Ressources documentaires pour Mandriva Linux et les Logiciels Libres

Billets dans la catégorie Linux

Mandriva 2009 Spring et les futures orientations de Mandriva

Ouf, je viens enfin de terminer de mettre

Is MIPS support an opportunity for Mandriva ?

In the first post of my Thoughts about Linux marketing series, at one point I talk about the fact that Mandriva was running under MIPS CPU with the G-dium netbook from EMTEC. Today I’ve read a very interesting article from Ars Technica about Andro

Thoughts about Linux marketing #2 : Mandriva and Grid-computing

Cloud-computing is the magic word theses days. But cloud-computing is based on Grid computing. So what Mandriva have to do with Grid computing ?

From wikipedia definition, Grid computing is the application of several computers to a single problem at the same time – usually to a scientific or technical problem that requires a great number of computer processing cycles or access to large amounts of data. Few people know about it, but Mandriva is involved in the Xtreem OS which aims at investigating and proposing new services that should be added to current operating systems to build Grid infrastructure in a simple way. What’s interesting with Xtreeem OS is the fact that even PDA ( or netbbok ) could be used in the grid system. The main objectives of the XtreemOS project are :

  • To build a reference open source Grid operating system based on Linux for PCs, clusters and mobile devices
  • To provide a simple Grid API compliant with POSIX while adding new functionalities and supporting Grid-aware applications
  • To identify fundamental functionalities to be embedded in Linux for secure application execution in Grid environments
  • To develop a set of self-healing OS services for secure resource management in very large dynamic grids
  • To aggregate cluster resources into powerful Grid nodes by integrating single system image mechanisms in Linux
  • To build an XtreemOS flavour for mobile devices enabling ubiquitous access to Grid resources
  • To validate the design and implementation of the XtreemOS Grid operating system with a set of real use cases in scientific and business domains on a large Grid testbed
  • To promote XtreemOS software and create communities of users and developers.

Partners like EADS, EDF, Nec, SAP, Red Flag and Mandriva are working together to build an ubiquitous grid computing system based on Linux. ISO can be download on the Mandriva mirrors, and RPMS for the Xtreem OS are available. Xtreem OS is based on Mandriva 2008.0, but it seems that a new version may be based on the Mandriva 2009.0. Xtreem OS features list is impressive and can be consult at One notable feature is XtreemFS , a distributed object-based filesystem that can be mounted from anywhere in the Internet, POSIX-compliant, with client-side caching support and available for various Linux distributions, MacOS X and Win32. It’s strange to see that when people are talking about cloud or grid-computing, we don’t have people from Mandriva giving their advice, and showing the experiences that they have accumulated … At least Mandriva present the Xtreem OS project at Salon Solutions Linux 2009. A video interview is also available on Vnunet french website

Interview d’Anne Nicolas la VP Engineering de Mandriva

Eh oui, il m’arrive de poster en fran

About Apple App Store like GUI for Linux

Last time I said that Linux had the technology to provide Apple App Store implementations. Please note that Apple is not the only one providing theses kind of features, as Microsoft, Nokia, Palm and RIM have their own. Of course presently this is done only in the mobile space. But what about doing this on desktop and workstations ? For example if Androïd based netbooks are really going to be shipped, then I’m pretty sure that Androïd Market will be available too. What preventing normal computers from having an application store ? Nothing.

Now look at what is needed to provide an application store :

  • A place to store and published the applications
  • An UI easy to use and allowing users to browse the application catalog
  • A way to allow users to pay and installing paying applications

Every Linux distribution on earth have the ability to store and published applications : just have a look at distributions FTP repositories. Now we just need the right UI, and really this may not be so hard to develop. Xandros understand it already as they are providing in their Presto system an Presto Application Store. If you look at Presto Application Store, you will noticed that the differences with common Linux package management systems are few but very important :

  1. Only the words software or applications are used
  2. Applications are sorted by categories, but you also have sub-categories like « Most popular », « New releases » or « Recommended »
  3. Proprietaries applications ( Acrobat Reader, Skype ) as Free ones are mixed. This is not new, and the same happen with current Linux packages management programs
  4. Only graphical applications are shown, and the notion of packages or dependencies are hidden. Indeed, no need to talk about this in theses kind of UI : just install the needed packages and tell that the application have been installed.
  5. Applications are represented with big icons. The only others informations which can be seen are : the size of the application ( I guess that the dependencies size are not taken into account ), the author ( Free applications could use the Packager or URL tag ) and the rate. IMHO the description of the application is missing, this could be the summary tag or eventually the description one. However as this is a web interface, if you click on the application icon, you will have access to the detailed description of the application, the license information, and some screenshots

