On Thursday 31 January 2008, Mateusz Mierzwinski wrote: > Talking about modularize kernel i think this is an gentoo mailing > list so every user know's his hardware - if not there is always > GOOGLE, Gentoo HowTo and Hardware Manual. Most drivers in kernel are > universal for one vendor family what makes more suitable to different > types of chipsets (revisions A, B etc...). There is also true that > maybee kernel modules are good for people with binary distro's but > Gentoo is source based distribution - thank god - and every user > should compile kernel for his hardware - modules not needed.
Rubbish. Let's say tomorrow I plug in a USB sound card, joystick and HSDPA modem. Today I do not have this hardware. Should I rebuild my kernel just to use a hotplug device that I borrowed for a few hours? No, thanks, I'm going to use modrobe. To get my sound card to work, I need a parameter "dell=m42". How should I easily pass this argument without modules? Should I have a webcam driver permanently loaded in kernel space just for the odd case where I decide to use it? 1995 called, they say they want their hardware back. > Cheap > code modules are also bad rule of cheap programmers, which don't know > system and kernel structures. Afterwords thats how making usage of > NDISWRAPPER is fundamental on Windows drivers hardware. <sigh> If a crap programmer writes a module, it will be crap and do $BAD_STUFF. How does this change if the crap programmer is forced to not write modules? Does he suddenly get enlightened and know what K&R have been telling him for years? CRAP PROGRAMMERS WRITE CRAP CODE. MODULES ARE COMPLETELY IRRELEVANT TO THIS. > If we speak about realtime preemption model i think that You are > mistaken saying that PC and realtime kernels (software) is not good > choice. My licentiate work on University of Silesia (Poland, > Katowice) is about usage of realtime services in computer LAN/WAN > networks. I digging some materials about RTOS and realtime preemption > model, realtime schedule algorithm and realtime applications critical > points programming. I don't know if PC + Realtime preemption model is > something wrong. When You need critical services for network such as > multiplexed SDH traffic control and violation prevention You must > have great power computer with RTOS, that can monitor min. 166MB/s > traffic full duplex. Now-days computers have enough power to stand > with RISC (Reduced Instruction Set Computing) machines - thats why > Sun Solaris has arrived on PC's. Another big step is RTLinux with > dual core - realtime core and Linux kernel working together. That type of usage is not my area of expertise, but I can tell that it's a niche market. If monolithicality is the correct design paradigm there, then the designer has the option of building a monolithic kernel. If you can coerce it to work on Intel cpus, well that's fine and dandy and attests to the power and adaptibility of Linux. But how does this support your assertion that modules are a bad idea? You have the choice to do it a better way in those circumstances. Meanwhile, the vast majority of server nd desktop deployments out there that truly do need kernel modules (including Gentoo) cna and should continue to use them. -- Alan McKinnon alan dot mckinnon at gmail dot com -- gentoo-user@lists.gentoo.org mailing list
