From kde-devel  Sat Sep 13 10:07:56 2003
From: Marco Krohn <marco.krohn () gmx ! de>
Date: Sat, 13 Sep 2003 10:07:56 +0000
To: kde-devel
Subject: Re: Storage implementation in KDE
X-MARC-Message: https://marc.info/?l=kde-devel&m=106344789020958

Hi,

> Why bother blurring the line at all? Just keep the traditional
> hierarchical organization. 

The traditional hierachical organization is illogical in the sense that it is 
not always possible to form proper subsets of categories. For instance does a 
picture of Albert Einstein belong to the folder physics/pics or pics/physics?
You do not have these kind of problems in a storage implementation.

> UI wise, the user is going to pick one item 
> out of that menu, so it doesn't matter much if its a folder or a
> category. My main concern is the UI. I really want to see a general
> purpose UI for this kind of stuff before I irrevocably commit all of my
> filesystems over to a database.

About two years ago I spent some thought on this, and I think it is not 
impossible to implement a storage which is user friendly. Since we have 
almost all the information that is contained in a traditional file system 
(+some extra information) it would even be possible to rebuild a traditional 
directory (dynamical hierarchy).

For example: if there are 50 images "pics physics einstein", 30 images "pics 
physics teller" and 30 papers "physics high-energy" than by counting how 
often the categories are used we find:

physics = 50 + 30 +30 = 110
pics = 50+30 = 80
einstein = 50
...

From that it should be possible to show at the top-level a number of 
directories (number depending on personal preferences), e.g. 

physics/
pics/
(einstein/)

physics splits up to:

physics/pics
physics/high-energy
(physics/einstein)

and so on...

Using additional free available information such as hypernyms ("is a kind of") 
would give additional meta information at no costs. For example have a look 
at the (free) wordnet:

http://www.cogsci.princeton.edu/cgi-bin/webwn1.7.1?
stage=2&word=tulip&posnumber=1&searchtypenumber=-2&senses=&showglosses=1


> A mockup with test data would be excellent to see.

I very much agree.

best regards,
  Marco


-- 
Marco Krohn
Theoretical Physics 
University of Hannover
 
>> Visit http://mail.kde.org/mailman/listinfo/kde-devel#unsub to unsubscribe <<