Mandriva adds a semantic layer to the KDE 4 desktop [LWN.net]

From:		Mandriva Press <press-AT-mandriva.com>
To:		lwn-AT-lwn.net
Subject:		Mandriva adds a semantic layer to the KDE 4 desktop
Date:		Wed, 4 Jul 2007 11:34:17 +0200 (CEST)

Paris, July 4th, 2007 - For Immediate Release
_____________________________________________________________________

On the occasion of aKademy 2007, the annual conference of KDE
developers and users, Mandriva is proud to announce together with the
NEPOMUK partners and the KDE community that the NEPOMUK-KDE project it
leads is poised to deliver a sea change in the Linux desktop
experience. NEPOMUK-KDE introduces semantic capabilities to the
upcoming release of the K Desktop Environment (KDE 4), providing an
interoperable framework that can be harnessed by all KDE applications
to allow annotating and interlinking any and all desktop objects.

62 years ago, in the famous article "As we may think", Vannevar Bush
aired the dream of a "Memex", a system that would let the user create
associative trails of data, hence becoming an "enlarged intimate
supplement to his memory". In the mean time, the advent of semantic
Web technologies combined with the tremendous progress of database
systems, the exponential growth of digital information stored on
personal computers, the power of graphical toolkits and the
inventiveness of open source communities have produced huge momentum
for a new generation of personal knowledge systems. In 2007, the time
of desktops acting as intelligent assistants has come.

Building upon existing standards and open source technologies, the
NEPOMUK-KDE framework is the first desktop-wide semantic system to
become a reality. The framework provides rich application programming
interfaces allowing for content annotation using well defined
relationships between the information elements and for advanced
content search from any KDE application. KDE Dolphin - the new KDE
file manager - and KDE PIM - an application suite for managing
personal information - count among the early adopters of the
innovative NEPOMUK libraries.

Sebastian Trueg, the leading architect of the NEPOMUK-KDE project,
explains: "NEPOMUK-KDE transforms the data islands currently
encapsulated in each application into a gigantic graph of knowledge
both at the desktop level and soon at the Web level".

With the goals of showcasing the latest advancements of the framework
and involving the community of end users, Mandriva is building a KDE 4
preview image integrating the latest NEPOMUK achievements. "The
objective is to expose the new desktop environment to the largest
number of users, developing its future in close collaboration with the
community it is made for, making the technology truly accessible and
easy to use", says David Barth, Mandriva Consumer Vice-President.

Mandriva and the NEPOMUK Consortium are extending the scope of the
project by bridging existing initiatives related to desktop metadata
management to make semantic features interoperable between different
technologies. Mandriva is also leading the implementation of similar
features on top of the Eclipse RCP and the Mozilla XUL frameworks, and
it is teaming up with XWiki and with the Stockholm Royal Institute of
Technology to design a community semantic help desk with rich desktop
extensions and P2P capabilities that will bring the knowledge sharing
process within the Mandriva community to a new level of efficiency.

StÃ©phane Lauriere, coordinator of the Mandriva Semantic Web
activities, says: "NEPOMUK paves the way for revolutionary
applications. It will considerably improve the efficiency of users in
their day to day work. NEPOMUK will have a direct impact on the way
people work, learn and communicate, empowering them with new methods
for easily creating knowledge, and for discovering contextually
relevant facts."

Francois Bancilhon, Mandriva's CEO, adds: "It is no surprise that the
innovations carried out by NEPOMUK stem from open source communities,
since collective knowledge creation is at the heart of the open source
movement. Mandriva is very proud to be part of a European team who is
realizing the next big thing in desktop computing."

A video demonstrating NEPOMUK's integration into Dolphin, the new KDE
file manager, can be found at DailyMotion:
http://www.dailymotion.com/liquidat/video/x2bf36_nepomukd...

