Clair que c'est plus facile d'être simple pour un système qui ne tourne que sur les x86-32, qui ne gère pas le multicoeur, qui n'a presque aucun pilote.
C'est le reproche de Linus: l'argument de la simplicité ne tient plus quand on veut vraiment répondre aux problématiques du monde réel.
Ne pas s'emballer quand même. La partie "What we assume" de cette page reste fort conséquente ! Comme tu le dis toi-même je ne suis pas certains que cela ait un quelconque bénéfice pratique pour l'instant.
Bon je crois que c'est le moment idéal pour ressortir un vieux post que j'avais écrit début 2009. Dans ce post je me contentais de traduire l'avis de Linus sur les micro-noyaux qui avait été posté sur le site realworldtech:
En relisant ce qu'il écrit, je trouve ça particulièrement lucide et dévastateur. Sa seconde réponse en particulier ou il explique l'attirance que l'on ressent envers le concept même des micro-noyaux et qu'il s'agit en fait "d'échapper à la réalité du fait que le design et l'implémentation d'OS complet est difficile et que c'est beaucoup de travail".
J'ai l'impression qu'Andrew Tanenbaum est un peu amer du succès de Linux et qu'il se cherche des excuses pour expliquer l'absence de percée des micro-noyaux.
Dans l'interview j'ai fait exprès de reprendre presque terme pour terme une question datant d'une vieille interview de 2007: "Is MINIX being taken up in industry at present?"
En 2007 il répondait qu'il ne savait pas et que MINIX n'avait pas eu assez de temps pour convaincre. Maintenant en 2011 il dit qu'il ne sait pas et que MINIX va bientôt faire un réel effort pour convaincre.
Autrement dit rien n'a changé....Comme l'écrit Linus dans le post realworldtech:
I'm sure that ten years from now, it will be something else. There's always an excuse.
LinuxFr.org : Do you still hack on MINIX today or do you only supervise your PhD students?
Andrew Tanenbaum : I don't personally write much code, but I have 3 full-time programmers working on MINIX 3. It is being developed quite rapidly.
LinuxFr.org : Is there a MINIX community outside of the VU University Amsterdam?
Andrew Tanenbaum : Yes. See this page. The CD-ROM image page has been visited 850,000 times since the end of 2005. The number of actual downloads is about half a million. The number of visits to the minix3.org site is over 2 million. There are tiny squares at the bottom of some of the MINIX 3 Web pages that lead to traffic statistics.
LinuxFr.org : By MINIX community I mean a broad developper community. I haven't seen patches contributed to MINIX by developpers outside of VU University. Perhaps you don't buy the bazaar model of developpement? Do you think a more controlled model improve the quality of the code?
Andrew Tanenbaum : We are definitely open to outsider developers and have had code contributed by others but, for quality reasons, we want it contributed to us rather than having other people put it in trunk themselves.
LinuxFr.org : Could you elaborate a little bit about the "reincarnation server" in MINIX? If a driver is known to be malfunctioning, is it really a good idea to restart it? How does it compare to Nooks?
Andrew Tanenbaum : Most failures are due to random timing errors and races. If you run the code again, the bug will still be there of course but it is very very unlikely to manifest itself again. Now it is better to find the bug, but it is better yet to report the bug and repair the code so it can continue anyway
Nooks was a one-off attempt to write 20,000 lines of code to hand wrap the Linux drivers, but it was obsolete within a year because all the drivers changed. Besides, it was not bulletproof because any line of wrapped code could still write on any word of RAM
LinuxFr.org : Is it possible to recover from failures in stateful components?
Andrew Tanenbaum : Not yet but we are working on it
LinuxFr.org : What new features will we see in MINIX in the coming months (version 3.2) or years?
Andrew Tanenbaum : We are now focused on three things: NetBSD compatibility, embedded systems, and reliability.
3.2.0 will have a lot of headers, libraries, and userland programs take from NetBSD, which is a very stable, mature system. The BSDs are quite popular as you may know. One of them is sold under the brand name "Macintosh" by Apple.
One of the things that sets MINIX 3 apart is the ability to recover from many fatal errors that would bring down other systems. We are continually pushing the edge further in this dimension. We also are working on live update. We can now replace—on-the-fly—all of the operating system except the tiny microkernel while the system is running and without a reboot and without affecting running processes. This is not in the release yet, but we have it working. There are a lot of applications in the world that would love to run 24/7/365 and never go down, not even to upgrade to a new version of the operating system. Certainly Windows and Linux can't do this.
We are also moving into the embedded systems world. Probably 95% or more of the 32-bit computers in the world are in embedded systems. We think MINIX 3 can score well there due to its modularity, reliability, small memory footprint, and BSD license. We know that many companies find the GPL so unacceptable that they won't use Linux for that reason. In this regard we might become a small BSD-licensed Linux replacement.
LinuxFr.org : Concerning the multicore support, what is the strategy? Is it possible to have a simple and reliable OS with multicore support? What do you think about Barrelfish?
Andrew Tanenbaum : Multicore is hard. We are doing essentially the same thing as Barrelfish: we regard our 48-core Intel SCC chip as 48 separate computers that happen to be physically chose to one another, but they don't share memory. We are splitting up the file system, the network stack and maybe other pieces to run on separate cores. The focus is really more on reliability but we are starting to work seriously to look at how multicore can improve reliability (e.g., but having other cores check on what the primary cores are doing, splitting work into smaller pieces to be divided over cores, etc.)
LinuxFr.org : I've seen MINIX participate in Google Summer of Code. Is it important for the project? Was it a success?
Andrew Tanenbaum : It is not crucial, but we have had good students and they have done good work. By and large we are happy with them.
