On Wednesday 14 March 2007 01:21:03 pm Anders Johansson wrote: > On Wednesday 14 March 2007 14:45, Kai Ponte wrote: > > On Wednesday 14 March 2007 01:05:26 am Marcus Meissner wrote: > > > > Someone at novel must have been able to predict or even after the > > > > fact see that removing smbfs AND usbfs support from the kernel is > > > > bound to lead into at least some 10.2 sales losses. The fact that a > > > > recompile is required is guarranteed to turn some customers away and > > > > both file systems are needed for basic functionality in some major > > > > applications. > > > > > > "sales losses" are quite difficult for a product that is mostly > > > downloaded. > > > > > > The next 10.2 kernel update will include USBFS again btw. > > > > Just out of curiosity - why are some items called "kernel modules" and > > others not? > > Because some items are modules that plug into the kernel and some aren't? > > The kernel runs with complete access to hardware, it can (with some > limitations) address memory that user space applications can't. Some things > need that while others don't. In the case of samba, there are user space > versions. You really only need the kernel support when you want to "mount" > a samba share and make it part of the linux virtual file system
Oh, okay. So those modules which want (or need) access to the hardware need to be compiled into the kernel? Now, wouldn't they use the HAL to get at the hardware? Just curious. > > > For example, I often use the Cisco VPN client to telecommute. It seems > > that every few weeks - I guess when kernel update happens - the client > > fails. I'm then forced to recompile. > > In the case of a vpn client I would tend to agree, I see very little reason > for placing that inside the kernel > > Incidentally (but I guess you know this) there is an ancient debate over > the relative merits of monolithic (i.e. do everything in kernel space) vs. > micro-kernels (kernels that do as little as possible in kernel space) Yep! In fact there's a beauty of a post with Mr. Torvalds arguing about it with Andy Tannenbaum back in '91. I remember reading it in the late '90s when I first switched to Linux on a few desktops and then my P133 laptop. http://groups.google.com/group/comp.os.minix/browse_frm/thread/c25870d7a41696d2/3f6b594a5b4eccb4?lnk=st&q=&rnum=1#3f6b594a5b4eccb4 http://tinyurl.com/34ojmg > > It usually ends up being a question of performance. For design, micro > kernels usually win hands down, while for performance, the micro kernels > haven't even reached the starting line when the monolithic kernels drink > the victory champagne Yeah, it kind of reminds me of the old relational vs. hierarchical debate in DBM systems. -- k -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