Tudor Groza, from DERI, will present the NEPOMUK framework during a
dedicated keynote session at aKademy 2007:
http://akademy2007.kde.org/conference/talks/58.php

About NEPOMUK

The NEPOMUK project is funded under the European Union's Sixth
Framework Programme. NEPOMUK brings together researchers, industrial
software developers and representative industrial users to develop a
comprehensive and multi-platform solution for extending the personal
desktop into a collaboration environment which supports both personal
information management and sharing of information via social and
organizational channels. Major partners of the project include DFKI,
DERI, SAP, Thales, and IBM (the full list available online, see
project website below). Mandriva is bringing its values to the NEPOMUK
project, sharing its vision for an open, simple and innovative desktop
that is accessible to all.
More information at: http://nepomuk.semanticdesktop.org

About Mandriva

Mandriva, formerly known as Mandrakesoft, is the publisher of Mandriva
Linux, an easy-to-use and innovative operating system. It is one of
the most popular Linux editions in the world. Dedicated to making open
source technologies accessible to all users, the company offers a full
range of products and services to individuals, enterprises and
government organizations. Mandriva products are available online in 80
languages and in more than 140 countries through dedicated channels.
Headquartered in Paris, France, the company is publicly traded on the
Euronext Marche Libre .
More information at: http://www.mandriva.com

Press contacts

Camille Repussard - crepussard@mandriva.com - +33 (0)1 40 41 17 58
Guillaume Martin - gmartin@mandriva.com - +33 (0)1 40 41 17 64
Caroline Garlatti - cgarlatti@mandriva.com - +33 (0)1 40 41 67 67


_____________________________________________________________________
You receive this email because you are in our journalist database. If
you do not want to receive information about us anymore, just reply to
this press release with "REMOVE ME" as a subject.

to post comments

Mandriva adds a semantic layer to the KDE 4 desktop

Posted Jul 6, 2007 1:16 UTC (Fri) by bluegecko (guest, #42312) [Link] (1 responses)

After reading all of the above, I am still not exactly sure what this "semantic layer" actually does. The video ain't much help either (which shows somebody adding tags to files). If this is for searching stuff, how is this different/more effective than Google desktop search ?

Mandriva adds a semantic layer to the KDE 4 desktop

Posted Jul 6, 2007 8:46 UTC (Fri) by ecureuil (guest, #3507) [Link]

Yes it is more powerful than Google Desktop Search because you can store
any kind of relations between pieces of data like :
1) This file talks about John
2) John is my boss
3) This file was an attachment to this e-mail

KDE4 will starts slowly with this new technology with simple stuff like
tagging and searching. It is important that the functionality works well
without impeding your work and scale without an excessive burden.
Analyzers have to be written for all the different types of files and
data. More innovative stuff will certainly comes when the fundations are
laid down properly.

Not The Unix Way?

Posted Jul 6, 2007 9:14 UTC (Fri) by Richard_J_Neill (subscriber, #23093) [Link] (12 responses)

Maybe I'm misunderstanding, but this seems to be a violation of "The Unix Way" i.e. lots of little tools, each good at one thing, *without* complex interdependencies, and where it is easy to change the permutations. Another principle is that (almost) every GUI application has a CLI application at its heart.

An example of this: many photo applications (eg digikam) insist on "importing photos" into their own structures, rather than allowing the user to stick with his own choice of filesystem layout. Will other programs (such as gthumb, which don't impose this) become "second class citizens"? Is there actually a risk of KDE becoming increasingly "incompatible" with the shell?

In particular, I hope KDE doesn't become monolithic: many parts of KDE are excellent (eg kwin,klipper,kdeprint), but there are some bits I prefer to avoid (eg arts,dcop). There are also some excellent KDE apps (eg konsole,konqueror,amarok), but these stand on their own (for instance, konqueror works much better with kmplayer than with kaffeine; amarok is great for playing music, but not for managing it.)

Not The Unix Way?

Posted Jul 6, 2007 9:58 UTC (Fri) by MathFox (guest, #6104) [Link] (3 responses)

The idea behind the "semantic layer" is that there is one interface to store metadata behind objects. If you've made pictures and mp3 recordings of your daughter singing a song, you can link the different media to the event. The event can be part of "July 2007" and "daughter"

IMO it is "the UNIX way" in a sense that small and simple programs (like gthumb) become more powerful because they can make better use of the relations layer, not imposing a structure on their own.

Not The Unix Way?

Posted Jul 6, 2007 11:40 UTC (Fri) by Richard_J_Neill (subscriber, #23093) [Link] (2 responses)

That makes sense - providing that it isn't a KDE-specific thing. I'd have expected it to be more on the lines of something freedesktop.org would provide. We'd also need a good CLI application to match.

However, I still think that metadata belongs *in* the files, not in a database. So if I move/copy a file, the metadata should stay embedded in it.
How does the semantic layer stuff maintain consistency when you do this?

Not The Unix Way?