LinuxFr.org : MINIX received a €2.5 million grant from the European Research Council. Does it mean MINIX is still geared more for academic purposes than for production in the real world?
Andrew Tanenbaum : No, not at all. The ERC very much wants the results to be commercialized. In fact, I just received a second ERC grant solely for the purposes of commercializing MINIX 3. We are going to port it to the ARM and do that starting in January.
LinuxFr.org : Why porting the userland utilities from NetBSD? Is the goal to become a BSD-like system?
Andrew Tanenbaum : We think NetBSD is a mature stable system. Linux is not nearly as well written and is changing all the time. NetBSD has something like 8000 packages. That is enough for us.
LinuxFr.org : What is the performance of MINIX 3 compared to other OS? Is there benchmarks somewhere?
Andrew Tanenbaum : It is surely somewhat slower although it is hard to make measurements that mean anything as there are so many parameters you can tweak that affect performance and it matters a lot what you measure. It is my observation that OS performance really isn't an issue at all any more. If you tripled the speed of your OS, Firefox wouldn't run 1 msec faster and diskbound applications (like data bases) probably wouldn't go much faster either.
What most ordinary users want is that it ALWAYS works. In engineering terms I would say this could be expressed as mean time to failure of 50 years. I'll say systems are reliable enough when no person I know personally has ever experienced a system crash.
LinuxFr.org : Thus one of the main goals of MINIX 3 is reliability. How can we evaluate the reliability of an OS?
Andrew Tanenbaum : The metric is normally mean time to failure, but it is hard to measure. We have done numerous fault-injection tests injecting millions of faults into parts of the system by overwriting the OS binary at run time. We have had only a handful of crashes and all of those were caused ultimately by bugs in the PCI bus hardware where faulty code did something that hung the PCI bus and we had to pull the plug out to fix it.
LinuxFr.org : If you could return in the past to change the MINIX original proprietary licence to the GPL licence, do you think your system might have become the dominant free OS today?
Andrew Tanenbaum : Never. The reason MINIX 3 didn't dominate the world has to do with one mistake I made about 1992. At that time I thought BSD was going to take over the world. It was a mature and stable system. I didn't see any point in competing with it, so I focused MINIX on education. Four of the BSD guys had just formed a company to sell BSD commercially. They even had a nice phone number: 1-800-ITS-UNIX. That phone number did them and me in. AT&T sued them over the phone number and the lawsuit took 3 years to settle. That was precisely the period Linux was launched and BSD was frozen due to the lawsuit. By the time it was settled, Linux had taken off. My mistake was not to realize the lawsuit would take so long and cripple BSD. If AT&T had not brought suit (or better yet, bought BSDI), Linux would never have become popular at all and BSD would dominate the world.
Now as we are starting to go commercial, we are realizing the value of the BSD license. Many companies refuse to make major investments in modifying Linux to suit their needs if they have to give the code to their competitors. We think that the BSD license alone will be a great help to us, as well as the small size, reliability, and modularity.
LinuxFr.org : Yet we see the contributions of numerous companies during each development cycles of the Linux kernel (Intel, IBM, Novell, Google, Oracle, Broadcom, Samsung, TI, Freescale, Fujitsu, etc).
Andrew Tanenbaum : Clearly some companies are willing to do this but I also know of quite a few companies that refuse to use Linux for this reason. It depends on the company and whether it feels its contribution would help its competitors. Also it is easy to overestimate what companies have contributed. Usually it is things that benefit them directly, like making sure their drivers and programs work on Linux.
"I'm still convinced that it's one of those ideas that sounds nice on paper, but ends up being a failure in practice, because in real life the real complexity is in the interactions, not in the individual modules. And microkernels strive to make the modules more independent, making the interactions more indirect and complicated. The separation essentially ends up also cutting a lot of obvious and direct communication channels".
What do you think about this answer?
Andrew Tanenbaum : I don't buy it. He is speculating about something he knows nothing about. Our modules are extremely well defined because they run in separate address spaces. If you want to change the memory manager, only one module is affected. Changing it in Linux is far more complicated because it is all spaghetti down there.
LinuxFr.org : Do you think the Linux success is a proof he was right or is it unrelated?
Andrew Tanenbaum : No, Linux "succeeded" because BSD was frozen out of the market by AT&T at a crucial time. That's just dumb luck. Also, success is relative. I run a political website that ordinary people read. On that site statistics show that about 5% is Linux, 30% is Macintosh (which is BSD inside) and the rest is Windows. These are ordinary people, not computer geeks. I don't think of 5% as that big a success story.
LinuxFr.org : So you don't think the so-called "fairness" of the GPL ("your improvements need to be available the same way the original is available"), the bazaar model and a great project leader triggered the contributions to the Linux kernel? For you it's only dumb luck?
Andrew Tanenbaum : Yes, I think Linux succeeded against BSD, which was a stable mature system at the time simply because BSDI got stuck in a lawsuit and was effectively stopped for several years.
There are only half a dozen or so open-source software projects that dominate their markets, like Apache and gcc. Even something like Firefox, which has received millions of dollars from Google in return for the little Google search box on top, is about to drop into third place after IE and Chrome. But if the EU were to ban IE for legal reasons for 3 years, you'd be amazed how fast Firefox would gain market share. If open-source and GPL were the way to go, you'd find hundreds of open-source products that dominated their markets instead of half a dozen.
LinuxFr.org : Since 2005, MINIX 3 has been designed to be an OS for resource-limited and embedded computers and no more as a teaching tool only. Is MINIX being taken up in industry at present?
Andrew Tanenbaum : In truth, one of the problems with the BSD license is that I don't know. Companies using it don't have to tell us anything. My experience is that a lot of companies don't like to talk about what technology they are using. But we are starting to make a real effort in embedded systems now.
