[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-look
Subject:    Re: Metadata use ideas collecting... :)
From:       Dave Leigh <dave.leigh () cratchit ! org>
Date:       2002-05-15 0:59:29
[Download RAW message or body]

On Tuesday 14 May 2002 17:00, Friedrich W. H. Kossebau wrote:
...
> And what do you expect from metadata?
>
> Friedrich

OK, since you ask... I'll try to do this with much emphasis on the "Look and 
Feel" since that IS the list's domain. I'll also do this as what I'd like to 
see as an ultimate result, understanding that there would be many steps along 
the way...

First, I'd go a lot farther than you described using the virtual filesystem 
database described earlier in the thread. My major complaint is that "file 
management systems" in general do a whole lot of STORING, but next to no 
MANAGEMENT. Lotus Notes has a system that does a lot of what we're talking 
about on the KDE list. Metadata are stored as a record, the file is stored as 
an attachment. Notes will do a full-text index on the file attachments. 

The text index is a must. I know from my time on a newspaper that people 
request articles and backissues like this: "I'd need to find the article 
about the..." Note that the query doesn't include the title or the author. 
People remember content, and in fuzzy terms. To a publication (especially a 
long lived one), being able to research past articles is a must for 
consistency and accuracy. I've already described how it's useful for IT 
development.

The thing is, this sort of thing needs to be next to be transparent (and 
failing that, very translucent). 

When you create a "directory" in the database you're actually creating a 
record to which other records that will be linked. It's basically nothing BUT 
metadata. This gives you the option to describe the files that will go into 
directory. Any file created here inherits this description, which can be 
further modified with a right-click on the icon and selecting "Properties." 
Drag a file to a new directory and it acquires the metadata for that 
directory as well. For the most part you don't have to do anything special. 
And a full-text index ensures that you don't need to manually define a 
bunch of keywords in order to have a reasonable chance of retrieving 
the correct file on a search.

And if you're going to put it in a database, you might as well include 
revision control (it will come in VERY handy, below). I visualize a file 
manager where the icons of files in use have an overlay of a user (similar to 
the "arrow" overlay on shortcuts), and a right-click can tell you who's got 
the file checked out, and can allow you to check out or check in a file. Look 
at the file properties dialogue and you should see a tab called "revisions" 
that will allow you see the revision history and select a revision to view or 
to which to revert. This is regardless of whether or not the application 
itself supports revision control. The revision control system can borrow 
technology from CVS to merge concurrent changes from multiple editors.

So much for putting things in. Now for retrieval:

While you could do a simple query like you describe, I think that 
technology's at a stage where it's time to start thinking in less restrictive 
terms. Imagine a File Search on steroids that understands natural language 
queries.  Then tie that in with IBM ViaVoice ("Get me all the recipes that 
include portabello mushrooms or lamb.") Using existing technologies from 
Oracle, Google, and IBM this could be done in (relatively) short order. All 
of the technologies exist; they need to be incorporated.

And a very cool benefit is that using the CVS features IN CONJUNCTION with 
the extended search capabilities you could do Lotus Notes-like replication. 
For an enterprise, this means that frequently accessed files can be 
replicated on geographically local fileservers. For an individual it means 
you can take your work with you. Like Briefcase, only better. You could 
replicate the files located by query. You could actually SAY to your 
computer, "Give me a local copy of all of the files related to project number 
10221." or (for a previously saved query) "Update my replica."

THAT's what I'd like to expect out of a full-featured solution including VFS. 
For legacy filesystems I don't expect much more than you do.


-- 
Dave Leigh, Consulting Systems Analyst
Cratchit.org
  http://www.cratchit.org
  864-427-7008 (direct)
  AIM or Yahoo!: leighdf
  MSN: leighdf29379@hotmail.com
  ICQ: 37839381

An actor's a guy who if you ain't talkin' about him, ain't listening.
		-- Marlon Brando

[prev in list] [next in list] [prev in thread] [next in thread]