Re: A Question about Metadata storage

[David Jeske <jeske home chat net>]

On Tue, Aug 18, 1998 at 05:06:13PM -0400, Dries, Joseph wrote:
> 	I know I kind of added a bit of fuel to the discussion with some of
> my earlier comments, but I was just curious to what extent options for
> storing the metadata were looked at. Some people are advocating storing them
> in the filesystems (ext2 for example) that support that functionality (re:
> HPFS from OS/2 and their Extended Attributes), or having a data-file on fs
> that don't support it (what OS/2 did on FAT filesystems); others have
> suggested NeXTstep-like .app/.lib directory wrappers (which makes this
> completely fs independent AND has the side-benefit of working over different
> OS's fs mounted via NFS, SMB, etc.). 

Just for the record, when I brought up the NeXTstep stuff, I wasn't
suggesting we use that structure for storing metadata. My logic was
something like this:

We don't have anything today which would use metadata. One important
use for metadata would be the whole filetype/app thing. So before we
really need to be able to store the mime-type of a file in metadata,
it would be nice if we had a robust way to install apps, and have them
export inforamtion to the system. This information includes
information about what mime types they can open.

So, I was advocating that we move to a NeXT-like encap/wrapper system
for installing items on the system, so that we no longer have to edit
text configuration files for mime-type->app mappings every time an app
is installed.

Then, down the road, when this nice filetype->app automapping scheme
is in place, it'll make more sense to have a metadata system which can
store the 'real' mimetype of a file, instead of just inferring the
mimetype from the extension.


-- 
David Jeske (N9LCA) + http://www.chat.net/~jeske/ + jeske@chat.net