LinuxFr.org : The mobile/tablet space is hot these days, does MINIX aim at this kind of devices? Is there an ARM port available?
Andrew Tanenbaum : We are starting an ARM port in January
LinuxFr.org : The embedded world was formerly a niche market for microkernels but nowadays embedded systems people are moving to Linux (with Android but also others solutions like MontaVista or Wind River). Why a monolithic kernel is able to win the embedded market?
Andrew Tanenbaum : I have seen reports on market share in the embedded world. Linux isn't winning it at all. About 30% have no operating system and another 30% have a home-brew operating system. The rest is divided into many tiny segments with QNX, Green Hills, VxWorks, Linux, NetBSD and many other players having a chunk.
In fact, one of the reasons we are going into the embedded world is that there is no dominant player there. There isn't even a big player. Android is only popular on phones and tablets. Nobody is controlling his car, aircraft, or MRI scanner using it. Also noteworthy is that Google and others are putting a huge effort to remove much of the code from the Linux kernel and replace it with BSD code in userland, both to make it simpler and get rid of the GPL.
We are 5 years ahead on this.
LinuxFr.org : The two most famous microkernels nowadays are MINIX and L4. What are the differences between these two systems?
Andrew Tanenbaum : L4 is very successful. Gernot Heiser in Australia managed to sell it to Qualcomm, so it is now running on 300 million phones in the world. Not bad for a microkernel. The biggest difference is that we run the entire operating system as a collection of servers in userland. They mostly run Linux as a whole unit in userland. This means that a single bug in Linux can crash Linux. While L4 will survive this, all that does is make the reboot faster. It doesn't solve the reliability problem.
LinuxFr.org : But what about the technical differences between L4 and MINIX core code? I suppose that all microkernels are not equal?
Andrew Tanenbaum : There are lots of technical differences between MINIX and L4, but the focus of the projects is very different. We have a microkernel so we can run a multiserver system. The entire focus is on multiserver and the ability to restart OS components independently. L4 is being used to run Linux and on phones. This has nothing to do with technical differences. We could have used L4 and they could have used MINIX but it is easier to use your own since you can change it as needs arise.
LinuxFr.org : What do you think about the formal verification methods and in particular the work on seL4 ?
Andrew Tanenbaum : They certainly did a nice job but it is really hard to verify if it works since the hardware itself is poorly defined and the OS has to deal with the hardware a lot.
LinuxFr.org : Do you think a capability-based OS like Coyotos is a good solution to increase security?
Andrew Tanenbaum : Capability systems have been around for 40 years. At the recent SOSP conference, M.I.T. professor Jack Dennis got an award for inventing them in 1967. They definitely have potential. My first distributed system, Amoeba, used them. Coyotos, however, is going nowhere.
LinuxFr.org : Finaly what are your dream for MINIX? Take over the world?
Andrew Tanenbaum : I am too modest for that. One thing I would like is that when microkernels take over the world, which I fully expect, is that we at least get a footnote. Many hypervisors are getting more and more functionality over time until they are indistinguishable from a microkernel, only not designed well. With the Android people and Microsoft moving more and more code into user space, we are also moving in that direction.
Being ahead of your time is never good. I published a paper in 1978 on something very close to the Java Virtual Machine, but we never got much credit for it although we were years ahead of Sun. Such is life sometimes.
Les x86 sont les seuls a avoir du SIMD sur 256bits avec AVX (les autres CPU ne fournissent que des small vectors sur 128bits)
Ce n'est pas vrai si on parle du loogson qui est décrit comme ayant "two 256-bit vector processing units in each core".
Et puis les autres processeurs sont optimisés spécialement pour être des monstres en performance/watt sur les flottants. Les x86 d'Intel (si on met à part le futur MIC) restent des processeurs généralistes.
Et aussi, à mon humble avis, cela aurait plus de sens de commencer à compter le temps à partir du passage en modération et pas à partir de la création.
Pourtant avec le lien "page perso" juste à côté de mon pseudo ce n'est pas très dur de savoir.
et je m’en fous en fait
Ça semble en effet très raisonnable.
Ceci dit cette méprise m'a permis de fantasmer un petit peu en m'imaginant en génie du crime comme Keyser Söze ;-)
je trouve juste que c'est cocasse de faire un journal plusieurs semaines plus tard pour en remettre une couche.
Le journal est juste la pour signaler la parution d'un article qui prend du recul par rapport au flot hagiographique qui a déferlé après sa mort.
Je ne sais pas ou tu vois de la haine. Est-ce que tu as lu l'article complet au moins ?
Tu pourra y voir les déclarations de Laurene Powell qui est la veuve de Steve Jobs :
"There are parts of his life and personality that are extremely messy, and that’s the truth" Powell tells Isaacson. "You shouldn’t whitewash it."
C'est juste son ex-femme qui recommande d'éviter de passer la pommade et de dire clairement qu'il y avait des aspects peu reluisants dans la personnalité de Jobs.
Le New Yorker est d'ailleurs très célèbre pour sa prudence et ses vérifications avant toute publication d'un article.
Dans l'article Wikipédia on trouve la phrase suivante : "Au sein de la profession, les équipes de The New Yorker chargées de la correction et de la vérification des faits sont réputées pour leur rigueur."
Mais bon laisse tomber. C'est juste les moules de la tribune qui veulent s'amuser un peu à mes dépends ;-)
sa volonté de sortir un produit impeccable contraste avec ce que font beaucoup d'entreprises.
C'est quand même un truc à relativiser fortement. Il avait la volonté de sortir un truc impeccable sur le plan du look et du design ainsi que de la facilité d'utilisation.
En revanche cette exigence de qualité de s'est jamais étendue à toute le "triperie" technique non visible de l'utilisateur.
