| 1 |
pk <peterk2@××××××××.se> wrote: |
| 2 |
> |
| 3 |
> Seriously, boot-critical would be something that the system cannot *boot |
| 4 |
> without*, which belongs in /. Everything else should be in /usr, i.e. |
| 5 |
> non-boot-critical. How hard is it to start *non-boot* (system) critical |
| 6 |
> *after* boot (things like sshd)? I do that today... |
| 7 |
|
| 8 |
For somebody who uses sshfs-fuse to mount /usr from another machine, |
| 9 |
sshd and fuse *are* boot critical. (And yes, this maybe a natural |
| 10 |
setup for home systems since in many settings this is more secure |
| 11 |
than using nfs for this.) |
| 12 |
|
| 13 |
But even without net-mounting the answer to "how hard is it to start ... |
| 14 |
after boot" the answer for modern kernels is: a lot. |
| 15 |
Modern kernels initialize modules simulataneously (i.e. in an |
| 16 |
unpredictable order). So you would have to remember and postpone these |
| 17 |
initializations which can produce all sorts of unexpected problems |
| 18 |
if you have complicated implicit dependencies. |
| 19 |
Older versions of udev did this in a somewhat primitive way |
| 20 |
(restarting failed services again), but obviously this is not |
| 21 |
a clean solution (since the failing could have other reasons). |
Archives