I must admit that I’m somewhat skeptic about using a web site to install application, I’d rather use a normal application. Indeed a possible limitation of a web based UI would be the fact that web site doesn’t know which applications are already installed or not on your system, and then it will display unneeded applications. However it’s true that developing an application with theses kinds of UI under Linux may be hard. Indeed several issues will make the task hard : interface theming, ability to jump to applications details information, ability to display applications screenshots and show them at bigger size, and of course animations effects. Now the needed framework are coming slowly to Linux : Clutter for Gtk application and Plasma ( or Qt QGraphicsView ) allow now to designed complexed interfaces with animations effects support. With Webkit support, you can even embedded a web interface in the application or use it to display external resources.

Contrary to popular belief, this is not the first time that a Linux distribution is trying to provide theses kind of features. If you look closely, you will noticed that Presto Application Store is based on Xandros CNR technology. OpenSuse have also 1-click install but which is still too much technical ( you see the entire package name, while installing an application will be started and it will show too much informations ). Mandriva also had the same kind of technology than OpenSuse in the past.

Please note that I don’t think that classic packages managers UI should be removed. IMHO they should be used by advanced users or for advanced task, like installing servers or CLI applications.

Presto Application Store screenshot

Thoughts about Linux marketing #1 : Linux and Netbooks

Here is a long series of articles about things that i think which could be improved concerning Mandriva PR. If possible, I will take a subject or some articles and show how Mandriva could have react or communicate about this. Today i will take about the netbook market and Mandriva situation.

As most people I think that Linux have failed concerning the Netbook market. Whereas people may consider that when Microsoft claims about the ownership of 96% of the netbook market as FUD, I still think that Linux failed to create the necessary turn-around. What’s interesting is the fact that a late player in the netbook market, Ubuntu, try to defend Linux positions on Netbook market. What can I say about all of this ?

  1. Linux marketing : in fact it’s more … Linux lack of marketing. Ever see a linux ads on TV or press magazine or in the street ? Everyday, and eventually everywhere I can see adds about Microsoft Windows or Office, or Mac products. Linux ? Nothing. At the same time we have so many « small » actors using the same product that they don’t know how to sell themselves and sell Linux.
  2. Choice : sorry, but when I’m going in the retail shop, I may have … 1 Linux netbook for 9 Windows ones. So there’s no choices. We should ask why … How do we deal with after-sales services ? The retail shops call centers don’t have the competences to debug a Linux issue, and except if this is done by the hardware manufacturer with help from the distribution, you will have a hard time. What about the documentation or quick start guide for the users ?
  3. Do we provide value-add ? free or paying games designed to run on the netbook, applications, others services, documentations or quick start guide, peripheral/accessories designed for the netbook ( webcam, DVD player/writer, DVB-T USB adapter, bluetooth adapters, … )

I looked at the arguments of Ubuntu concerning Linux vs Windows. In short we have : no virus,, Linux reactivity and stability … Sorry but this is shit ! No virus ? Frankly most users don’t really care about this argument : they just live with it. Firefox and ? They are available under Windows, and sometimes they perform better under Windows and it’s easy to install them and have the latest version. Linux stability ? Again this is not an argument. Users have friends who are going to use CCleaner, TuneUp utilities to resurrect their computer, they will have Avira AntiVir or Avast to mostly protect their computer, and they will be able to use most of their preferred application, use skype or MSN with webcam chat, play all flash games and some little games that will work on their computers. When doing a presentation and willing to use clone mode display, this will work out of the box, and wireless connection will be most of the time reliable. In short, a Netbook running Windows XP, with a friends who know how to cure your computer, is top notch and all that most users may want. So what can a Linux netbook could provide ?