- iOS en faux multitâche pendant des années.
- Un coeur OSX tout pourri avec un Mach antédiluvien auquel on ajoute pleins de trucs histoire de cumuler les défauts des noyaux monolithiques et des micro-noyaux.
- Un système de fichier HFS+ tout vieux.
Dès qu'on creuse un peu on découvre que cette quête de la "perfection" se borne juste à la carrosserie.
Bof, ça fait sans doute du fric en plus pour la MoFo donc c'est une bonne nouvelle. J'aime autant qu'un organisme à but non lucratif qui produit du logiciel libre et qui a d'excellents principes soit bien financé.
Et puis quelle est la différence fondamentale avec Google dans la barre de recherche ?
Bon c'est un peu chiant parce qu'elle est saucissonnée entre plusieurs pages mais ça reste très intéressant.
Une citation à propos de la nouvelle manie de Microsoft de faire des procès:
Personally, I am of the opinion that going to court is a sign of a business that is dying. I am not saying that Microsoft is dying, it's a sign, not a complete indicator. But I think in this case it's somewhat true that their court and legal strategy has changed because they lost a lot of power outside the lawsuits and now they are trying to game the system and trying to stay on top. If I was Microsoft, looking at Android on cellphone and Apple on both cellphones and laptops and desktops, I would be very, very nervous, I bet they are. Nobody likes them.
peux-tu refuser/te désinscrire lorsque tu es appelé à faire un don (donc quand ce don devient urgent et nécessaire) ?
Tu peux évidemment refuser à tout moment. Même quand tu es allongé sur la table et que l'anesthésiste s'approche.
Mais bon après ça doit être dur de se regarder dans la glace et de se dire qu'on s'était engagé à aider quelqu'un mais que finalement on n'est pas allé jusqu'au bout. C'est pour ça que j'ai pris le temps de bien y penser avant de signer et que j'en ai parlé avec des proches pour avoir leur avis.
Avec les techniques d'aphérèse (cytaphérèse), plus besoin de rester hospitalisé une demie semaine et de recourir à une anesthésie générale (du moins ça devient extrêmement rare). Il suffit de quelques heures de prélèvement et de deux jours d'arrêt maladie.
Hmm...ce n'est pas ce que j'avais entendu lors de la session d'information préalable à l'inscription sur le registre. Ils ont bien insisté sur le fait que le prélèvement pouvait être fait par voie intra-osseuse avec hospitalisation à la clé. Je ne sais pas quelles sont les proportions par rapport à la cytaphérèse mais c'est une donnée à prendre en compte au moment de s'engager.
Bien entendu le truc le plus important auquel il faut réfléchir avant de signer c'est la longueur de l'engagement. On peut être appelé des années ou même des décennies après avoir accepté. Ça doit faire drôle de signer un truc à 20 ans quand on est étudiant et de recevoir un coup de fil à 45 ans quand on a une femme et trois gamins.
Forces spéciales n’est pas un film d’action (...) Le contexte est très réaliste
A noter que le réalisateur du film, Stéphane Rybojad, est connu pour avoir fait plusieurs reportages sur les forces spéciales françaises (par exemple sur la sélection pour entrer dans les commandos-marine). Il connait donc son sujet.
Une petite interview sur le blog de Jean-Dominique Merchet:
[^] # Prétérition
Posté par patrick_g (site web personnel) . En réponse à la dépêche Entretien avec Andrew Tanenbaum à propos de MINIX. Évalué à 9.
Quand à moi je ne parle pas du fait qu'il m'a renvoyé ses réponses dans un .doc alors que j'avais envoyé mes questions dans un .txt !
[^] # Re: Fiabilité
Posté par patrick_g (site web personnel) . En réponse à la dépêche Entretien avec Andrew Tanenbaum à propos de MINIX. Évalué à 9.
Clair que c'est plus facile d'être simple pour un système qui ne tourne que sur les x86-32, qui ne gère pas le multicoeur, qui n'a presque aucun pilote.
C'est le reproche de Linus: l'argument de la simplicité ne tient plus quand on veut vraiment répondre aux problématiques du monde réel.
[^] # Re: Avis de Linus Torvalds sur les micro-noyaux
Posté par patrick_g (site web personnel) . En réponse à la dépêche Entretien avec Andrew Tanenbaum à propos de MINIX. Évalué à 3.
Ne pas s'emballer quand même. La partie "What we assume" de cette page reste fort conséquente ! Comme tu le dis toi-même je ne suis pas certains que cela ait un quelconque bénéfice pratique pour l'instant.
[^] # Re: Mauvaise foi
Posté par patrick_g (site web personnel) . En réponse à la dépêche Entretien avec Andrew Tanenbaum à propos de MINIX. Évalué à 5.
http://www.electoral-vote.com/
# Avis de Linus Torvalds sur les micro-noyaux
Posté par patrick_g (site web personnel) . En réponse à la dépêche Entretien avec Andrew Tanenbaum à propos de MINIX. Évalué à 9. Dernière modification le 17 novembre 2011 à 17:09.
Bon je crois que c'est le moment idéal pour ressortir un vieux post que j'avais écrit début 2009. Dans ce post je me contentais de traduire l'avis de Linus sur les micro-noyaux qui avait été posté sur le site realworldtech:
https://linuxfr.org/nodes/24318/comments/1019366
En relisant ce qu'il écrit, je trouve ça particulièrement lucide et dévastateur. Sa seconde réponse en particulier ou il explique l'attirance que l'on ressent envers le concept même des micro-noyaux et qu'il s'agit en fait "d'échapper à la réalité du fait que le design et l'implémentation d'OS complet est difficile et que c'est beaucoup de travail".
