01micko on the first page wrote:Vision
OK, alpha has been out some days now and the bugs are turning up. On to that.
The immediate plan is to compile 3.4.52 with devtmpfs and f2fs support. I'll add my f2fs usb stick program, and also write up a FULL install to f2fs-usb program. This will take full advantage of f2fs usage of nand storage. After all, whacking an ext filesystem in a save file on top of f2fs probably nulifies any advantage gained by using f2fs. For a FRUGAL type situation, we need to develop f2fs SAVE files. That's down the track. The big advantage of this will be that writes will not have to be restricted so USB installs so using AUFS will act like pupmode=12 installs (on a fast hard drive).
So why the trouble? I believe puppy's niche is USB installs. I use them extensively. Currently, the system of caching everything in RAM until a set interval (default 30 minutes) and saving back to the save file has it's limitations.This configuration is known as "pupmode=13". I don't particularly like it and I think kirk and jamesbond dumped that method in FatDog (confirmation of that please?).
- The 'snapmerge' (as it is known) can take up to several minutes and the system cops a substantial performance penalty.
- Using FAT filesystems (and NTFS for that matter) is not Linux friendly as those FS do not respect UNIX permissions
- RAM can fill up in that 30 minute period and cause RAM crash, then everything in that period is lost. Sure, that could be monitored but the systems that suffer these problems (old ones) then cop another performance penalty by adding another daemon.
***************************
I will stick with k3.10 series for the high end (PAE) version. I will wait until 3.10.1 before compiling. With the newer kernels though and an older lib base (slackware-14.0's base is approaching 12 months old, that is ancient in Linux Years) we run the risk of poor driver support in the graphics stack. A few kernel devs are also Xorg devs (Alan Cox comes to mind) so the kernel and Xorg are undergoing some parallel developments.
So there you have it, my rough plan.
Mick
--------------------------------------------------------------------------------
Woohoo! I got the f2fs patches to work cleanly and compile in k3.4.52.
Below are the patches for safe keeping.
NOTE: there is an alteration from standard to accomodate AUFS in kconfig and Makefile.
Now pick up those heads and shoulders.anikin wrote:Hi 01micko,
This one is the fastest, coolest and smoothest of all the Slackos I've tested on my eeepc (Intel atom n270/945GSE). It even feels faster than ThinSlacko. With every release, these new 3.x.x.x kernels are getting friendlier to atom processors. I'm frustrated by your decision to downgrade the kernel. And it's surprising that you have already made up your mind at this very early stage, when the testing has barely started. Really frustrated and disappointed. No peace in the valley ... deep sigh.