  • App Store : Do people know the Apple App Store ? You know, the famous : For whatever you will want to do in your life, there will have an application in Apple App Store. What a brilliant and simple marketing idea. Do you know that Linux have « Apple App Store » like features since years ? Yeah, just start Yast/Synaptics/rpmdrake/yourpreferredpackagemanager, and you will have access to thousands of applications. Now look at the differences between linux and Apple. Apple use icons and screenshots for the applications, mix free and paying application, sort them by categories, rate them and eventually show by default the preferred application. Now look at Linux, application are sorted in sometimes obscure categories, everything is text, no image, no screenshots, and no applications are recommended. You only have Free applications, and you don’t have paying applications or demo and ability to buy them directly. One is for end user, the other one is for technical users. Linux is using ugly and obscure names like package, media, repositories, dependencies, conflict. Why can’t we have commercial paying applications shown in Linux applications list ? This channel/media could be disabled by default, but this could be interesting. Why can’t we pay directly for the application from the Linux package manager ? Mandriva have, so part of the infrastructure is available. A user open an account on The user could provide all needed informations to allow payments, and then when installing a paying application, he could just confirm the payment directly from the application, or by starting a browser on the corresponding page. Do you know that we have already something like this ? look at the Hercules eCafe, a netbook running Mandriva Linux and Windows. You have a dedicated site where you can find tutorials and documentation, but also you have access to some games specifically designed for the netbook ! With this netbook you can even have online storage
  • Ecology/GreenIT : How is Linux concerning this ? If linux based Netbooks use less power than Windows based ones, why don’t we talk about Linux as being green-friendly ? Can Linux help fighting global warming ? does Linux allow to use less boxes/packages as you download most of the stuff from internet ? What’s the advantages of intensive internet usage ? What about the development model based mostly on people working at home ( less transport usage, lower power consumption in office building ). Even for OEM they could communicate about the fact they are using a « green » OS. Is Linux a new capitalism ?
  • A new model : Windows is a whole world of compatible hardwares and softwares. Apple tell you to think different. And what about Linux ? We need serious marketing in Linux landscape … technicians/programmers are bad at marketing and without proper marketing, an unknown product can’t work and have success. Even the biggest keep doing marketing ( L’Oreal, Nike, Adidas, Coca Cola, IBM, HP, Microsoft ).
  • ARM based-netbook : really, I don’t think that ARM based netbook could be successful. I have the same kind of skepticism concerning Google Androïd. However what piss me off is the fact that when talking about ARM based netbooks, we talk about future and not yet released netbooks running Xandros or Ubuntu, and what about Mandriva Gdium. Ok, it is not ARM based, but instead MIPS based, but it is running Linux only, allow to share your documents with the USB key, and have accessories designed to run with it.
  • Windows 7 won’t be an opportunity : at the time Windows 7 will be out, Netbook will be powerful enough to make it run … As more time pass, Netbook are more expensive and more powerful. Linux price is no longer an argument, Linux need to provide values.

I guess that Mandriva needs a vision about what they want to do, what they want to show and how they want you to feel …

Dconf in GNOME 3.0 : one step further to Windows registry ?

Today, while reading Weekly Edition for April 9 concerning GNOME 3.0, I noticed the part about dconf. Dconf aims to replace … gconf already. Yeah, we can’t keep a simple technology as simple as reading and writing application configurations settings more than 8 years … Whereas I was not agree with the fact that gconf was somewhat like the Windows registry, at least it was using plain text files which could be delete or read from CLI with standard tools. However with dconf things are about to change. I don’t know if this GNOME dconf is the same than the dconf project on fd.o for which Aaron Seigo had some harsh words in 2005, but really this GNOME dconf seems just wrong. Really I do prefer gconf finally !

The big problem for me comes from this sentence from the dconf project page : Having all of the keys in a single compact binary format also avoids the intense fragmentation problems currently experienced by the tree-of-directories-of-xml-files approach. Hell yes, you read correctly : dconf will be using a single binary file … We just reinvent Windows registry. If really dconf is going to hold all the settings in one binary file, we are going to have a single point of failure. Do people remember the Ext4 data loss saga ? Now image this with the famous single binary file containing all your applications settings … You see the point ? you may end up with a zero file or corrupted configuration database. This corruption will not just impact the key being modified ( as in gconf ) or the application’s configuration file ( as in KDE ), but _all_ the applications which are storing information in dconf … Really, I don’t why they want to do this …gconf is too slow ? Why about doing the KDE way by grouping several keys in one text/xml file ? On top of that having a binary file means that users/sysadmins will no longer be able to browse or edit the keys from CLI … Really I don’t understand …