J'ai l'impression qu'Andrew Tanenbaum est un peu amer du succès de Linux et qu'il se cherche des excuses pour expliquer l'absence de percée des micro-noyaux.
Dans l'interview j'ai fait exprès de reprendre presque terme pour terme une question datant d'une vieille interview de 2007: "Is MINIX being taken up in industry at present?"
En 2007 il répondait qu'il ne savait pas et que MINIX n'avait pas eu assez de temps pour convaincre. Maintenant en 2011 il dit qu'il ne sait pas et que MINIX va bientôt faire un réel effort pour convaincre.
Autrement dit rien n'a changé....Comme l'écrit Linus dans le post realworldtech:
# Interview with Andrew Tanenbaum: Original version
Posté par patrick_g (site web personnel) . En réponse à la dépêche Entretien avec Andrew Tanenbaum à propos de MINIX. Évalué à 10. Dernière modification le 17 novembre 2011 à 14:14.
LinuxFr.org : Do you still hack on MINIX today or do you only supervise your PhD students?
Andrew Tanenbaum : I don't personally write much code, but I have 3 full-time programmers working on MINIX 3. It is being developed quite rapidly.
LinuxFr.org : Is there a MINIX community outside of the VU University Amsterdam?
Andrew Tanenbaum : Yes. See this page. The CD-ROM image page has been visited 850,000 times since the end of 2005. The number of actual downloads is about half a million. The number of visits to the minix3.org site is over 2 million. There are tiny squares at the bottom of some of the MINIX 3 Web pages that lead to traffic statistics.
LinuxFr.org : By MINIX community I mean a broad developper community. I haven't seen patches contributed to MINIX by developpers outside of VU University. Perhaps you don't buy the bazaar model of developpement? Do you think a more controlled model improve the quality of the code?
Andrew Tanenbaum : We are definitely open to outsider developers and have had code contributed by others but, for quality reasons, we want it contributed to us rather than having other people put it in trunk themselves.
LinuxFr.org : Could you elaborate a little bit about the "reincarnation server" in MINIX? If a driver is known to be malfunctioning, is it really a good idea to restart it? How does it compare to Nooks?
Andrew Tanenbaum : Most failures are due to random timing errors and races. If you run the code again, the bug will still be there of course but it is very very unlikely to manifest itself again. Now it is better to find the bug, but it is better yet to report the bug and repair the code so it can continue anyway
Nooks was a one-off attempt to write 20,000 lines of code to hand wrap the Linux drivers, but it was obsolete within a year because all the drivers changed. Besides, it was not bulletproof because any line of wrapped code could still write on any word of RAM
LinuxFr.org : Is it possible to recover from failures in stateful components?
Andrew Tanenbaum : Not yet but we are working on it
LinuxFr.org : What new features will we see in MINIX in the coming months (version 3.2) or years?
Andrew Tanenbaum : We are now focused on three things: NetBSD compatibility, embedded systems, and reliability.
3.2.0 will have a lot of headers, libraries, and userland programs take from NetBSD, which is a very stable, mature system. The BSDs are quite popular as you may know. One of them is sold under the brand name "Macintosh" by Apple.
One of the things that sets MINIX 3 apart is the ability to recover from many fatal errors that would bring down other systems. We are continually pushing the edge further in this dimension. We also are working on live update. We can now replace—on-the-fly—all of the operating system except the tiny microkernel while the system is running and without a reboot and without affecting running processes. This is not in the release yet, but we have it working. There are a lot of applications in the world that would love to run 24/7/365 and never go down, not even to upgrade to a new version of the operating system. Certainly Windows and Linux can't do this.
We are also moving into the embedded systems world. Probably 95% or more of the 32-bit computers in the world are in embedded systems. We think MINIX 3 can score well there due to its modularity, reliability, small memory footprint, and BSD license. We know that many companies find the GPL so unacceptable that they won't use Linux for that reason. In this regard we might become a small BSD-licensed Linux replacement.
LinuxFr.org : Concerning the multicore support, what is the strategy? Is it possible to have a simple and reliable OS with multicore support? What do you think about Barrelfish?
Andrew Tanenbaum : Multicore is hard. We are doing essentially the same thing as Barrelfish: we regard our 48-core Intel SCC chip as 48 separate computers that happen to be physically chose to one another, but they don't share memory. We are splitting up the file system, the network stack and maybe other pieces to run on separate cores. The focus is really more on reliability but we are starting to work seriously to look at how multicore can improve reliability (e.g., but having other cores check on what the primary cores are doing, splitting work into smaller pieces to be divided over cores, etc.)
LinuxFr.org : I've seen MINIX participate in Google Summer of Code. Is it important for the project? Was it a success?
Andrew Tanenbaum : It is not crucial, but we have had good students and they have done good work. By and large we are happy with them.
LinuxFr.org : MINIX received a €2.5 million grant from the European Research Council. Does it mean MINIX is still geared more for academic purposes than for production in the real world?
Andrew Tanenbaum : No, not at all. The ERC very much wants the results to be commercialized. In fact, I just received a second ERC grant solely for the purposes of commercializing MINIX 3. We are going to port it to the ARM and do that starting in January.
LinuxFr.org : Why porting the userland utilities from NetBSD? Is the goal to become a BSD-like system?
Andrew Tanenbaum : We think NetBSD is a mature stable system. Linux is not nearly as well written and is changing all the time. NetBSD has something like 8000 packages. That is enough for us.
LinuxFr.org : What is the performance of MINIX 3 compared to other OS? Is there benchmarks somewhere?
Andrew Tanenbaum : It is surely somewhat slower although it is hard to make measurements that mean anything as there are so many parameters you can tweak that affect performance and it matters a lot what you measure. It is my observation that OS performance really isn't an issue at all any more. If you tripled the speed of your OS, Firefox wouldn't run 1 msec faster and diskbound applications (like data bases) probably wouldn't go much faster either.
