DSL Ideas and Suggestions :: Any plans for...?



I am not very familier with the workings of the kernel. Could someone explain or point me to a good site where I could see what the difference between 2.4 and 2.6 is? I have no complaints about the 2.4, and I use older hardware so I definetely wouldnt want to lose that support, but I have seen alot of posts like this and just wanted to know why so many want it.
Many of the people that want it don't have any idea why other than it's new and shiny ;) For those that do have a legitimate reason, it's probably one of the following:

Performance -- the 2.6 kernel features many improvements in process scheduling (deciding which bits of competing code get to run when). This can improve performance (or at least perceived performance) by keeping apps like media players and the desktop from being slowed down by "background" processing. Note that what you're really doing is just favoring the stuff the user "sees" and slowing down background tasks (vs. a 2.4 kernel) but that's generally desirable from the user's standpoint. 2.6 also reduces overhead for things like spawning new processes and a number of interprocess communication mechanisms which makes it highly desirable for servers.

Drivers -- some device subsystems saw significant improvements in 2.6 and though some of those have been backported to 2.4, some have not. Hotplug/coldplug is much improved in 2.6. In addition, drivers for new hardware are starting to be written exclusively for 2.6. Some get backported to 2.4, but there are no guarantees.

SMP -- symmetric multi-processing is much improved in 2.6. There are far fewer spots in the kernel where interrupts on both CPUs are disabled or resource locks taken by one CPU hold up operation on the other(s). That means better SMP performance (and also better "realtime" performance on single-cpu systems).

The tradeoff of course is size -- both in terms of "on-disk" and "in-RAM" memory usage. The benefits are real, but are of far more value to newer high-end systems than old hardware. The other issue is that though 2.6 gets extensive testing on newer systems, it doesn't see anywhere near the same level of use on old hardware. The result being that 2.4 is less likely to have problems when running on an old "untested" system than 2.6. So long as DSL continues to cater to older low-resource systems, staying with 2.4 seems like the best solution.
FYI,

Tried a DSL-like livecd with a 2.6 kernel (INSERT project) using QEMU that comes with DSL embedded.

The kernel bootup and overall response to the detection process was noticably slower than the 2.4 kernel version of INSERT or for DSL itself.
kopsis, thanks for taking the time to explain those details. I think I have a much better understanding of what is going on. From what you have told me I feel pretty satisfied sticking with 2.4
Quote (mikshaw @ Aug. 12 2005,16:53)
a 2.6.x based DSL at a size of 50mb would require a pile of extensions to return it to the useability of the current 2.4 DSL.  It would also not support a lot of the older hardware which DSL is famous for supporting.

Any you sure?  There is the 2.6-tiny project... and if KNOPPIX (have they?) has upgraded to 2.6 then their auto-detection stuff could be pulled over just like it was long ago.

Not saying this should be done, just that IF Knoppix is on 2.6 by now then it's not as complicated as everyone thinks...

Also, what hardware support was dropped from the 2.6 kernels?
Next Page...
original here.