Re: [gentoo-user] Re: USE flag 'split-usr' is now global

[Kai Peter <kp@lists.openqmail.org>]

On 2019-08-08 09:43, Neil Bothwick wrote:
> On Wed, 07 Aug 2019 14:24:22 +0100, Mick wrote:
> 
>> > As opposed to splitting binaries across four directories based on
>> > arbitrary decisions made in the last century? :P
>> 
>> LOL!  You're missing the most important part:  across different fs and
>> partition layouts.
>> 
>> Look, the pyramids were built before the last century, but that 
>> doesn't
>> mean we should try to balance them upside down if they work fine as
>> they are.
> 
> Why not? As long as it's optional and controlled by a USE flag :)
> 
>> Clearly different use cases have different requirements and
>> correspondingly different optimal solutions.  Can I please keep my
>> bin/sbin directories separate and ditto for /usr and its subbies?  :-)
> 
> The /usr / split makes sense when you want different filesystems,
> something I used to do but don't have any need for now. I've yet to see 
> a
> convincing argument for the bin/sbin split in either location.

Let me jump back into this hi-jacked thread somewhere. Unfortunately I 
didn't found an answer to the future directions at gentoo regarding the 
usr merge. And my fear is that the merge will be forced. As I use gentoo 
as a meta distro really, this change _could_ be put _me_ into trouble. 
But my 'gentoo-lfs' is not the typical use case.

Now, even that there are my individual requirements behind, some short 
points of my POV to some points in the whole thread. All IMHO.

An initramfs is a elegant and - more important - a very flexible 
solution. It is not required, but it makes things easier. It could solve 
more things than having /usr at a separate fs or loading drivers. I 
would also define it as a bootloader (like Rich) in a chain. More 
abstract I would see a classical bootloader like grub as a kernel by 
itself.

I see 3 stages of security concerns of an initramfs:

1. trust yourself and build your own one
2. trust gentoo and use gentoo's initramfs
3. download one (from a warez site) and stay hacked

The usr merge itself is questionable workaround to consolidate things. 
Now, a consolidation is a good thing in general. But what is the real 
difference to have a symlink /bin against having a folder? I couldn't 
follow the given arguments.

The core issue is the under laying folder structure. And depending on 
this hard coded path's. We still relying at a 50 years old concept - and 
time was moving on. Don't ask, I haven't a solution (yet). Just the 
dream/vision of the perfect system :). One point of the vision is to 
have a core (linux) OS (e.g. an extended stage3) separated from all user 
programs (similar to a ring0 kernel isolation).

 From all exiting concepts and implementations the best parts should be 
used. Not sure if this is ever possible. But forcing users to use a 
solution which is not required by 99% of them is a very bad thing. These 
'1-percent-solutions' have to be optional.

Anyway, just an opinion beside the mainstream. I encourage people to 
have their own. :) It is not hard to create pro and con arguments. And I 
left enough room for misunderstandings.


-- 
Sent with eQmail-1.10.3 beta - a fork of djb's famous qmail