What most ordinary users want is that it ALWAYS works. In engineering terms I would say this could be expressed as mean time to failure of 50 years. I'll say systems are reliable enough when no person I know personally has ever experienced a system crash.
LinuxFr.org : Thus one of the main goals of MINIX 3 is reliability. How can we evaluate the reliability of an OS?
Andrew Tanenbaum : The metric is normally mean time to failure, but it is hard to measure. We have done numerous fault-injection tests injecting millions of faults into parts of the system by overwriting the OS binary at run time. We have had only a handful of crashes and all of those were caused ultimately by bugs in the PCI bus hardware where faulty code did something that hung the PCI bus and we had to pull the plug out to fix it.
LinuxFr.org : If you could return in the past to change the MINIX original proprietary licence to the GPL licence, do you think your system might have become the dominant free OS today?
Andrew Tanenbaum : Never. The reason MINIX 3 didn't dominate the world has to do with one mistake I made about 1992. At that time I thought BSD was going to take over the world. It was a mature and stable system. I didn't see any point in competing with it, so I focused MINIX on education. Four of the BSD guys had just formed a company to sell BSD commercially. They even had a nice phone number: 1-800-ITS-UNIX. That phone number did them and me in. AT&T sued them over the phone number and the lawsuit took 3 years to settle. That was precisely the period Linux was launched and BSD was frozen due to the lawsuit. By the time it was settled, Linux had taken off. My mistake was not to realize the lawsuit would take so long and cripple BSD. If AT&T had not brought suit (or better yet, bought BSDI), Linux would never have become popular at all and BSD would dominate the world.
Now as we are starting to go commercial, we are realizing the value of the BSD license. Many companies refuse to make major investments in modifying Linux to suit their needs if they have to give the code to their competitors. We think that the BSD license alone will be a great help to us, as well as the small size, reliability, and modularity.
LinuxFr.org : Yet we see the contributions of numerous companies during each development cycles of the Linux kernel (Intel, IBM, Novell, Google, Oracle, Broadcom, Samsung, TI, Freescale, Fujitsu, etc).
Andrew Tanenbaum : Clearly some companies are willing to do this but I also know of quite a few companies that refuse to use Linux for this reason. It depends on the company and whether it feels its contribution would help its competitors. Also it is easy to overestimate what companies have contributed. Usually it is things that benefit them directly, like making sure their drivers and programs work on Linux.
LinuxFr.org : In this LinuxFR interview with Linus (http://linuxfr.org/nodes/85904/comments/1230981) he gave his opinion on microkernels:
What do you think about this answer?
Andrew Tanenbaum : I don't buy it. He is speculating about something he knows nothing about. Our modules are extremely well defined because they run in separate address spaces. If you want to change the memory manager, only one module is affected. Changing it in Linux is far more complicated because it is all spaghetti down there.
LinuxFr.org : Do you think the Linux success is a proof he was right or is it unrelated?
Andrew Tanenbaum : No, Linux "succeeded" because BSD was frozen out of the market by AT&T at a crucial time. That's just dumb luck. Also, success is relative. I run a political website that ordinary people read. On that site statistics show that about 5% is Linux, 30% is Macintosh (which is BSD inside) and the rest is Windows. These are ordinary people, not computer geeks. I don't think of 5% as that big a success story.
LinuxFr.org : So you don't think the so-called "fairness" of the GPL ("your improvements need to be available the same way the original is available"), the bazaar model and a great project leader triggered the contributions to the Linux kernel? For you it's only dumb luck?
Andrew Tanenbaum : Yes, I think Linux succeeded against BSD, which was a stable mature system at the time simply because BSDI got stuck in a lawsuit and was effectively stopped for several years.
There are only half a dozen or so open-source software projects that dominate their markets, like Apache and gcc. Even something like Firefox, which has received millions of dollars from Google in return for the little Google search box on top, is about to drop into third place after IE and Chrome. But if the EU were to ban IE for legal reasons for 3 years, you'd be amazed how fast Firefox would gain market share. If open-source and GPL were the way to go, you'd find hundreds of open-source products that dominated their markets instead of half a dozen.
LinuxFr.org : Since 2005, MINIX 3 has been designed to be an OS for resource-limited and embedded computers and no more as a teaching tool only. Is MINIX being taken up in industry at present?
Andrew Tanenbaum : In truth, one of the problems with the BSD license is that I don't know. Companies using it don't have to tell us anything. My experience is that a lot of companies don't like to talk about what technology they are using. But we are starting to make a real effort in embedded systems now.
LinuxFr.org : The mobile/tablet space is hot these days, does MINIX aim at this kind of devices? Is there an ARM port available?
Andrew Tanenbaum : We are starting an ARM port in January
LinuxFr.org : The embedded world was formerly a niche market for microkernels but nowadays embedded systems people are moving to Linux (with Android but also others solutions like MontaVista or Wind River). Why a monolithic kernel is able to win the embedded market?
Andrew Tanenbaum : I have seen reports on market share in the embedded world. Linux isn't winning it at all. About 30% have no operating system and another 30% have a home-brew operating system. The rest is divided into many tiny segments with QNX, Green Hills, VxWorks, Linux, NetBSD and many other players having a chunk.
In fact, one of the reasons we are going into the embedded world is that there is no dominant player there. There isn't even a big player. Android is only popular on phones and tablets. Nobody is controlling his car, aircraft, or MRI scanner using it. Also noteworthy is that Google and others are putting a huge effort to remove much of the code from the Linux kernel and replace it with BSD code in userland, both to make it simpler and get rid of the GPL.
