[klibc] Polling again: switch the list over to subscriber-only?

[maximilian attems]

On Fri, 21 Mar 2008, H. Peter Anvin wrote:

> Well, I'm not going to give up on it quite yet, but I have to admit that 
> the amount of effort I've spent on it for the last year has been limited.
> 
> I've had some discussions with some other kernel people about the 
> potential for getting it to the point where it hopefully would be 
> accepted, and there is a fair bit of interest in it.

cool.

i've seen it also been added on the TODO list of util-linux-ng. :)
what i'd like to see is to have this fstype recognotion stuff moved
over to them.
udev's lib seem to fail for all non bog standard installs.
that's why ubuntu gets away by using it for rootfs detection,
while debian relies on fstype.
 
> So the objections aren't to klibc itself, but rather to what is 
> currently available (from the kernel's perspective) on top of it.  There 
> are a lot of ideas, but several of them are extremely hard to implement 
> while still out of tree, which causes an ugly chicken-and-egg problem.
> 
> 	-hpa

i'd like to see current declared with latest little fixes declared as 1.6
and have work starting on 1.6.X based on the stdio branch.
mdadm upstream has been receptive on adding patches to have it work
against klibc, what is missing is mostly on our side.
lvm2 seems more like a closed rh circle, but i guess with more handwaving
patches will land there too.

to summarize klibc supports right now module-init-tools and udev, what
would be needed is cryptsetup, lvm2 and mdadm. (in a later run stripped
down busybox configs too, but that is less urgent)

-- 
maks

ps concerning the ml, why not move to marc?