Intel Drivers Love Days : to make a better Intel eXPerience

Sorry for writing about this so late, unfortunately I’ve been very busy at the beginning of this week 🙁 So Colin Guthrie asked last week to test the Intel drivers in order to track possible regressions. An announce have been done on the Mandriva english forum too. The goal was to test the Intel 2.6.3 drivers + libdrm 2.4.5 versus the Intel drivers 2.6.99 with libdrm 2.4.6. Indeed many people report speed regressions or display corruption when using the Intel drivers. The following stuff should be checked :

  1. Check if you have 3D performances regressions with Intel 2.6.3 or Intel drivers 2.6.99
  2. Check if performances regressions can be fixed by using UXA accel method or GEM
  3. Check if 3D performances are correct with the Intel drivers, and notably test compositing and 3D effects in KDE Kwin and Compiz
  4. .

  5. Check also if Xv and video playback is working correctly at fullscreen and even with a busy system. Video playback could be test with compositing enabled or disabled.
  6. Report any issues with dual displays setup ( notably Clone mode ), and suspend/resume issues.

Users can found the Intel 2.6.99 drivers in the main/testing cooker repositories. So to install theses drivers, just enable the main testing media, and then install libdrm2, mesa-7.4 and x11-driver-video-intel- packages. Thoses wiling to test to UXA acceleration method just need to put in the Device section of /etc/X11/xorg.conf the following line : Option « AccelMethod » « UXA »

As GEM support have been integrated and enabled by default in Mandriva kernel, users willing to disable GEM can just use the following module option to disable GEM : options i915 gem_enable=0. To enable GEM, just replace 0 by 1. Users should try to disable GEM « when slow 3D performance is noticed and when the Xorg.log makes reference to problems enabling tiling due to it being rejected by the kernel ». This module option could be add at boot time, or add in a i915.conf file located in /etc/modprobe.d/ directory as follow :
echo « options i915 gem_enable=1 » > /etc/modprobe.d/i915.conf

After 1 week of tests and reports, the following issues have been fixed or are about to be resolved :

  • DRM was disabled when using speedboot : mdv bug #49490
  • Performances was reported to be better with GEM disabled under kernel for some older Intel cards
  • Intel acceleration method choice, and GEM support may be add soon in XFdrake. XFdrake will then allow to choose the acceleration method to use ( XAA ?, EXA, UXA ) and eventually enable or disable GEM support in the kernel.
  • As already spotted by Wanderlei Antonio Cavassin, glxgears can’t be used as a 3D benchmark to report 3D regression. The best thing is to use the phoronix test suite by installing the phoronix-test-suite package, and then calling the following command to start the GUI : phoronix-test-suite gui. The menu entry can be found also in the Tools -> System tools category. People willing to test 3D performances may be willing to install and run the GLmark or the Doom 3 tests. Those willing to test 2D operations and notably RENDER should use gtkperf, qgears2 and Render bench ( and eventually x11perf ). To test video playback, the « Mplayer Video playback tests » should be used.

Please note that each time you report an issue, a regression or an improvement concerning the Intel drivers, you should also give the following informations :

  • Provide the Intel and libdrm version : rpm -q x11-driver-video-intel libdrm2
  • Give your Intel graphic card version : lspcidrake | grep VGA
  • Provide the content of the Device section of your /etc/X11/xorg.conf file in order to know for example the accel method you are using
  • Notify if speedboot is enabled or not, and eventually provide relevant /var/log/Xorg.0.log content

We do hope that for the Mandriva 2009.1 Spring release, most Intel issues will be fixed and that users will have the best possible experience.


Why IBM may want to buy Sun and consequences for Mandriva

There are many rumors on the net concerning IBM buying Sun. Slashdot
today even announce the pricetag : 7 billion $. They were also many question about the motivations
behind this move. Here are my thoughts.