We are 5 years ahead on this.
LinuxFr.org : The two most famous microkernels nowadays are MINIX and L4. What are the differences between these two systems?
Andrew Tanenbaum : L4 is very successful. Gernot Heiser in Australia managed to sell it to Qualcomm, so it is now running on 300 million phones in the world. Not bad for a microkernel. The biggest difference is that we run the entire operating system as a collection of servers in userland. They mostly run Linux as a whole unit in userland. This means that a single bug in Linux can crash Linux. While L4 will survive this, all that does is make the reboot faster. It doesn't solve the reliability problem.
LinuxFr.org : But what about the technical differences between L4 and MINIX core code? I suppose that all microkernels are not equal?
Andrew Tanenbaum : There are lots of technical differences between MINIX and L4, but the focus of the projects is very different. We have a microkernel so we can run a multiserver system. The entire focus is on multiserver and the ability to restart OS components independently. L4 is being used to run Linux and on phones. This has nothing to do with technical differences. We could have used L4 and they could have used MINIX but it is easier to use your own since you can change it as needs arise.
LinuxFr.org : What do you think about the formal verification methods and in particular the work on seL4 ?
Andrew Tanenbaum : They certainly did a nice job but it is really hard to verify if it works since the hardware itself is poorly defined and the OS has to deal with the hardware a lot.
LinuxFr.org : Do you think a capability-based OS like Coyotos is a good solution to increase security?
Andrew Tanenbaum : Capability systems have been around for 40 years. At the recent SOSP conference, M.I.T. professor Jack Dennis got an award for inventing them in 1967. They definitely have potential. My first distributed system, Amoeba, used them. Coyotos, however, is going nowhere.
LinuxFr.org : Finaly what are your dream for MINIX? Take over the world?
Andrew Tanenbaum : I am too modest for that. One thing I would like is that when microkernels take over the world, which I fully expect, is that we at least get a footnote. Many hypervisors are getting more and more functionality over time until they are indistinguishable from a microkernel, only not designed well. With the Android people and Microsoft moving more and more code into user space, we are also moving in that direction.
Being ahead of your time is never good. I published a paper in 1978 on something very close to the Java Virtual Machine, but we never got much credit for it although we were years ahead of Sun. Such is life sometimes.
[^] # Re: AVX un futur atout pour les x86?
Posté par patrick_g (site web personnel) . En réponse à la dépêche Le Top 500 de novembre 2011. Évalué à 5. Dernière modification le 15 novembre 2011 à 11:35.
Ce n'est pas vrai si on parle du loogson qui est décrit comme ayant "two 256-bit vector processing units in each core".
Et puis les autres processeurs sont optimisés spécialement pour être des monstres en performance/watt sur les flottants. Les x86 d'Intel (si on met à part le futur MIC) restent des processeurs généralistes.
[^] # Re: Slogan vaseux
Posté par patrick_g (site web personnel) . En réponse à la dépêche Flash Player sur mobile est gelé : HTML 5 a gagné !. Évalué à 4.
Enjoy : http://www.geekculture.com/joyoftech/joyarchives/1615.html
# Modération
Posté par patrick_g (site web personnel) . En réponse à l’entrée du suivi Nouvelles stats pour la partie modération. Évalué à 5 (+0/-0).
Et aussi, à mon humble avis, cela aurait plus de sens de commencer à compter le temps à partir du passage en modération et pas à partir de la création.
[^] # Re: Jobs est cité dans le livre de management
Posté par patrick_g (site web personnel) . En réponse au journal Steve Jobs décortiqué dans le New Yorker. Évalué à 5.
Pourtant avec le lien "page perso" juste à côté de mon pseudo ce n'est pas très dur de savoir.
Ça semble en effet très raisonnable.
Ceci dit cette méprise m'a permis de fantasmer un petit peu en m'imaginant en génie du crime comme Keyser Söze ;-)
[^] # Re: Juste par curiosité
Posté par patrick_g (site web personnel) . En réponse au journal Quand la directive Guéant commence à me les briser menues.... Évalué à 5.
Ça doit être Bull je pense. En France je ne vois qu'eux.
[^] # Re: Garbage collector
Posté par patrick_g (site web personnel) . En réponse au journal Steve Jobs décortiqué dans le New Yorker. Évalué à 10.
Le journal est juste la pour signaler la parution d'un article qui prend du recul par rapport au flot hagiographique qui a déferlé après sa mort.
Je ne sais pas ou tu vois de la haine. Est-ce que tu as lu l'article complet au moins ?
Tu pourra y voir les déclarations de Laurene Powell qui est la veuve de Steve Jobs :
C'est juste son ex-femme qui recommande d'éviter de passer la pommade et de dire clairement qu'il y avait des aspects peu reluisants dans la personnalité de Jobs.
[^] # Re: Garbage collector
Posté par patrick_g (site web personnel) . En réponse au journal Steve Jobs décortiqué dans le New Yorker. Évalué à 9.
Le New Yorker est d'ailleurs très célèbre pour sa prudence et ses vérifications avant toute publication d'un article.
Dans l'article Wikipédia on trouve la phrase suivante : "Au sein de la profession, les équipes de The New Yorker chargées de la correction et de la vérification des faits sont réputées pour leur rigueur."
Mais bon laisse tomber. C'est juste les moules de la tribune qui veulent s'amuser un peu à mes dépends ;-)
[^] # Re: Jobs est cité dans le livre de management
Posté par patrick_g (site web personnel) . En réponse au journal Steve Jobs décortiqué dans le New Yorker. Évalué à 10.
C'est quand même un truc à relativiser fortement. Il avait la volonté de sortir un truc impeccable sur le plan du look et du design ainsi que de la facilité d'utilisation.