Posted Jul 6, 2007 12:33 UTC (Fri) by MathFox (guest, #6104) [Link] (1 responses)

If the "semantic interface" is simple enough to use in KDE-C++, it shouldn't be too hard to create a GNOME-C adapter for it.

About metadata belonging *in* the files: In what file do you store a relation between files? Furthermore there is the "minor" problem that many file formats don't support metadata of the format required for the semantic layer, it would require tools that can handle "all file formats". And performing a complete file system scan to collect all metadata is a costly operation. I would think about a database that identifies files by hashes of the file contents; such a scheme could work with "off line storage" too.

Not The Unix Way?

Posted Jul 7, 2007 0:33 UTC (Sat) by Richard_J_Neill (subscriber, #23093) [Link]

You're right about the problems. But I think that the semantic database has got to cope gracefully with editing/renaming/moving of the files, even when they are moved off the computer entirely, and the database loses sync.

Because this is such a hard problem to solve, I'd prefer to do without the semantic stuff, rather than sacrifice the ability to use the unix-shell commands (or risk losing the integrity of the database). [inotify may help a bit here].

Lastly, files which don't have metadata are usually simpler formats such as .txt and emails. So the problem isn't that bad. Anyway, we have grep, and locate (so what more could one ask for?!)

dcop

Posted Jul 6, 2007 11:39 UTC (Fri) by rfunk (subscriber, #4054) [Link] (7 responses)

Funny you mention disliking dcop. That's a big part of what makes KDE
exceedingly scriptable and generally adds a lot of command-line power.
Dig into some amarok extension scripts for some examples.

arts, of course, is scheduled to go away in KDE 4; nobody likes it.

dcop

Posted Jul 6, 2007 12:43 UTC (Fri) by tzafrir (subscriber, #11501) [Link] (2 responses)

Adds command-line power: yes. But only when the GUI is up.

dcop

Posted Jul 6, 2007 15:16 UTC (Fri) by rfunk (subscriber, #4054) [Link] (1 responses)

Er, when the GUI isn't up then KDE and the remote control that dcop provides are
irrelevant.

.... As long as the applications store their data in a way that it can be retrieved without
the GUI (whether with the same program or a different one). That's the key.

dcop

Posted Jul 6, 2007 15:40 UTC (Fri) by tzafrir (subscriber, #11501) [Link]

Remote control of programs is a means to an end. What I need is to be able to do useful things with the data.

Consider an example from the beginning of the thread: images collection of digikam. Or the mail folders of kmail. Or your addressbook.

I'm not saying that the interface that dcop is not highly useful. I do say that it still doesn't help if the GUI itself isn't up.

dcop

Posted Jul 6, 2007 12:57 UTC (Fri) by Los__D (guest, #15263) [Link]

Isn't DBUS DCOP++, but without being desktop dependant?

dcop

Posted Jul 7, 2007 0:24 UTC (Sat) by Richard_J_Neill (subscriber, #23093) [Link] (2 responses)

The reason I dislike dcop is that so few of the controls I want are actually implemented! That isn't always true; some applications provide everything one could want (eg amarok, klipper). For example this is a very useful shell script:

#!/bin/bash
#Pipe something to stdin of this command; it will end up on the clipboard.
dcop klipper klipper setClipboardContents "`cat`"

However, the dcop commands that I really want to have are usually missing. For example, every option in kcontrol should be dcop-enabled. That way, I could write a shell script to automate KDE configuration every time I set up a new system. (copying ~/.kde brings too much cruft with it).
Or, I'd like to be able to have the old (pre 3.4) functionality of the Show-Desktop button back.

P.S. Amarok doesn't need dcop

dcop

Posted Jul 7, 2007 19:52 UTC (Sat) by tzafrir (subscriber, #11501) [Link]

> hat way, I could write a shell script to automate KDE configuration
> every time I set up a new system. (copying ~/.kde brings too much
> cruft with it).

Again, dcop needs the server to be up in order for it to be configured. You'd have ot run KDE first. Which is unlike what you can do with manipulation of configuration files.

(At least you can manipulate configuration files. With gnome you depend on gconf to be up even for mere configuration)

dcop

Posted Jul 8, 2007 14:53 UTC (Sun) by tmg (guest, #46149) [Link]

>That way, I could write a shell script to automate KDE configuration
>every time I set up a new system. (copying ~/.kde brings too much cruft
>with it).
You may want to try kwriteconfig / kreadconfig, which seem to even work
without KDE running. These commands probably do exactly what you want.