  1. Technology control : IBM is using many Sun-based technologies in its
    software stack : Java ( Websphere, Eclipse, … ), ( L
    otus Symphony
    ). So buying Sun will allow IBM to control directly theses
    technologies. Of course some products will eventually disappear like Star
    Office, NetBeans.
  2. Virtualization technologies for mid-servers : IBM with the mainframes have
    a long history of virtualization technologies, however theses technolgies can’t
    be used outside of the scope of the IBM mainframes. x86 based servers rely
    more on products like VMWare or Virtualbox ( eventually Xen ) for virtualization.
    With the acquisition of Sun, IBM will have the control of Virtualbox and thus may
    be less dependant from VMWare and even be in direct concurrence.
  3. low-end and mid database market : Again IBM have a super product called
    DB2 but which is not really used on mid-servers. 2 OpenSources technologies
    dominate this market : MySQL and PostgreSQL. Sun have presently the control
    of MySQL, and also employ some of the lead dev of PostgreSQL. Now maybe
    that IBM should make a choice between MySQL and PostgreSQL. Feature-wise,
    PostgreSQL is more interesting and advanced, however IBM may fear that
    PostgreSQL may become a competitor in the futur for DB2.
  4. low-end and mid-system OS : IBM with AIX control the O.S and the
    hardware for big systems ( notably mainframes, HTC market, … ). However for
    low-end and mid x86 servers, IBM is using Linux onr which it has no control.
    Red Hat will have more expertise to maintain its distribution for example, so
    someone will buy the IBM hardwxare, buy RedHat, but use Red Hat for support
    services concerning the OS. However IBM is
    one of the first IT consulting company in the world. So if IBM sell a IBM server
    running Linux, a part of the IT consulting revenue will go to the Linux
    distribution maker. By buying Sun, IBM will have also OpenSolaris : a user
    friendly x86 Unix based Unix operating system supported by a community and
    Open Source. OpenSolaris could then be used in place of Linux on IBM
    servers, and then IBM could ship its servers with OpenSolaris and control all
    the revenus stream with a complete software stack going from the operating
    system ( OpenSolaris ) to the database (DB2, MySQL ) and a language/SDK to
    develop and build the applications (
    Java, Websphere, Eclipse, … ). For this stack, the best placed to provide
    consulting services would be … IBM.

So IMHO the IBM strategy is to come back in the mass x86 market by
targeting both the x86 servers, and eventually the workstations. IBM will then
be able to be at the same time in the high-end market ( mainframes, HPC, big
big servers ), but also in the x86 mass market.

By looking at this, others vendors could be interested to do the same move,
and notably a french one … Bull. Bull provide
high-end servers with the Novascale ( Itanium based ) running Linux and
mainframes running their own Operating System ( GCOS ). Bull is involved in many
open sources projects like NFSv4,
Ext4, KDB or JOnAS. Bull is also an IT Consulting company, however they
provide third parties Linux system. What if Bull was able to provide its own Linux
distribution ? For which they could have control and provide full and complete
support ? IMHO this could be a good opportunity for Mandriva and Bull to work
closer, and especially for Bull to buy Mandriva. This will give a boost to
Mandriva credibility in the professional market, and allow Bull to ship its own
Linux. Wait & See, but I do hope this will come to reality in the future. Bull and
together is one of my dream for a first class european player in the Linux and
server market.

Here is a link to another point of view, in french, concerning
IBM/Sun by Denis Szalkowski : Quel impact du
rachat de Sun par IBM ?


Compiz is an evil dying hack

Often I said that Compiz was an evil hack as it was replacing the native and
eventually well tested desktop windowmanager with a new code unstested and
unstable. Now Compiz is at an important point of its life, and some decisions
may make it become irrelevant in the future.

After all the hype of the beginning of Compiz, their issue with Beryl fork, and
then their rebirth with Compiz Fusion, people may have think that Compiz future
would have been bright. However as Compiz dev saw that their code was
incorrect and an ugly aggregate of different codes, they decide to rewrite it for
Compiz 0.9.0
. They plan to use C++, improve the plugin system, allow it to
run without any opengl/composite, etc … Theses are sensible move, however
IMHO Compiz lost momentum and will soon became an irrelevant technology.
The reasons ?

  1. 3D compositing support in Kwin : with KDE4, Kwin now support some very
    compositing features and effects
    . With this, KDE users no longer need to
    run Compiz. WIth Kwin they have a stable 3D effects implementation,
    supporting Xrender compositing, and having the ability to detect automatically if
    the system can support compositing and then fall back gracefully in non-
    composited mode.
  2. GNOME 3.0 and future Gnome-shell. It seems that Gnome 3.0 will have a
    new gnome-shell
    where the panel and the window manager will be integrated
    . With this
    Compiz will have a hard time as compiz users may loose some GNOME
    functionnalities like the panel … On top of that Metacity + Clutter ( aka Mutter ) may
    implements their own effects thus making compiz irrelevant for Gnome

So whereas it’s sad for the compiz dev, I’m happy to see tyhat finally in the
end, hacks and workaround like Compiz will come to and end.



My Tweets