Re: A Proposal for Metadata

[Tom Tromey <tromey cygnus com>]

Daniel> I tend to agree with your opinions here, but I also tend to
Daniel> think that they are far too file oriented.

FWIW: I assume that we are always talking about "files" in the VFS
sense.  So the use of this word, "file", is really just a shorthand.
I agree this isn't clear from the document I wrote.

Daniel> My reflexion comes from Web metadata (the one I'm most used
Daniel> to), and especially the work done at W3C on metadata:
Daniel> http://www.w3.org/Metadata/

I read the RDF syntax spec.  This is interesting stuff, but it seems
to complicated for what we're doing.

Daniel> I beg of seamless access of remote data from my future Gnome
Daniel> desktop, metadata handling included !

I agree.

Daniel>    - Metadata is Data, so to your question:
Daniel>      "If the implementation represents metadata as a file
Daniel>       associated with the parent file, can that metadata file
Daniel>       have metadata itself?"  Daniel> I would clearly answer
Daniel>       yes !

The problem with this answer is that you must then give the metadata
itself a name.

Daniel> - Second the metadata is often far more useful than data (I
Daniel>      guess in average we do more ls on files than we look at
Daniel>      or edit their content).

Well, the metadata is used more frequently, but that doesn't make it
more useful.  This probably sounds like a distinction without a
difference.

Let's put it this way: if my metadata is destroyed (in the Gnome
model), things still work, though perhaps not as nicely.  If the data
itself is destroyed, chaos results.


Anyway, I think in general we are thinking the same things.  I agree
that every addressable object should have metadata.  But I don't think
we should deal with designing this until we actually have some notion
of what these addressable objects will look like.  I don't mind if we
have to change the (VFS-) file metadata implementation at some later
date.

Tom