Thanks for your comment, Jean-François.
> 
> I think it NEEDS to be fixed.
> Otherwise, even if one cleans the mess one day and begin again from scratch,
> he's stuck to keep the same tags structure forever; or, he will move/add
> something and begin another mess!
> 
> Who is completely sure the tags structure he (intends to) uses just now will
> be the same he wants in some future. Not me, for sure...

Not me neither :-)

And it's not only a matter of tags structure but also of tags language.
The same tags, or indexes, may have different representations in different
languages, even if they refer to exactly the same things.
As an example, if one wants to keep the location where images have been
shot, e.g. country and city, this is definitive information. But the
tags list may differ.
In French :
paris  -> Lieu/France/Paris
vienne -> Lieu/Autriche/Vienne

And if I tag for English users, the same locations become :
paris  -> Location/France/Paris
vienne -> Location/Austria/Vienna
or in German :
paris  -> Ort/Frankreich/Paris
vienne -> Ort/Österreich/Wien
etc.

It's a very complicated problem and my personal option is to use
a vocabulary of keys. Can be a controlled vocabulary or just a
managed vocabulary, but something unambiguous.
Keeping track of the keys (somewhere in XMP data) and the tags
translation from a vocabulary file allows to modify the definitions
of keys, reorganisation, language switching, and rebuild the tags
structure.

Tags tree are only useful to make organisation more readable and
logical, but in fine, the important thing is the terminal keywords,
all what is used to define a « subject ».
Applications that use keywords, use a subject flat list of words,
never hierarchy. E.g. a web albums builder may insert in generated
XHTML code some definitions as :
  <meta name="keywords" content="xxxx, zzzz, ..." />
(That's one of the reasons why I prefer to have metadata in files,
to allow other applications to use the work done.)

In the above example, from the same keys, paris or vienne, the final
subject could be Paris or Vienne, or Paris or Vienna, or Paris or Wien,
depending on a vocabulary translation table. But the keys are eternal.
The important thing, IMHO, is to keep aware of the difference between
a tag semantic and a tag structured representation.

But this is off topic anyway :-)

Regards,
Jean-François