It should be highly file-type aware, to offer a preview of the file and to populate recommended tags. A list of handlers should all be given the chance to look at the file and suggest zero or more tagging schemas, and zero or more previews for the GUI. The user can then select a tagging schema with a radio button and then tweak the tags by hand, then import. There should also be scope for plugins to offer data-stream filters along with the preview UI, such as image rotation or re-encoding media files. Items selected for importing should be buffered into a persistent queue file until an explicit "upload this batch" button is pressed, causing them to go in as a single library delta in the archive, and the source files to be deleted.

alaric added on 2012-05-04 10:42:17 UTC:
I've had some further thoughts on this. I feel it's important to get files into the vault as soon as possible, even if they can't be manually classified. So a better workflow might be:

1. Bulk-import files into a specified archive tag; the system assigns automatic metadata from the files as they go in, and specifies workflow=needs-audit as well.
2. The snazzy GUI tagging tool is based, instead, around updating the metadata of existing objects in the archive, and supports advanced search, including the ability to easily produce all workflow-needs-audit objects in a list for examination.

When importing, we should try and guess MIME types for files.

Directories should be "inode/directory"

See http://standards.freedesktop.org/shared-mime-info-spec/shared-mime-info-spec-latest.html#idm140625828597376 for some other interesting MIME types.

The manifest generator outlined in [33fd928177] is a sufficiently snazzy archive importer.

This ticket refers to a GUI (HTML-based?) for searching for groups of objects in a given archive, and then performing tagging operations upon them.

That should include adding values to properties, removing properties, and overriding properties (remove all existing values and then add one or more new values).

Viewing the audit trail for a given object to see its previous properties would also be good, but might be a new ticket.

For image tagging in particular, it might be nice to include thumbnails of objects we can thumbnail. Either way, a link to download the object will be useful for all types.

A library of suggested properties, based on the MIME type, will also help for rapid tagging of common types.