En revanche cette exigence de qualité de s'est jamais étendue à toute le "triperie" technique non visible de l'utilisateur.
- iOS en faux multitâche pendant des années.
- Un coeur OSX tout pourri avec un Mach antédiluvien auquel on ajoute pleins de trucs histoire de cumuler les défauts des noyaux monolithiques et des micro-noyaux.
- Un système de fichier HFS+ tout vieux.
Dès qu'on creuse un peu on découvre que cette quête de la "perfection" se borne juste à la carrosserie.
[^] # Re: Twitter
Posté par patrick_g (site web personnel) . En réponse à la dépêche Firefox 8 est disponible. Évalué à 10.
Bof, ça fait sans doute du fric en plus pour la MoFo donc c'est une bonne nouvelle. J'aime autant qu'un organisme à but non lucratif qui produit du logiciel libre et qui a d'excellents principes soit bien financé.
Et puis quelle est la différence fondamentale avec Google dans la barre de recherche ?
# Interview de Linus Torvalds sur mukware.com
Posté par patrick_g (site web personnel) . En réponse au journal Linux 3.2 en route. Évalué à 10.
Je profite de ce journal pour signaler une interview de Linus ici : http://www.muktware.com/news/2855
Bon c'est un peu chiant parce qu'elle est saucissonnée entre plusieurs pages mais ça reste très intéressant.
Une citation à propos de la nouvelle manie de Microsoft de faire des procès:
[^] # Re: Don de moelle osseuse
Posté par patrick_g (site web personnel) . En réponse au journal Don de soi. Évalué à 4.
Tu peux évidemment refuser à tout moment. Même quand tu es allongé sur la table et que l'anesthésiste s'approche.
Mais bon après ça doit être dur de se regarder dans la glace et de se dire qu'on s'était engagé à aider quelqu'un mais que finalement on n'est pas allé jusqu'au bout. C'est pour ça que j'ai pris le temps de bien y penser avant de signer et que j'en ai parlé avec des proches pour avoir leur avis.
[^] # Re: détails : performance des cpu
Posté par patrick_g (site web personnel) . En réponse au journal Fujitsu atteint la première place du top500. Évalué à 5.
J'avais posté des liens vers les pdf des specs du processeur dans ma dernière news top500.
# Don de moelle osseuse
Posté par patrick_g (site web personnel) . En réponse au journal Don de soi. Évalué à 3.
Hmm...ce n'est pas ce que j'avais entendu lors de la session d'information préalable à l'inscription sur le registre. Ils ont bien insisté sur le fait que le prélèvement pouvait être fait par voie intra-osseuse avec hospitalisation à la clé. Je ne sais pas quelles sont les proportions par rapport à la cytaphérèse mais c'est une donnée à prendre en compte au moment de s'engager.
Bien entendu le truc le plus important auquel il faut réfléchir avant de signer c'est la longueur de l'engagement. On peut être appelé des années ou même des décennies après avoir accepté. Ça doit faire drôle de signer un truc à 20 ans quand on est étudiant et de recevoir un coup de fil à 45 ans quand on a une femme et trois gamins.
# Forces spéciales
Posté par patrick_g (site web personnel) . En réponse à la dépêche Intouchables, Forces Spéciales et Polisse : le cinéma français se porte bien. Évalué à 7.
A noter que le réalisateur du film, Stéphane Rybojad, est connu pour avoir fait plusieurs reportages sur les forces spéciales françaises (par exemple sur la sélection pour entrer dans les commandos-marine). Il connait donc son sujet.
Une petite interview sur le blog de Jean-Dominique Merchet:
http://www.marianne2.fr/blogsecretdefense/Stephane-Rybojad-On-me-reproche-de-ne-pas-avoir-fait-un-film-a-charge-contre-l-armee_a419.html
[^] # Re: C'est fameux
Posté par patrick_g (site web personnel) . En réponse au journal Journal bookmark The Stallman Dialogues. Évalué à 7.
En fait la phrase la plus drôle de toute c'est celle-ci:
Putain ça doit pas être facile à organiser ça :-)
[^] # Re: C'est fameux
Posté par patrick_g (site web personnel) . En réponse au journal Journal bookmark The Stallman Dialogues. Évalué à 4.
Activiste politique jusqu'au bout des ongles !
# Hop, hop, hop !
Posté par patrick_g (site web personnel) . En réponse au journal Un an de projet libre : bilan. Évalué à 10.
Tu t'es trahi ;-)
Fallait modifier la phrase avant de poster ce journal car, là, ça incite à chercher l'origine du copier/coller : http://gelnior.wordpress.com/2011/10/24/un-an-de-projet-libre-bilan/
Ceci dit, merci quand même parce que le journal est intéressant et que le projet à l'air sympa.
[^] # Re: Petites erreurs dans le descriptif sur Nouveau
Posté par patrick_g (site web personnel) . En réponse à la dépêche Sortie du noyau Linux 3.1. Évalué à 5.
Merci beaucoup pour tes précisions.
# TOMOYO
Posté par patrick_g (site web personnel) . En réponse au journal Choisir un module de sécurité linux. Évalué à 6. Dernière modification le 25 octobre 2011 à 13:23.
Pour TOMOYO il y a des live-CD et des tutos très bien foutus qui permettent de jouer avec pour voir à quoi ça ressemble.
Tutorial d'utilisation du Live-CD Ubuntu 10.04 : http://tomoyo.sourceforge.jp/1.8/ubuntu10.04-live.html.en
Tutorial d'utilisation du Live-CD CentOS 6 : http://tomoyo.sourceforge.jp/1.8/centos6-live.html.en
Un volontaire pour écrire une dépêche sur l'utilisation de ce module ? Avec le live-CD cela devrait être assez simple.